wip(ci): point sharded acceptance tests at the canonical executable with QA_TUI_BIN

Co-Authored-By: Claude Sonnet 5.5 <[email protected]>

.github/AUTHOR_MAP

@@ -4,11 +4,12 @@
 #   alias = Display Name <[email protected]>
 #
 # The right-hand side must use GitHub's numeric noreply address so a harvested
-# commit authored with it links to the contributor's account. Only commit
-# authorship puts someone in the repository's contributor list; a co-author
-# trailer shows on their profile but not there. The left-hand side may be a
-# GitHub login, old-style noreply address, raw email from a contributor commit,
-# or local machine email seen in older harvested history.
+# commit authored with it, or a Co-authored-by trailer in the commit's final
+# trailer block, links to the contributor's account. GitHub resolves the
+# address by its numeric id, so the id must be the contributor's own
+# (`gh api users/<login> --jq .id`). The left-hand side may be a GitHub login,
+# old-style noreply address, raw email from a contributor commit, or local
+# machine email seen in older harvested history.
 
 hmbown = Hmbown <[email protected]>
 [email protected] = Hmbown <[email protected]>
@@ -50,6 +51,7 @@ Oliver-ZPLiu = Oliver-ZPLiu <[email protected]>
 idling11 = idling11 <[email protected]>
 Hanmiao Li = idling11 <[email protected]>
 [email protected] = idling11 <[email protected]>
+894876246 = idling11 <[email protected]>
 angziii = angziii <[email protected]>
 aboimpinto = aboimpinto <[email protected]>
 Paulo Aboim Pinto = aboimpinto <[email protected]>
@@ -79,19 +81,24 @@ Vince = Liu-Vince <[email protected]>
 [email protected] = Liu-Vince <[email protected]>
 LI-Jialu = LI-Jialu <[email protected]>
 JiarenWang = JiarenWang <[email protected]>
-wdw8276 = wdw8276 <3972439+wdw8276@users.noreply.github.com>
+wdw8276 = kitty <3972439+mars-base@users.noreply.github.com>
+mars-base = kitty <[email protected]>
+[email protected] = kitty <[email protected]>
 pengyou200902 = pengyou200902 <[email protected]>
 linzhiqin2003 = linzhiqin2003 <[email protected]>
 LING71671 = LING71671 <[email protected]>
+[email protected] = LING71671 <[email protected]>
 JasonOA888 = JasonOA888 <[email protected]>
 Inference1 = Inference1 <[email protected]>
 hongqitai = hongqitai <[email protected]>
 gordonlu = gordonlu <[email protected]>
-angus-guo = gus <217034332+angus-[email protected]>
-gus.guo@tec-do.com = gus <217034332+angus-[email protected]>
+angus-guo = gus <217034332+im[email protected]>
+imgusguo = gus <217034332+im[email protected]>
+[email protected] = gus <[email protected]>
 gaord = gaord <[email protected]>
 Ben Gao = gaord <[email protected]>
 [email protected] = gaord <[email protected]>
+bengao168 = gaord <[email protected]>
 MXAntian = MXAntian <[email protected]>
 [email protected] = MXAntian <[email protected]>
 DarrellThomas = Darrell Thomas <[email protected]>
@@ -247,6 +254,7 @@ [email protected] = zhuowp <[email protected]>
 [email protected] = 秋月凉梦 <[email protected]>
 
 asto18089 = asto18089 <[email protected]>
+[email protected] = asto18089 <[email protected]>
 xiechimon = Xmon Dai <[email protected]>
 yrk111222 = Rongkun Yan <[email protected]>
 VincentCorleone = Vincent Corleone <[email protected]>
@@ -278,3 +286,38 @@ Felixkw12 = Felix Wunderlich <[email protected]>
 [email protected] = Felix Wunderlich <[email protected]>
 Felix Wunderlich = Felix Wunderlich <[email protected]>
 [email protected] = luochen211 <[email protected]>
+
+AiurArtanis = AiurArtanis <[email protected]>
+[email protected] = AiurArtanis <[email protected]>
+Apeiron0w0 = Apeiron <[email protected]>
+[email protected] = Apeiron <[email protected]>
+[email protected] = ChaceLyee2101 <[email protected]>
+AmosLeeeeeeee = ChaceLyee2101 <[email protected]>
+ddlw = ddlw <[email protected]>
+mo-vic = movic <[email protected]>
+[email protected] = movic <[email protected]>
+[email protected] = merchloubna70-dot <[email protected]>
+[email protected] = merchloubna70-dot <[email protected]>
+xulongzhe = luola <[email protected]>
+[email protected] = luola <[email protected]>
+ZhulongNT = SONG ZHENRUI <[email protected]>
+[email protected] = SONG ZHENRUI <[email protected]>
+xiluoduyu = xiluoduyu <[email protected]>
+[email protected] = xiluoduyu <[email protected]>
+wlon = Stephen Xu <[email protected]>
+[email protected] = Stephen Xu <[email protected]>
+krisclarkdev = Kristopher Clark <[email protected]>
+THatch26 = THatch26 <[email protected]>
+jeoor = Kayro <[email protected]>
+harvey2011888 = harvey2011888 <[email protected]>
+duanchao-lab = DC <[email protected]>
+dumbjack = dumbjack <[email protected]>
+[email protected] = dumbjack <[email protected]>
+AquaZhang = IceTear <[email protected]>
+disappear00 = idontknow <[email protected]>
+namezzb = sailorZ <[email protected]>
+alphapku-ux = Alpha Wu <[email protected]>
+BX166 = BX166 <[email protected]>
+c020627 = c020627 <[email protected]>
+[email protected] = Parman Mohammadalizadeh <[email protected]>
+Parman Mohammadalizadeh = Parman Mohammadalizadeh <[email protected]>

.github/scripts/release-workflows.test.js

@@ -32,6 +32,7 @@ function namedStep(source, name) {
 }
 
 const ci = read(".github/workflows/ci.yml");
+const ciSuite = read(".github/workflows/ci-test-suite.yml");
 const nightly = read(".github/workflows/nightly.yml");
 const candidate = read(".github/workflows/release-candidate.yml");
 const artifacts = read(".github/workflows/release-artifacts.yml");
@@ -46,28 +47,212 @@ const archiveInstaller = read("scripts/release/install.sh");
 const cliDispatcher = read("crates/cli/src/lib.rs");
 const runbook = read("docs/RELEASE_RUNBOOK.md");
 
-const ciTestJob = ci.slice(ci.indexOf("\n  test:\n"));
-assert.match(
-  ciTestJob,
-  /matrix\.os == 'ubuntu-latest'.*github\.event_name == 'pull_request'/,
-  "heavy pull requests must select the real Ubuntu test lane",
-);
-assert.match(
-  ciTestJob,
-  /cargo nextest run --workspace --all-features --locked --profile ci/,
-  "the Ubuntu pull-request lane must run workspace nextest",
-);
-assert.match(
-  ciTestJob,
-  /name: Linux test location \(CNB\)/,
-  "the non-PR CNB fallback must be named explicitly",
-);
+function jobBlock(source, job) {
+  const match = source.match(new RegExp(`^  ${job}:\\n([\\s\\S]*?)(?=^  \\S|(?![\\s\\S]))`, "m"));
+  assert.ok(match, `missing job: ${job}`);
+  return match[1];
+}
 
+function jobIf(source, job) {
+  const lines = jobBlock(source, job).split("\n");
+  const at = lines.findIndex((line) => /^    if:/.test(line));
+  assert.notEqual(at, -1, `${job} must declare an if`);
+  const first = lines[at].replace(/^    if:\s*/, "");
+  if (first !== ">-") return first.replace(/^'(.*)'$/, "$1");
+  const folded = [];
+  for (let i = at + 1; i < lines.length && /^      \S/.test(lines[i]); i += 1) folded.push(lines[i].trim());
+  return folded.join(" ");
+}
+
+function evaluate(expression, context) {
+  return Boolean(vm.runInNewContext(expression, context));
+}
+
+// The sharded suite: one archive per OS, partitions of it in parallel shards,
+// doctests once, and a check that the shards add up to the archive's list.
 assert.match(
-  namedStep(ciTestJob, "Build canonical executable for acceptance tests"),
+  namedStep(ciSuite, "Archive test binaries"),
+  /cargo nextest archive --workspace --all-features --locked --profile ci/,
+  "the suite must archive the same workspace, features and profile the old single run used",
+);
+assert.match(
+  namedStep(ciSuite, "Build canonical executable for acceptance tests"),
   /cargo build -p codewhale-cli --bin codewhale --all-features --locked/,
   "Engine acceptance must build the canonical executable first",
 );
+assert.match(
+  namedStep(ciSuite, "Run doctests"),
+  /cargo test --workspace --all-features --locked --doc/,
+  "nextest does not run doctests; the suite must run them once per OS",
+);
+const shardRun = namedStep(ciSuite, "Run tests");
+assert.match(shardRun, /--partition "count:\$\{\{ matrix\.shard \}\}\/\$\{\{ inputs\.shards \}\}"/);
+assert.match(shardRun, /-E "not \(\$NATIVE_FILTER\)"/, "every shard must exclude the native tests that shard 1 runs apart");
+assert.match(namedStep(ciSuite, "Run native containment tests"), /-E "\$NATIVE_FILTER"/);
+assert.match(ciSuite, /nextest-shard-count\.py check counts/, "the suite must prove the shards add up to the archive's list");
+for (const name of ["Run tests", "Run native containment tests", "Run doctests"]) {
+  const step = namedStep(ciSuite, name);
+  assert.match(step, /shell: bash/, `${name} must invoke the POSIX helper on Windows too`);
+}
+assert.match(ciSuite, /RUST_MIN_STACK: '16777216'/);
+
+// The native receipt script requires its tests in one JUnit report, so the
+// shard that runs them must select exactly the script's list.
+const receiptScript = read("scripts/compiled-host-native-receipt.py");
+const receiptNames = [
+  receiptScript.match(/^COMPILED = "([^"]+)"$/m)[1],
+  receiptScript.match(/^MEMORY = "([^"]+)"$/m)[1],
+  ...[...receiptScript.match(/^WINDOWS = \[([\s\S]*?)^\]/m)[1].matchAll(/"([^"]+)"/g)].map((m) => m[1]),
+];
+const nativeFilter = ciSuite.match(/NATIVE_FILTER: >-\n([\s\S]*?)\n    steps:/)[1];
+assert.deepEqual(
+  [...nativeFilter.matchAll(/test\(=([^)]+)\)/g)].map((m) => m[1]).sort(),
+  receiptNames.sort(),
+  "NATIVE_FILTER must list exactly the tests scripts/compiled-host-native-receipt.py requires",
+);
+
+// Shard counts: each caller's shards equals the length of its shard_list, the
+// suite's own list agrees with it, and every OS keeps a pipeline.
+const suiteCallers = {};
+for (const job of ["test-ubuntu", "test-macos", "test-windows"]) {
+  const block = jobBlock(ci, job);
+  const os = block.match(/^      os: (\S+)$/m)[1];
+  const shards = Number(block.match(/^      shards: (\d+)$/m)[1]);
+  const list = JSON.parse(block.match(/^      shard_list: '(.+)'$/m)[1]);
+  assert.deepEqual(list, Array.from({ length: shards }, (_, i) => i + 1), `${job} shard_list must be 1..shards`);
+  assert.match(block, /uses: \.\/\.github\/workflows\/ci-test-suite\.yml/);
+  assert.equal(jobIf(ci, job), jobIf(ci, "test-ubuntu"), `${job} must share the suite condition`);
+  suiteCallers[os] = shards;
+}
+assert.deepEqual(Object.keys(suiteCallers).sort(), ["macos-latest", "ubuntu-latest", "windows-latest"]);
+
+// Required contexts. The ruleset requires exactly "Lint" and "Test
+// (windows-latest)". Only Lint and the compile-only test-windows-pr job may
+// produce them; a shard, build or suite check must never share a name.
+const jobNames = (source) => [...source.matchAll(/^  [A-Za-z0-9_-]+:\n    name: (.+)$/gm)].map((m) => m[1]);
+assert.deepEqual(
+  [...jobNames(ci), ...jobNames(ciSuite)].filter((name) => /^Test \(/.test(name) || name === "Lint"),
+  ["Lint", "Test (windows-latest)"],
+  "only Lint and the compile-only Windows leg may use the required context names",
+);
+assert.match(jobBlock(ci, "test-windows-pr"), /^    name: Test \(windows-latest\)$/m);
+assert.equal(jobIf(ci, "test-windows-pr"), "github.event_name == 'pull_request'");
+assert.match(jobBlock(ci, "test-windows-pr"), /cargo check --workspace --all-features --all-targets|cargo check --workspace --all-targets --all-features --locked/);
+
+// Each gate that pull requests lost on 2026-11-07 (99cb5a0f4f) runs on a heavy
+// pull request again, plus on dispatch; a docs-only pull request keeps skipping
+// them; a scoped dispatch skips all of them.
+const gates = ["safety-gate", "msrv", "test-ubuntu", "test-macos", "test-windows", "portable-boundaries", "npm-wrapper-smoke", "mobile-smoke"];
+for (const job of gates) {
+  const guard = jobIf(ci, job);
+  const base = { inputs: { workspace_test_mode: "nextest" } };
+  const run = (event, heavy, extra = {}) =>
+    evaluate(guard, {
+      github: { event_name: event },
+      needs: { changes: { outputs: { heavy: String(heavy), mobile: "true", scoped: "false", ...extra } } },
+      ...base,
+    });
+  assert.equal(run("pull_request", true), true, `${job} must run on a heavy pull request`);
+  if (job !== "npm-wrapper-smoke") {
+    assert.equal(run("pull_request", false), false, `${job} must skip a docs-only pull request`);
+  }
+  assert.equal(run("workflow_dispatch", true), true, `${job} must run on a full dispatch`);
+  assert.equal(run("workflow_dispatch", true, { scoped: "true" }), false, `${job} must skip a scoped dispatch`);
+}
+assert.equal(
+  evaluate(jobIf(ci, "mobile-smoke"), {
+    github: { event_name: "pull_request" },
+    needs: { changes: { outputs: { heavy: "true", mobile: "false", scoped: "false" } } },
+  }),
+  false,
+  "mobile smoke keeps its path filter on pull requests",
+);
+for (const job of ["test-ubuntu", "test-macos", "test-windows"]) {
+  assert.equal(
+    evaluate(jobIf(ci, job), {
+      github: { event_name: "workflow_dispatch" },
+      inputs: { workspace_test_mode: "shared-process-twice" },
+      needs: { changes: { outputs: { heavy: "true", scoped: "false" } } },
+    }),
+    false,
+    `${job} must yield to the shared-process qualification`,
+  );
+}
+const sharedGuard = jobIf(ci, "test");
+for (const [event, mode, scoped, expected] of [
+  ["workflow_dispatch", "shared-process-twice", "false", true],
+  ["workflow_dispatch", "nextest", "false", false],
+  ["workflow_dispatch", "shared-process-twice", "true", false],
+  ["pull_request", "nextest", "false", false],
+]) {
+  assert.equal(
+    evaluate(sharedGuard, {
+      github: { event_name: event },
+      inputs: { workspace_test_mode: mode },
+      needs: { changes: { outputs: { heavy: "true", scoped } } },
+    }),
+    expected,
+    `shared-process job: ${event} ${mode} scoped=${scoped}`,
+  );
+}
+assert.equal(
+  evaluate(jobIf(ci, "scoped-test"), { needs: { changes: { outputs: { scoped: "true" } } } }),
+  true,
+);
+assert.equal(
+  evaluate(jobIf(ci, "scoped-test"), { needs: { changes: { outputs: { scoped: "false" } } } }),
+  false,
+);
+assert.match(ci, /^      expected_sha:\n        description: .*\n        required: true\n        type: string$/m, "expected_sha stays required");
+for (const input of ["crate", "filter", "os"]) {
+  assert.match(ci, new RegExp(`^      ${input}:\\n        description: .*\\n        required: false\\n`, "m"), `scoped dispatch input ${input}`);
+}
+
+// Concurrency: a new push to the same pull request or non-main branch cancels
+// the older run; runs on main and scoped dispatches never cancel anything.
+const concurrency = ci.match(/^concurrency:\n  group: (.+)\n  cancel-in-progress: (.+)$/m);
+assert.ok(concurrency, "ci.yml must declare its concurrency");
+const strip = (value) => value.replace(/^\$\{\{ /, "").replace(/ \}\}$/, "");
+const format = (template, ...args) => template.replace(/\{(\d)\}/g, (_, i) => args[Number(i)]);
+const groupOf = (context) =>
+  vm.runInNewContext(strip(concurrency[1]), { ...context, format });
+const cancelOf = (context) => vm.runInNewContext(strip(concurrency[2]), context);
+const dispatch = (ref, sha, inputs = {}) => ({
+  github: { event_name: "workflow_dispatch", ref, sha, workflow: "CI", run_id: `run-${sha}` },
+  inputs: { crate: "", filter: "", ...inputs },
+});
+const pr = (number, sha) => ({
+  github: { event_name: "pull_request", ref: `refs/pull/${number}/merge`, sha, workflow: "CI", event: { pull_request: { number } }, run_id: `run-${sha}` },
+  inputs: {},
+});
+assert.equal(groupOf(pr(7, "a")), groupOf(pr(7, "b")), "a new push to a pull request shares its group");
+assert.notEqual(groupOf(pr(7, "a")), groupOf(pr(8, "a")));
+assert.equal(cancelOf(pr(7, "a")), true);
+assert.equal(groupOf(dispatch("refs/heads/ci/x", "a")), groupOf(dispatch("refs/heads/ci/x", "b")), "a new dispatch on the same ci branch shares its group");
+assert.notEqual(groupOf(dispatch("refs/heads/ci/x", "a")), groupOf(dispatch("refs/heads/ci/y", "a")));
+assert.equal(cancelOf(dispatch("refs/heads/ci/x", "a")), true);
+assert.notEqual(groupOf(dispatch("refs/heads/main", "a")), groupOf(dispatch("refs/heads/main", "b")), "runs on main never share a group");
+assert.equal(cancelOf(dispatch("refs/heads/main", "a")), false, "runs on main never cancel");
+assert.notEqual(
+  groupOf(dispatch("refs/heads/ci/x", "a", { crate: "codewhale-tui" })),
+  groupOf(dispatch("refs/heads/ci/x", "a")),
+  "a scoped dispatch never shares the full run's group",
+);
+assert.equal(cancelOf(dispatch("refs/heads/ci/x", "a", { filter: "test(x)" })), false, "a scoped dispatch cancels nothing");
+
+// Security for pull requests from forks: the workflows use pull_request (no
+// pull_request_target), read-only permissions, no secrets, and write caches
+// only from main.
+for (const [name, source] of [["ci.yml", ci], ["ci-test-suite.yml", ciSuite]]) {
+  assert.doesNotMatch(source, /pull_request_target/, `${name} must not use pull_request_target`);
+  assert.doesNotMatch(source, /\$\{\{\s*secrets\./, `${name} must not read secrets`);
+  assert.match(source, /^permissions:\n  contents: read\n/m, `${name} must keep read-only default permissions`);
+  assert.doesNotMatch(source.replace(/^permissions:\n  contents: read\n/m, ""), /^\s+(?:contents|actions|id-token|pull-requests|packages):\s*write/m, `${name} must not grant write permissions`);
+  for (const match of source.matchAll(/Swatinem\/rust-cache@[^\n]*\n(?:\s+(?:with|cache-bin):[^\n]*\n)*\s+save-if: ([^\n]+)/g)) {
+    assert.match(match[1], /refs\/heads\/main|false/, `${name}: rust-cache must save only on main`);
+  }
+  assert.doesNotMatch(source, /actions\/cache(?:\/save)?@/, `${name} must not write caches outside rust-cache and sccache`);
+}
 const npmSmokeJob = ci.match(/^  npm-wrapper-smoke:\n([\s\S]*?)(?=^  \S)/m)?.[1];
 assert.ok(npmSmokeJob, "CI must retain the npm-wrapper job");
 assert.match(
@@ -79,13 +264,13 @@ assert.match(
   /run: node scripts\/release\/npm-wrapper-smoke\.js/,
 );
 const npmSmokeSteps = npmSmokeJob.split(/(?=^      - )/m).slice(1);
-// Exercise the workflow's actual Boolean guards. The smoke is manual-only
-// (workflow_dispatch); pull requests, pushes and schedules skip the job.
+// Exercise the workflow's actual Boolean guards. The smoke runs on pull
+// requests and dispatches (§180); pushes and schedules skip the job.
 const npmSmokeCases = [
   // name, event, heavy, OS, trusted, cache success, execute, Linux deps, CNB
-  ["own PR", "pull_request", true, "ubuntu-latest", true, true, false, false, false],
-  ["fork PR", "pull_request", true, "ubuntu-latest", false, true, false, false, false],
-  ["PR cache failure", "pull_request", true, "ubuntu-latest", false, false, false, false, false],
+  ["own PR", "pull_request", true, "ubuntu-latest", true, true, true, true, false],
+  ["fork PR", "pull_request", true, "ubuntu-latest", false, true, true, true, false],
+  ["PR cache failure", "pull_request", true, "ubuntu-latest", false, false, true, true, false],
   ["light PR", "pull_request", false, "ubuntu-latest", true, true, false, false, false],
   ["manual Ubuntu", "workflow_dispatch", true, "ubuntu-latest", true, true, true, true, false],
   ["main Ubuntu", "push", true, "ubuntu-latest", true, true, false, false, false],
@@ -95,15 +280,22 @@ const npmSmokeCases = [
   ["light main", "push", false, "ubuntu-latest", true, true, false, false, false],
   ["schedule", "schedule", true, "ubuntu-latest", true, true, false, false, false],
 ];
-// The sccache GitHub Actions backend is main-only, mirroring rust-cache's
-// save-if: pull requests never install or enable it (cache bloat, PLAN D).
+// The sccache GitHub Actions backend is read-only off main: every run reads
+// main's cache, and only main writes it (SCCACHE_GHA_RW_MODE), mirroring
+// rust-cache's save-if (cache bloat, PLAN D).
+for (const [name, source] of [["ci.yml", ci], ["ci-test-suite.yml", ciSuite]]) {
+  const enabled = source.match(/SCCACHE_GHA_ENABLED=true/g) ?? [];
+  const modes = source.match(/if \[ "\$\{GITHUB_REF\}" = refs\/heads\/main \]; then rw=READ_WRITE; else rw=READ_ONLY; fi\n\s+echo "SCCACHE_GHA_RW_MODE=\$\{rw\}"/g) ?? [];
+  assert.ok(enabled.length > 0, `${name} must enable sccache`);
+  assert.equal(modes.length, enabled.length, `${name}: every sccache enable must be read-only off main`);
+  assert.doesNotMatch(source, /sccache-action[^\n]*\n(?:[^\n]*\n){0,4}\s+if: [^\n]*github\.ref == 'refs\/heads\/main'/, `${name}: sccache must install on every run`);
+}
 const sccacheInstallStep = "mozilla-actions/[email protected]";
 for (const [label, event, heavy, os, trusted, cache, execute, linuxDeps, cnb] of npmSmokeCases) {
   const ref = event === "pull_request" ? "refs/pull/1/merge" : "refs/heads/main";
-  const onMain = ref === "refs/heads/main";
-  const installed = execute && onMain;
+  const installed = execute;
   const context = {
-    needs: { changes: { outputs: { heavy: String(heavy), trusted: String(trusted) } } },
+    needs: { changes: { outputs: { heavy: String(heavy), trusted: String(trusted), scoped: "false" } } },
     github: { event_name: event, ref },
     matrix: { os },
     steps: { sccache: { outcome: installed ? (cache ? "success" : "failure") : "skipped" } },
@@ -688,7 +880,8 @@ assert.doesNotMatch(
 // itself runs none: its parity job calls release-parity.yml.
 let hermeticInvocations = 0;
 for (const [label, workflow, expected] of [
-  ["CI", ci, 6],
+  ["CI", ci, 5],
+  ["CI test suite", ciSuite, 3],
   ["release", release, 0],
   ["release parity", parityWorkflow, 4],
   ["CNB", cnb, 3],
@@ -707,11 +900,6 @@ for (const [label, workflow, expected] of [
   }
   hermeticInvocations += commands.length;
 }
-for (const name of ["Run tests", "Run doctests"]) {
-  const step = namedStep(ciTestJob, name);
-  assert.match(step, /shell: bash/, `${name} must invoke the POSIX helper on Windows too`);
-  assert.match(step, /RUST_MIN_STACK: '16777216'/);
-}
 console.log(`Hermetic Rust workflow invocations OK: ${hermeticInvocations} checks passed.`);
 
 const nextest = read(".config/nextest.toml");
@@ -853,6 +1041,7 @@ for (const [name, source] of [
   ["release-parity.yml", parityWorkflow],
   ["release-republish.yml", republish],
   ["ci.yml", ci],
+  ["ci-test-suite.yml", ciSuite],
   ["nightly.yml", nightly],
 ]) {
   assert.deepEqual(

.github/workflows/ci-test-suite.yml

@@ -0,0 +1,394 @@
+name: CI test suite
+
+# Called by ci.yml, once per operating system, so each OS is its own pipeline:
+# one build of the whole workspace's test binaries, parallel shards that run
+# partitions of that one nextest archive, and a check that the shards together
+# ran exactly the tests the archive lists. Running per OS means Linux reports
+# as soon as Linux is done instead of waiting for the slowest build.
+#
+# Trust boundary: this runs with the caller's event. For a pull request from a
+# fork that is the read-only token and no secrets; nothing here writes a cache
+# (rust-cache and sccache save only on main) or uses a write permission. The
+# artifacts passed between jobs live inside one run.
+
+on:
+  workflow_call:
+    inputs:
+      os:
+        description: Hosted runner label
+        required: true
+        type: string
+      shards:
+        description: Number of shards, equal to the length of shard_list
+        required: true
+        type: number
+      shard_list:
+        description: JSON array of shard numbers, for example [1,2,3,4]
+        required: true
+        type: string
+      build_jobs:
+        description: CARGO_BUILD_JOBS for the test build
+        required: true
+        type: string
+      debug:
+        description: Debug-table level for the test build (CARGO_PROFILE_*_DEBUG)
+        required: true
+        type: string
+
+permissions:
+  contents: read
+
+env:
+  CARGO_TERM_COLOR: always
+  CARGO_INCREMENTAL: 0
+  RUSTFLAGS: -Dwarnings
+  RUST_MIN_STACK: 8388608
+
+jobs:
+  build:
+    name: Build
+    # A cold hosted Mac compiled for 65 minutes with one Cargo job (run
+    # 43216596942) and the old one-job cap was 210; this job only compiles, so
+    # it keeps real margin without the test run in it.
+    timeout-minutes: 150
+    runs-on: ${{ inputs.os }}
+    env:
+      # Test builds compile all features of the whole workspace. One Cargo job
+      # kept hosted runners inside memory (exit 143 on 2026-10-19 with four);
+      # trimmed debug tables shrink compiler and linker memory so Linux and
+      # Windows can run two. Assertions and every test are unchanged.
+      CARGO_BUILD_JOBS: ${{ inputs.build_jobs }}
+      CARGO_PROFILE_DEV_DEBUG: ${{ inputs.debug }}
+      CARGO_PROFILE_TEST_DEBUG: ${{ inputs.debug }}
+    steps:
+      - uses: actions/checkout@v7
+      - name: Test Windows installer PATH helper
+        if: inputs.os == 'windows-latest'
+        shell: pwsh
+        run: ./scripts/installer/update-user-path.tests.ps1
+      - name: Install NSIS for Windows installer regression
+        if: inputs.os == 'windows-latest'
+        shell: pwsh
+        run: ./scripts/installer/install-nsis.ps1
+      - name: Test Windows installer PATH regression
+        if: inputs.os == 'windows-latest'
+        shell: pwsh
+        run: ./scripts/installer/installer-path-regression.tests.ps1 -AllowUserPathMutation
+      - uses: dtolnay/rust-toolchain@stable
+      - name: Install macOS LLVM tools
+        if: inputs.os == 'macos-latest'
+        run: rustup component add llvm-tools --toolchain stable
+      - uses: mozilla-actions/[email protected]
+        id: sccache
+        continue-on-error: true
+        # Every run reads the GitHub Actions cache main wrote; only main writes (see Enable sccache).
+      - name: Enable sccache
+        if: steps.sccache.outcome == 'success'
+        shell: bash
+        run: |
+          echo "SCCACHE_GHA_ENABLED=true" >> "${GITHUB_ENV}"
+          if [ "${GITHUB_REF}" = refs/heads/main ]; then rw=READ_WRITE; else rw=READ_ONLY; fi
+          echo "SCCACHE_GHA_RW_MODE=${rw}" >> "${GITHUB_ENV}"
+          echo "RUSTC_WRAPPER=sccache" >> "${GITHUB_ENV}"
+          echo "SCCACHE_IGNORE_SERVER_IO_ERROR=1" >> "${GITHUB_ENV}"
+      - name: Install Linux system dependencies
+        if: inputs.os == 'ubuntu-latest'
+        run: |
+          bash scripts/ci-apt-install.sh libdbus-1-dev pkg-config bubblewrap apparmor-profiles
+          sh scripts/prepare-linux-test-sandbox.sh
+      - name: Ubuntu build headroom
+        # Reuse the same fixed hosted-runner provisioning as Safety and Lint.
+        # Exit 143 is a shutdown observation, not proof of OOM or timeout.
+        if: inputs.os == 'ubuntu-latest'
+        shell: bash
+        run: bash scripts/prepare-ubuntu-build-headroom.sh
+      - uses: Swatinem/rust-cache@v2
+        with:
+          cache-bin: false
+          save-if: ${{ github.ref == 'refs/heads/main' }}
+      - uses: taiki-e/install-action@nextest
+      - uses: actions/[email protected]
+        with:
+          node-version: 22
+      - name: Build canonical executable for acceptance tests
+        shell: bash
+        run: cargo build -p codewhale-cli --bin codewhale --all-features --locked
+        env:
+          RUSTC_WRAPPER: ${{ inputs.os != 'windows-latest' && env.RUSTC_WRAPPER || '' }}
+      - name: Windows Terminal clipboard-to-draft acceptance
+        if: inputs.os == 'windows-latest'
+        timeout-minutes: 5
+        shell: pwsh
+        run: node scripts/release/windows-terminal-clipboard.mjs target/debug/codewhale.exe "${{ runner.temp }}/windows-clipboard"
+      - name: Retain Windows Terminal clipboard-to-draft evidence
+        if: ${{ !cancelled() && inputs.os == 'windows-latest' }}
+        uses: actions/upload-artifact@043fb46d1a93c77aae656e7c1c64a875d1fc6a0a # v7
+        with:
+          name: windows-clipboard-${{ github.sha }}-${{ github.run_attempt }}
+          path: ${{ runner.temp }}/windows-clipboard/
+          if-no-files-found: error
+          retention-days: 7
+      - name: Archive test binaries
+        # One compile per OS. Shards run this archive with
+        # `cargo nextest run --archive-file`, so the suite is the same test
+        # binaries, built once. Doctests are not in the archive; they run
+        # below, once, while the library artifacts are still warm.
+        shell: bash
+        run: |
+          set -euo pipefail
+          mkdir -p "$RUNNER_TEMP/test-archive" "$RUNNER_TEMP/test-expected"
+          cargo nextest archive --workspace --all-features --locked --profile ci \
+            --archive-file "$RUNNER_TEMP/test-archive/nextest-archive.tar.zst"
+          exe=codewhale
+          if [ "$RUNNER_OS" = Windows ]; then exe=codewhale.exe; fi
+          cp "target/debug/$exe" "$RUNNER_TEMP/test-archive/$exe"
+        env:
+          RUSTC_WRAPPER: ${{ inputs.os != 'windows-latest' && env.RUSTC_WRAPPER || '' }}
+      - name: Count the tests the archive selects
+        shell: bash
+        run: |
+          set -euo pipefail
+          cargo nextest list --archive-file "$RUNNER_TEMP/test-archive/nextest-archive.tar.zst" \
+            --workspace-remap . --profile ci --message-format json > "$RUNNER_TEMP/test-listing.json"
+          tests="$(python3 -I scripts/ci/nextest-shard-count.py list "$RUNNER_TEMP/test-listing.json")"
+          printf '{"os": "%s", "tests": %s, "shards": %s}\n' "${{ inputs.os }}" "$tests" "${{ inputs.shards }}" \
+            > "$RUNNER_TEMP/test-expected/expected-${{ inputs.os }}.json"
+          cat "$RUNNER_TEMP/test-expected/expected-${{ inputs.os }}.json"
+      - name: Run doctests
+        # nextest does not run doctests; this keeps them once per OS.
+        shell: bash
+        run: sh scripts/with-hermetic-test-home.sh cargo test --workspace --all-features --locked --doc
+        env:
+          RUSTC_WRAPPER: ${{ inputs.os != 'windows-latest' && env.RUSTC_WRAPPER || '' }}
+          RUST_MIN_STACK: '16777216'
+      - name: Lockfile drift guard
+        run: git diff --exit-code -- Cargo.lock
+      - name: sccache stats
+        if: steps.sccache.outcome == 'success'
+        continue-on-error: true
+        shell: bash
+        run: sccache --show-stats
+      - name: Retain test archive for the shards
+        uses: actions/upload-artifact@043fb46d1a93c77aae656e7c1c64a875d1fc6a0a # v7
+        with:
+          name: test-archive-${{ inputs.os }}
+          path: ${{ runner.temp }}/test-archive/
+          if-no-files-found: error
+          retention-days: 1
+          compression-level: 0
+      - name: Retain expected test count
+        uses: actions/upload-artifact@043fb46d1a93c77aae656e7c1c64a875d1fc6a0a # v7
+        with:
+          name: test-expected-${{ inputs.os }}
+          path: ${{ runner.temp }}/test-expected/
+          if-no-files-found: error
+          retention-days: 7
+
+  shard:
+    name: Shard ${{ matrix.shard }}/${{ inputs.shards }}
+    needs: build
+    timeout-minutes: 60
+    runs-on: ${{ inputs.os }}
+    strategy:
+      # A failure in one shard must not erase the others' results; the suite
+      # gate needs every conclusion.
+      fail-fast: false
+      matrix:
+        shard: ${{ fromJSON(inputs.shard_list) }}
+    env:
+      CODEWHALE_EXT_HOST_TESTS: '1'
+      CODEWHALE_EXT_HOST_BUN_TESTS: '1'
+      # Give test threads the stack the product gives itself; the reasoning is
+      # on the shared-process step above and in scripts/dev-cargo.sh.
+      RUST_MIN_STACK: '16777216'
+      # The exact-image Native containment tests (and, on Windows, the LPAC
+      # owner and profile tests) that scripts/compiled-host-native-receipt.py
+      # requires in ONE JUnit report. Shard 1 runs them in their own nextest
+      # invocation and every shard excludes them from its partition, so each
+      # still runs exactly once. release-workflows.test.js keeps this list equal
+      # to the script's.
+      NATIVE_FILTER: >-
+        package(codewhale-engine) & kind(lib) & (
+        test(=extension_host::tests::compiled_native_host_cannot_read_secrets_or_write_outside_its_data_dir) |
+        test(=extension_host::tests::compiled_native_host_memory_cap_is_enforced) |
+        test(=extension_host::windows::tests::windows_native_lpac_node_owner_boundary) |
+        test(=extension_host::windows::tests::windows_native_lpac_bun_owner_boundary) |
+        test(=extension_host::windows::tests::windows_native_lpac_compiled_owner_boundary) |
+        test(=extension_host::windows::tests::windows_native_profiles_are_distinct_and_acl_grant_refuses_junctions) |
+        test(=extension_host::windows::tests::windows_argv_and_environment_keep_exact_values_and_reject_nul) |
+        test(=extension_host::windows::tests::windows_profile_retirement_removes_only_its_grants_and_inherited_data_on_restarts) |
+        test(=extension_host::windows::tests::windows_profile_directory_budget_and_recorded_identity_refuse_before_overwrite))
+    steps:
+      - uses: actions/checkout@v7
+      - name: Point terminal failure diagnostics at the runner temp directory
+        shell: bash
+        run: echo "QA_PTY_DIAGNOSTICS_DIR=${RUNNER_TEMP}/pty-failures" >> "${GITHUB_ENV}"
+      - uses: dtolnay/rust-toolchain@stable
+      - name: Install Linux system dependencies
+        if: inputs.os == 'ubuntu-latest'
+        run: |
+          bash scripts/ci-apt-install.sh libdbus-1-dev pkg-config bubblewrap apparmor-profiles
+          sh scripts/prepare-linux-test-sandbox.sh
+      - uses: taiki-e/install-action@nextest
+      - uses: actions/[email protected]
+        # The extension-host integration tests spawn the real bundle under
+        # Node >= 22.19; CODEWHALE_EXT_HOST_TESTS makes a missing Node a
+        # failure instead of a silent skip.
+        with:
+          node-version: 22
+      - uses: oven-sh/setup-bun@0c5077e51419868618aeaa5fe8019c62421857d6 # v2.2
+        with:
+          bun-version: 1.4.0
+      - name: Build matching compiled-host input for required Native containment receipts
+        shell: bash
+        run: bash scripts/ci/compiled-host-test-input.sh
+      - uses: actions/download-artifact@9000827ccba6bdab643e8b6fd33ac0654aef8333 # v8
+        with:
+          name: test-archive-${{ inputs.os }}
+          path: ${{ runner.temp }}/test-archive
+      - name: Place the canonical executable for acceptance tests
+        shell: bash
+        run: |
+          set -euo pipefail
+          mkdir -p target/debug
+          cp "$RUNNER_TEMP"/test-archive/codewhale* target/debug/
+          if [ "$RUNNER_OS" != Windows ]; then chmod +x target/debug/codewhale; fi
+          # The archive extracts test binaries to a different layout than a
+          # one-job build, so acceptance tests would not find the executable
+          # beside them; QA_TUI_BIN names it explicitly.
+          exe=codewhale
+          if [ "$RUNNER_OS" = Windows ]; then exe=codewhale.exe; fi
+          echo "QA_TUI_BIN=${GITHUB_WORKSPACE}/target/debug/${exe}" >> "${GITHUB_ENV}"
+      - name: Run native containment tests
+        if: matrix.shard == 1
+        shell: bash
+        run: |
+          set -euo pipefail
+          mkdir -p "$RUNNER_TEMP/test-count" "$RUNNER_TEMP/nextest-extract"
+          sh scripts/with-hermetic-test-home.sh cargo nextest run \
+            --archive-file "$RUNNER_TEMP/test-archive/nextest-archive.tar.zst" \
+            --extract-to "$RUNNER_TEMP/nextest-extract" --extract-overwrite \
+            --workspace-remap . --profile ci -E "$NATIVE_FILTER"
+          cp target/nextest/ci/junit.xml "$RUNNER_TEMP/native-junit.xml"
+          python3 -I scripts/ci/nextest-shard-count.py junit "$RUNNER_TEMP/native-junit.xml" > "$RUNNER_TEMP/native-count.txt"
+      - name: Run tests
+        # Same test binaries as `cargo test`, run by cargo-nextest: one
+        # process per test, all runner cores busy, slow tests named instead
+        # of stalling the binary. `.config/nextest.toml` serializes the PTY
+        # binary and bounds the integration binary that spawns the real
+        # executable; retries are off, so a flake is a red run, not a hidden
+        # one. This shard runs partition k of N of everything except the
+        # native containment tests; test-suite proves the shards add up to the
+        # archive's full list.
+        shell: bash
+        run: |
+          set -euo pipefail
+          mkdir -p "$RUNNER_TEMP/test-count" "$RUNNER_TEMP/nextest-extract"
+          monitor=
+          if [ "$RUNNER_OS" = Linux ]; then
+            # Headroom trace: if the runner is terminated again, the last lines
+            # say whether memory or disk ran out.
+            ( while sleep 30; do
+                echo "[headroom] $(free -m | awk '/^Mem:/{print "mem used " $3 "/" $2 " MiB"} /^Swap:/{print "swap used " $3 "/" $2 " MiB"}' | paste -sd ' ') disk free $(df -h / | awk 'NR==2{print $4}')"
+              done ) &
+            monitor=$!
+          fi
+          status=0
+          sh scripts/with-hermetic-test-home.sh cargo nextest run \
+            --archive-file "$RUNNER_TEMP/test-archive/nextest-archive.tar.zst" \
+            --extract-to "$RUNNER_TEMP/nextest-extract" --extract-overwrite \
+            --workspace-remap . --profile ci \
+            --partition "count:${{ matrix.shard }}/${{ inputs.shards }}" \
+            -E "not ($NATIVE_FILTER)" || status=$?
+          if [ -n "$monitor" ]; then kill "$monitor" 2>/dev/null || true; fi
+          if [ "$status" -ne 0 ]; then exit "$status"; fi
+          tests="$(python3 -I scripts/ci/nextest-shard-count.py junit target/nextest/ci/junit.xml)"
+          if [ -f "$RUNNER_TEMP/native-count.txt" ]; then tests=$((tests + $(cat "$RUNNER_TEMP/native-count.txt"))); fi
+          printf '{"os": "%s", "shard": %s, "tests": %s}\n' "${{ inputs.os }}" "${{ matrix.shard }}" "$tests" \
+            > "$RUNNER_TEMP/test-count/shard-${{ inputs.os }}-${{ matrix.shard }}.json"
+          cat "$RUNNER_TEMP/test-count/shard-${{ inputs.os }}-${{ matrix.shard }}.json"
+      - name: Retain sealed terminal failure diagnostics
+        if: failure()
+        uses: actions/upload-artifact@043fb46d1a93c77aae656e7c1c64a875d1fc6a0a # v7
+        with:
+          name: terminal-failures-${{ runner.os }}-${{ matrix.shard }}-${{ github.sha }}-${{ github.run_attempt }}
+          path: ${{ runner.temp }}/pty-failures/
+          include-hidden-files: true
+          if-no-files-found: ignore
+          retention-days: 7
+      - name: Export exact compiled-image Native containment receipt
+        # JUnit is from the same archive's native invocation: no second
+        # compile. The official entire CI run must be green before release can
+        # consume this artifact. Missing/changed inputs or silent test skips
+        # are red.
+        if: matrix.shard == 1
+        shell: bash
+        run: |
+          set -euo pipefail
+          test "$(git rev-parse HEAD)" = "$GITHUB_SHA"
+          git diff --exit-code
+          git diff --cached --exit-code
+          python3 scripts/compiled-host-native-receipt.py \
+            --junit "$RUNNER_TEMP/native-junit.xml" \
+            --compiled-image "$CODEWHALE_COMPILED_HOST_TEST_BINARY" \
+            --bundle crates/engine/extension-host/dist/codewhale-extension-host.mjs \
+            --bun "$CODEWHALE_COMPILED_HOST_TEST_BUN" --source-sha "$GITHUB_SHA" \
+            --image-platform "$CODEWHALE_COMPILED_HOST_TEST_PLATFORM" --image-arch "$CODEWHALE_COMPILED_HOST_TEST_ARCH" \
+            --host-sha256 "$CODEWHALE_COMPILED_HOST_TEST_SHA256" \
+            --runtime-sha256 "$CODEWHALE_COMPILED_HOST_TEST_RUNTIME_SHA256" \
+            --bundle-sha256 "$CODEWHALE_COMPILED_HOST_TEST_BUNDLE_SHA256" \
+            --output "$RUNNER_TEMP/native-compiled-host"
+      - name: Retain exact compiled image and Native receipt
+        if: matrix.shard == 1
+        uses: actions/upload-artifact@043fb46d1a93c77aae656e7c1c64a875d1fc6a0a # v7
+        with:
+          name: native-compiled-host-${{ runner.os }}-${{ runner.arch }}
+          path: ${{ runner.temp }}/native-compiled-host/
+          if-no-files-found: error
+          retention-days: 30
+      - name: Lockfile drift guard
+        run: git diff --exit-code -- Cargo.lock
+      - name: Retain shard test count
+        uses: actions/upload-artifact@043fb46d1a93c77aae656e7c1c64a875d1fc6a0a # v7
+        with:
+          name: test-count-${{ inputs.os }}-${{ matrix.shard }}
+          path: ${{ runner.temp }}/test-count/
+          if-no-files-found: error
+          retention-days: 7
+
+  check:
+    # The job that answers "did this OS's whole suite pass": the build and every
+    # shard succeeded, and the shards together ran exactly the tests the archive
+    # lists. Not a required check (§180).
+    name: Suite check
+    needs: [build, shard]
+    if: '!cancelled()'
+    timeout-minutes: 10
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v7
+        with:
+          sparse-checkout: scripts/ci
+      - name: Every build and shard succeeded
+        shell: bash
+        env:
+          BUILD_RESULT: ${{ needs.build.result }}
+          SHARD_RESULT: ${{ needs.shard.result }}
+        run: |
+          echo "build: ${BUILD_RESULT}"
+          echo "shard: ${SHARD_RESULT}"
+          if [ "${BUILD_RESULT}" != success ] || [ "${SHARD_RESULT}" != success ]; then
+            echo "::error::The sharded suite did not complete successfully."
+            exit 1
+          fi
+      - uses: actions/download-artifact@9000827ccba6bdab643e8b6fd33ac0654aef8333 # v8
+        with:
+          name: test-expected-${{ inputs.os }}
+          path: counts
+      - uses: actions/download-artifact@9000827ccba6bdab643e8b6fd33ac0654aef8333 # v8
+        with:
+          pattern: test-count-${{ inputs.os }}-*
+          path: counts
+          merge-multiple: true
+      - name: Shards ran every listed test exactly once
+        run: python3 -I scripts/ci/nextest-shard-count.py check counts

.github/workflows/ci.yml

@@ -16,13 +16,32 @@ on:
         options:
           - nextest
           - shared-process-twice
+      crate:
+        description: Scoped run, in a few minutes. One workspace package (for example codewhale-tui); with crate or filter set only the scoped test runs, not the full suite
+        required: false
+        default: ''
+        type: string
+      filter:
+        description: Scoped run, in a few minutes. A nextest filterset (for example 'test(auto_review)'), alone or with crate; with crate or filter set only the scoped test runs
+        required: false
+        default: ''
+        type: string
+      os:
+        description: Operating systems for a scoped run, comma separated (ubuntu-latest, macos-latest, windows-latest, or all); ignored by a full run
+        required: false
+        default: ubuntu-latest
+        type: string
 
 permissions:
   contents: read
 
+# A new push to a pull request, or a new dispatch on the same non-main branch
+# (an agent's ci/<task> branch), cancels the older run still in progress.
+# Runs on main are keyed by SHA and never cancel each other, and a scoped
+# dispatch gets its own group so it cannot cancel a full run on the same branch.
 concurrency:
-  group: ${{ github.event_name == 'pull_request' && format('ci-pr-{0}', github.event.pull_request.number) || format('ci-{0}-{1}', github.workflow, github.sha) }}
-  cancel-in-progress: ${{ github.event_name == 'pull_request' }}
+  group: ${{ github.event_name == 'pull_request' && format('ci-pr-{0}', github.event.pull_request.number) || (github.event_name == 'workflow_dispatch' && (inputs.crate != '' || inputs.filter != '') && format('ci-scoped-{0}', github.run_id)) || (github.ref != 'refs/heads/main' && format('ci-ref-{0}', github.ref)) || format('ci-{0}-{1}', github.workflow, github.sha) }}
+  cancel-in-progress: ${{ github.ref != 'refs/heads/main' && !(github.event_name == 'workflow_dispatch' && (inputs.crate != '' || inputs.filter != '')) }}
 
 env:
   CARGO_TERM_COLOR: always
@@ -45,6 +64,8 @@ jobs:
       mobile: ${{ steps.detect.outputs.mobile }}
       actions: ${{ steps.detect.outputs.actions }}
       trusted: ${{ steps.trust.outputs.trusted }}
+      scoped: ${{ steps.scope.outputs.scoped }}
+      scoped_os: ${{ steps.scope.outputs.scoped_os }}
     steps:
       - name: Classify event trust
         id: trust
@@ -64,6 +85,48 @@ jobs:
           else
             echo "trusted=false" >> "$GITHUB_OUTPUT"
           fi
+      - name: Classify scoped dispatch
+        id: scope
+        shell: bash
+        env:
+          EVENT_NAME: ${{ github.event_name }}
+          SCOPE_CRATE: ${{ inputs.crate }}
+          SCOPE_FILTER: ${{ inputs.filter }}
+          SCOPE_OS: ${{ inputs.os }}
+        run: |
+          set -euo pipefail
+          # A dispatch with a crate or a filter runs only the scoped test job.
+          # Inputs reach the shell only through env, and the crate and OS
+          # names are validated before anything uses them.
+          scoped=false
+          scoped_os='["ubuntu-latest"]'
+          if [[ "${EVENT_NAME}" == "workflow_dispatch" && ( -n "${SCOPE_CRATE}" || -n "${SCOPE_FILTER}" ) ]]; then
+            scoped=true
+            if [[ -n "${SCOPE_CRATE}" && ! "${SCOPE_CRATE}" =~ ^[A-Za-z0-9_-]+$ ]]; then
+              echo "::error::crate must be a package name (letters, digits, - and _)." >&2
+              exit 1
+            fi
+            oses=()
+            IFS=',' read -r -a requested <<< "${SCOPE_OS:-ubuntu-latest}"
+            for name in "${requested[@]}"; do
+              name="${name//[[:space:]]/}"
+              case "${name}" in
+                ubuntu|ubuntu-latest|linux) oses+=("ubuntu-latest") ;;
+                macos|macos-latest) oses+=("macos-latest") ;;
+                windows|windows-latest) oses+=("windows-latest") ;;
+                all) oses+=("ubuntu-latest" "macos-latest" "windows-latest") ;;
+                "") ;;
+                *)
+                  echo "::error::os entries must be ubuntu-latest, macos-latest, windows-latest or all." >&2
+                  exit 1
+                  ;;
+              esac
+            done
+            if [[ "${#oses[@]}" -eq 0 ]]; then oses=("ubuntu-latest"); fi
+            scoped_os="$(printf '%s\n' "${oses[@]}" | sort -u | jq -R . | jq -sc .)"
+          fi
+          echo "scoped=${scoped}" >> "${GITHUB_OUTPUT}"
+          echo "scoped_os=${scoped_os}" >> "${GITHUB_OUTPUT}"
       - uses: actions/checkout@v7
         with:
           fetch-depth: 0
@@ -197,6 +260,8 @@ jobs:
 
   versions:
     name: Version drift
+    needs: changes
+    if: needs.changes.outputs.scoped != 'true'
     timeout-minutes: 15
     runs-on: ubuntu-latest
     steps:
@@ -274,6 +339,8 @@ jobs:
 
   plugin-conversion:
     name: Plugin conversion
+    needs: changes
+    if: needs.changes.outputs.scoped != 'true'
     timeout-minutes: 5
     runs-on: ubuntu-latest
     steps:
@@ -293,6 +360,8 @@ jobs:
 
   integrations:
     name: Integrations
+    needs: changes
+    if: needs.changes.outputs.scoped != 'true'
     timeout-minutes: 15
     runs-on: ubuntu-latest
     steps:
@@ -349,7 +418,7 @@ jobs:
   extension-host-runtimes:
     name: Extension host runtimes (${{ matrix.os }})
     needs: changes
-    if: needs.changes.outputs.heavy == 'true'
+    if: needs.changes.outputs.heavy == 'true' && needs.changes.outputs.scoped != 'true'
     timeout-minutes: 15
     strategy:
       fail-fast: false
@@ -388,6 +457,8 @@ jobs:
 
   vscode-extension:
     name: VS Code extension
+    needs: changes
+    if: needs.changes.outputs.scoped != 'true'
     timeout-minutes: 15
     runs-on: ubuntu-latest
     defaults:
@@ -421,7 +492,7 @@ jobs:
   safety-gate:
     name: Safety gate
     needs: changes
-    if: github.event_name == 'workflow_dispatch' && needs.changes.outputs.heavy == 'true'
+    if: needs.changes.outputs.heavy == 'true' && needs.changes.outputs.scoped != 'true'
     timeout-minutes: 30
     runs-on: ubuntu-latest
     steps:
@@ -431,16 +502,18 @@ jobs:
           toolchain: stable
       - uses: mozilla-actions/[email protected]
         id: sccache
-        # The GitHub Actions cache backend is main-only, mirroring
-        # rust-cache's save-if: PR runs wrote thousands of refs/pull/N
-        # entries that pushed the repo past its 10 GiB cache cap.
-        if: github.ref == 'refs/heads/main'
+        # Every run reads the cache main wrote; only main writes it
+        # (SCCACHE_GHA_RW_MODE in Enable sccache), mirroring rust-cache's
+        # save-if. PR runs once wrote thousands of refs/pull/N entries that
+        # pushed the repo past its 10 GiB cache cap.
         continue-on-error: true
       - name: Enable sccache
         if: steps.sccache.outcome == 'success'
         shell: bash
         run: |
           echo "SCCACHE_GHA_ENABLED=true" >> "${GITHUB_ENV}"
+          if [ "${GITHUB_REF}" = refs/heads/main ]; then rw=READ_WRITE; else rw=READ_ONLY; fi
+          echo "SCCACHE_GHA_RW_MODE=${rw}" >> "${GITHUB_ENV}"
           echo "RUSTC_WRAPPER=sccache" >> "${GITHUB_ENV}"
           echo "SCCACHE_IGNORE_SERVER_IO_ERROR=1" >> "${GITHUB_ENV}"
       - name: Install Linux system dependencies
@@ -475,6 +548,7 @@ jobs:
   lint:
     name: Lint
     needs: changes
+    if: needs.changes.outputs.scoped != 'true'
     # Measured on 89ee629e with runner headroom: clippy 10 min, runtime-contract
     # budget 16 min, and persistence-backlog still running at the old 45-minute
     # cancellation. Commands, filters and budgets are unchanged.
@@ -494,13 +568,15 @@ jobs:
         # Cache bootstrap failures (e.g. GitHub 504s fetching the sccache
         # binary) degrade to an uncached build instead of failing product CI.
         continue-on-error: true
-        # Main-only GitHub Actions cache backend; see the Safety gate job.
-        if: needs.changes.outputs.heavy == 'true' && github.ref == 'refs/heads/main'
+        # Every run reads the GitHub Actions cache main wrote; only main writes (see Enable sccache).
+        if: needs.changes.outputs.heavy == 'true'
       - name: Enable sccache
         if: needs.changes.outputs.heavy == 'true' && steps.sccache.outcome == 'success'
         shell: bash
         run: |
           echo "SCCACHE_GHA_ENABLED=true" >> "${GITHUB_ENV}"
+          if [ "${GITHUB_REF}" = refs/heads/main ]; then rw=READ_WRITE; else rw=READ_ONLY; fi
+          echo "SCCACHE_GHA_RW_MODE=${rw}" >> "${GITHUB_ENV}"
           echo "RUSTC_WRAPPER=sccache" >> "${GITHUB_ENV}"
           echo "SCCACHE_IGNORE_SERVER_IO_ERROR=1" >> "${GITHUB_ENV}"
       - name: Install Linux system dependencies
@@ -725,7 +801,7 @@ jobs:
     # fail the build there while stable stays green (#6543).
     name: MSRV check (1.89)
     needs: changes
-    if: github.event_name == 'workflow_dispatch' && needs.changes.outputs.heavy == 'true'
+    if: needs.changes.outputs.heavy == 'true' && needs.changes.outputs.scoped != 'true'
     timeout-minutes: 45
     runs-on: ubuntu-latest
     steps:
@@ -746,7 +822,7 @@ jobs:
   workflow-rlm-cache:
     name: Workflow RLM cache
     needs: changes
-    if: needs.changes.outputs.workflow == 'true'
+    if: needs.changes.outputs.workflow == 'true' && needs.changes.outputs.scoped != 'true'
     timeout-minutes: 30
     runs-on: ubuntu-latest
     steps:
@@ -754,14 +830,15 @@ jobs:
       - uses: dtolnay/rust-toolchain@stable
       - uses: mozilla-actions/[email protected]
         id: sccache
-        # Main-only GitHub Actions cache backend; see the Safety gate job.
-        if: github.ref == 'refs/heads/main'
+        # Every run reads the GitHub Actions cache main wrote; only main writes (see Enable sccache).
         continue-on-error: true
       - name: Enable sccache
         if: steps.sccache.outcome == 'success'
         shell: bash
         run: |
           echo "SCCACHE_GHA_ENABLED=true" >> "${GITHUB_ENV}"
+          if [ "${GITHUB_REF}" = refs/heads/main ]; then rw=READ_WRITE; else rw=READ_ONLY; fi
+          echo "SCCACHE_GHA_RW_MODE=${rw}" >> "${GITHUB_ENV}"
           echo "RUSTC_WRAPPER=sccache" >> "${GITHUB_ENV}"
           echo "SCCACHE_IGNORE_SERVER_IO_ERROR=1" >> "${GITHUB_ENV}"
       - uses: Swatinem/rust-cache@v2
@@ -772,261 +849,64 @@ jobs:
         run: sh scripts/with-hermetic-test-home.sh cargo test -p codewhale-workflow --locked
 
   test:
-    name: Test
+    # Manual shared-process qualification only (workspace_test_mode =
+    # shared-process-twice): the GH6698 evidence of two libtest passes at one
+    # exact revision on one hosted Ubuntu runner. Every other run executes the
+    # suite through test-build, test-shard and test-suite below.
+    name: Shared-process tests
     needs: changes
-    if: github.event_name == 'workflow_dispatch'
-    # Contexts "Test (ubuntu-latest)" / "Test (macos-latest)" /
-    # "Test (windows-latest)" derive from job name + matrix.os and are
-    # independent of runs-on. For light changes the macOS/Windows legs only
-    # echo a skip line, so run them on ubuntu instead of queueing for scarce
-    # macOS/Windows runners. Pull requests run only the Linux leg here; the
-    # required "Test (windows-latest)" context comes from test-windows-pr.
-    # The ternary is safe: matrix.os is always a non-empty literal, so
-    # runs-on can never evaluate to empty.
-    # A cold GitHub-hosted Mac spent 77-80 minutes in Test on 2026-09-23
-    # (fork PRs #6431, #6417), too close to the old 90-minute limit.
-    # On 2026-10-04 a PR leg spent 111 minutes compiling with no cache and
-    # was cancelled at 120 with 12,356 tests passed and none failed. The
-    # limit stays a literal (release-workflows.test.js, #5496); 210 bounds
-    # cold compilation and the complete suite without disabling test deadlines.
+    if: github.event_name == 'workflow_dispatch' && inputs.workspace_test_mode == 'shared-process-twice' && needs.changes.outputs.heavy == 'true' && needs.changes.outputs.scoped != 'true'
+    # A cold GitHub-hosted runner spent 111 minutes compiling with no cache on
+    # 2026-10-04. The limit stays a literal (release-workflows.test.js, #5496);
+    # 210 bounds cold compilation and both complete passes.
     timeout-minutes: 210
-    # macOS legs go to the self-hosted Mac ONLY when all three hold: the
-    # change is heavy, the event is trusted (not a fork PR), and the
-    # CW_SELF_HOSTED_MAC repo variable is 'true'. That variable is the kill
-    # switch: unset it and every leg falls back to GitHub-hosted runners
-    # immediately, with no commit — important because an offline
-    # self-hosted runner queues jobs forever, which is worse than a slow one.
-    runs-on: ${{ needs.changes.outputs.heavy != 'true' && 'ubuntu-latest' || (matrix.os == 'macos-latest' && needs.changes.outputs.trusted == 'true' && vars.CW_SELF_HOSTED_MAC == 'true' && fromJSON('["self-hosted","macOS","ARM64","codewhale-mac"]')) || matrix.os }}
-    strategy:
-      # A failure on one desktop platform must not erase evidence from the
-      # other one. We need both conclusions to diagnose and release safely.
-      fail-fast: false
-      matrix:
-        # Pull requests run the Linux workspace suite only; the macOS and
-        # Windows suites run in final validation. The explicit manual
-        # libtest qualification uses one hosted Ubuntu job, never a fork PR or
-        # self-hosted runner; normal dispatches keep all three legs.
-        os: ${{ fromJSON((github.event_name == 'pull_request' || (github.event_name == 'workflow_dispatch' && inputs.workspace_test_mode == 'shared-process-twice')) && '["ubuntu-latest"]' || '["ubuntu-latest","macos-latest","windows-latest"]') }}
+    runs-on: ubuntu-latest
     steps:
-      - name: Skip tests for light change
-        if: needs.changes.outputs.heavy != 'true'
-        run: echo "No executable Rust changes detected; preserving required Test context."
       - uses: actions/checkout@v7
-        if: needs.changes.outputs.heavy == 'true' && (matrix.os != 'ubuntu-latest' || github.event_name == 'workflow_dispatch' || github.event_name == 'pull_request')
-      - name: Test Windows installer PATH helper
-        if: needs.changes.outputs.heavy == 'true' && matrix.os == 'windows-latest'
-        shell: pwsh
-        run: ./scripts/installer/update-user-path.tests.ps1
-      - name: Install NSIS for Windows installer regression
-        if: needs.changes.outputs.heavy == 'true' && matrix.os == 'windows-latest'
-        shell: pwsh
-        run: ./scripts/installer/install-nsis.ps1
-      - name: Test Windows installer PATH regression
-        if: needs.changes.outputs.heavy == 'true' && matrix.os == 'windows-latest'
-        shell: pwsh
-        run: ./scripts/installer/installer-path-regression.tests.ps1 -AllowUserPathMutation
       - uses: dtolnay/rust-toolchain@stable
-        if: needs.changes.outputs.heavy == 'true' && (matrix.os != 'ubuntu-latest' || github.event_name == 'workflow_dispatch' || github.event_name == 'pull_request')
-      - name: Install macOS LLVM tools
-        if: needs.changes.outputs.heavy == 'true' && matrix.os == 'macos-latest'
-        run: rustup component add llvm-tools --toolchain stable
-      - name: Omit macOS build debug metadata
-        if: needs.changes.outputs.heavy == 'true' && matrix.os == 'macos-latest'
-        shell: bash
-        # Match the existing full-workspace CNB gate. Keep debug assertions and
-        # every test; only disposable build debug tables change on this lane.
-        run: printf '%s\n' 'CARGO_PROFILE_DEV_DEBUG=0' 'CARGO_PROFILE_TEST_DEBUG=0' >> "$GITHUB_ENV"
       - uses: mozilla-actions/[email protected]
         id: sccache
         continue-on-error: true
-        # Main-only GitHub Actions cache backend; see the Safety gate job.
-        if: needs.changes.outputs.heavy == 'true' && (matrix.os != 'ubuntu-latest' || github.event_name == 'workflow_dispatch' || github.event_name == 'pull_request') && github.ref == 'refs/heads/main'
+        # Every run reads the GitHub Actions cache main wrote; only main writes (see Enable sccache).
       - name: Enable sccache
-        if: needs.changes.outputs.heavy == 'true' && (matrix.os != 'ubuntu-latest' || github.event_name == 'workflow_dispatch' || github.event_name == 'pull_request') && steps.sccache.outcome == 'success'
+        if: steps.sccache.outcome == 'success'
         shell: bash
         run: |
           echo "SCCACHE_GHA_ENABLED=true" >> "${GITHUB_ENV}"
+          if [ "${GITHUB_REF}" = refs/heads/main ]; then rw=READ_WRITE; else rw=READ_ONLY; fi
+          echo "SCCACHE_GHA_RW_MODE=${rw}" >> "${GITHUB_ENV}"
           echo "RUSTC_WRAPPER=sccache" >> "${GITHUB_ENV}"
           echo "SCCACHE_IGNORE_SERVER_IO_ERROR=1" >> "${GITHUB_ENV}"
       - name: Install Linux system dependencies
-        if: needs.changes.outputs.heavy == 'true' && matrix.os == 'ubuntu-latest' && (github.event_name == 'workflow_dispatch' || github.event_name == 'pull_request')
         run: |
           bash scripts/ci-apt-install.sh libdbus-1-dev pkg-config bubblewrap apparmor-profiles
           sh scripts/prepare-linux-test-sandbox.sh
       - name: Ubuntu build headroom
         # Reuse the same fixed hosted-runner provisioning as Safety and Lint.
         # Exit 143 is a shutdown observation, not proof of OOM or timeout.
-        if: needs.changes.outputs.heavy == 'true' && matrix.os == 'ubuntu-latest' && (github.event_name == 'workflow_dispatch' || github.event_name == 'pull_request')
         shell: bash
         run: bash scripts/prepare-ubuntu-build-headroom.sh
       - uses: Swatinem/rust-cache@v2
-        if: needs.changes.outputs.heavy == 'true' && (matrix.os != 'ubuntu-latest' || github.event_name == 'workflow_dispatch' || github.event_name == 'pull_request')
         with:
           cache-bin: false
           save-if: ${{ github.ref == 'refs/heads/main' }}
-      - uses: taiki-e/install-action@nextest
-        if: needs.changes.outputs.heavy == 'true' && (matrix.os != 'ubuntu-latest' || github.event_name == 'workflow_dispatch' || github.event_name == 'pull_request')
-      - name: Check portable config policy boundary
-        if: needs.changes.outputs.heavy == 'true' && matrix.os == 'ubuntu-latest' && (github.event_name == 'workflow_dispatch' || github.event_name == 'pull_request')
-        shell: bash
-        run: |
-          python3 -B scripts/test_check_command_config_proof.py
-          sh scripts/check-portable-config.sh
-      - name: Check portable core navigation boundary
-        if: needs.changes.outputs.heavy == 'true' && matrix.os == 'ubuntu-latest' && (github.event_name == 'workflow_dispatch' || github.event_name == 'pull_request')
-        shell: bash
-        run: |
-          python3 -B scripts/test_check_command_core_navigation_proof.py
-          sh scripts/check-portable-core-navigation.sh
       - uses: actions/[email protected]
         # The extension-host integration tests spawn the real bundle under
         # Node >= 22.19; CODEWHALE_EXT_HOST_TESTS below makes a missing Node a
         # failure instead of a silent skip.
-        if: needs.changes.outputs.heavy == 'true' && (matrix.os != 'ubuntu-latest' || github.event_name == 'workflow_dispatch' || github.event_name == 'pull_request')
         with:
           node-version: 22
       - uses: oven-sh/setup-bun@0c5077e51419868618aeaa5fe8019c62421857d6 # v2.2
-        if: needs.changes.outputs.heavy == 'true' && (matrix.os != 'ubuntu-latest' || github.event_name == 'workflow_dispatch' || github.event_name == 'pull_request')
         with:
           bun-version: 1.4.0
       - name: Build matching compiled-host input for required Native containment receipts
-        if: needs.changes.outputs.heavy == 'true' && (matrix.os != 'ubuntu-latest' || github.event_name == 'workflow_dispatch' || github.event_name == 'pull_request')
         shell: bash
-        run: |
-          set -euo pipefail
-          bun_binary="$(bun -p 'process.execPath')"
-          host_binary="$(node -p "require('path').join(process.env.RUNNER_TEMP, 'codewhale-extension-host' + (process.platform === 'win32' ? '.exe' : ''))")"
-          compiled_info="$(node crates/engine/extension-host/compile-host.mjs --bun "$bun_binary" --output "$host_binary")"
-          printf 'CODEWHALE_COMPILED_HOST_TEST_BINARY=%s\n' "$host_binary" >> "$GITHUB_ENV"
-          node - "$host_binary" "$bun_binary" "$compiled_info" >> "$GITHUB_ENV" <<'JS'
-          const fs = require('node:fs'), crypto = require('node:crypto');
-          const digest = (file) => crypto.createHash('sha256').update(fs.readFileSync(file)).digest('hex');
-          const info = JSON.parse(process.argv[4]);
-          if (!info.platform || !info.arch || info.platform !== info.compiler?.platform || info.arch !== info.compiler?.arch) throw new Error('image/compiler identity mismatch');
-          console.log('CODEWHALE_COMPILED_HOST_TEST_PLATFORM=' + info.platform);
-          console.log('CODEWHALE_COMPILED_HOST_TEST_ARCH=' + info.arch);
-          console.log('CODEWHALE_COMPILED_HOST_TEST_SHA256=' + digest(process.argv[2]));
-          console.log('CODEWHALE_COMPILED_HOST_TEST_BUN=' + process.argv[3]);
-          console.log('CODEWHALE_COMPILED_HOST_TEST_RUNTIME_SHA256=' + digest(process.argv[3]));
-          console.log('CODEWHALE_COMPILED_HOST_TEST_BUNDLE_SHA256=' + digest('crates/engine/extension-host/dist/codewhale-extension-host.mjs'));
-          JS
-          # The required hermetic full invocation executes the shared actual
-          # compiled Native secret/history/write scenario on each OS; Windows
-          # additionally requires Node/Bun/compiled LPAC owner/reparse probes.
-          # Missing requested compiled input or a failed sandbox is red.
-      - name: Hold this machine's build lock (self-hosted)
-        # The self-hosted Mac is also a developer machine: local agents build
-        # through scripts/dev-cargo.sh, which holds this lock, and two Cargo
-        # builds at once exhaust its memory. Opt-in: the runner's `.env` names
-        # the same file as CODEWHALE_BUILD_LOCK_FILE. The holder is a
-        # background process; the runner kills orphans when the job ends, so a
-        # cancelled or crashed job cannot leave the lock held.
-        # (`env.*` in `if:` cannot see the runner's `.env`, so the opt-in is
-        # checked in the script.)
-        if: needs.changes.outputs.heavy == 'true' && runner.environment == 'self-hosted'
-        shell: bash
-        run: |
-          if [ -z "${CODEWHALE_BUILD_LOCK_FILE:-}" ]; then
-            echo "CODEWHALE_BUILD_LOCK_FILE is not set on this runner; building without the machine lock."
-            exit 0
-          fi
-          hold="$RUNNER_TEMP/cw-build-lock.hold"
-          ready="$RUNNER_TEMP/cw-build-lock.ready"
-          log="$RUNNER_TEMP/cw-build-lock.log"
-          touch "$hold"
-          rm -f "$ready"
-          # The single quotes are deliberate: the inner sh expands $1/$2.
-          # shellcheck disable=SC2016
-          nohup python3 scripts/build-lock.py "$CODEWHALE_BUILD_LOCK_FILE" -- \
-            sh -c 'touch "$1"; while [ -e "$2" ]; do sleep 2; done' _ "$ready" "$hold" \
-            >"$log" 2>&1 &
-          shown=0
-          until [ -e "$ready" ]; do
-            if [ "$shown" -eq 0 ] && [ -s "$log" ]; then cat "$log"; shown=1; fi
-            sleep 2
-          done
-          cat "$log"
-          # Scripts inside this job already run under the lock.
-          echo "CODEWHALE_BUILD_LOCK_HELD=1" >> "$GITHUB_ENV"
+        run: bash scripts/ci/compiled-host-test-input.sh
       - name: Build canonical executable for acceptance tests
-        if: needs.changes.outputs.heavy == 'true' && (matrix.os != 'ubuntu-latest' || github.event_name == 'workflow_dispatch' || github.event_name == 'pull_request')
         shell: bash
         run: cargo build -p codewhale-cli --bin codewhale --all-features --locked
-        env:
-          RUSTC_WRAPPER: ${{ matrix.os != 'windows-latest' && env.RUSTC_WRAPPER || '' }}
-      - name: Windows Terminal clipboard-to-draft acceptance
-        if: needs.changes.outputs.heavy == 'true' && matrix.os == 'windows-latest'
-        timeout-minutes: 5
-        shell: pwsh
-        run: node scripts/release/windows-terminal-clipboard.mjs target/debug/codewhale.exe "${{ runner.temp }}/windows-clipboard"
-      - name: Retain Windows Terminal clipboard-to-draft evidence
-        if: ${{ !cancelled() && matrix.os == 'windows-latest' && needs.changes.outputs.heavy == 'true' }}
-        uses: actions/upload-artifact@043fb46d1a93c77aae656e7c1c64a875d1fc6a0a # v7
-        with:
-          name: windows-clipboard-${{ github.sha }}-${{ github.run_attempt }}
-          path: ${{ runner.temp }}/windows-clipboard/
-          if-no-files-found: error
-          retention-days: 7
-      - name: Run tests
-        # Same test binaries as `cargo test`, run by cargo-nextest: one
-        # process per test, all runner cores busy, slow tests named instead
-        # of stalling the binary. `.config/nextest.toml` serializes the PTY
-        # binary and bounds the integration binary that spawns the real
-        # executable; retries are off, so a flake is a red run, not a hidden
-        # one. nextest does not run doctests — the next step keeps them.
-        if: needs.changes.outputs.heavy == 'true' && (matrix.os != 'ubuntu-latest' || github.event_name == 'workflow_dispatch' || github.event_name == 'pull_request') && !(github.event_name == 'workflow_dispatch' && inputs.workspace_test_mode == 'shared-process-twice')
-        shell: bash
-        run: |
-          monitor=
-          if [ "$RUNNER_OS" = Linux ]; then
-            # Headroom trace: if the runner is terminated again, the last lines
-            # say whether memory or disk ran out.
-            ( while sleep 30; do
-                echo "[headroom] $(free -m | awk '/^Mem:/{print "mem used " $3 "/" $2 " MiB"} /^Swap:/{print "swap used " $3 "/" $2 " MiB"}' | paste -sd ' ') disk free $(df -h / | awk 'NR==2{print $4}')"
-              done ) &
-            monitor=$!
-          fi
-          status=0
-          sh scripts/with-hermetic-test-home.sh cargo nextest run --workspace --all-features --locked --profile ci || status=$?
-          if [ -n "$monitor" ]; then kill "$monitor" 2>/dev/null || true; fi
-          exit "$status"
-        env:
-          CARGO_BUILD_JOBS: '1'
-          QA_PTY_DIAGNOSTICS_DIR: ${{ runner.temp }}/pty-failures
-          CODEWHALE_EXT_HOST_TESTS: '1'
-          CODEWHALE_EXT_HOST_BUN_TESTS: '1'
-          # sccache 0.17 panics resolving its config directory under the
-          # isolated Windows home before Cargo can compile or run any test.
-          # Bypass only that optional cache; keep the full suite and isolation.
-          RUSTC_WRAPPER: ${{ matrix.os != 'windows-latest' && env.RUSTC_WRAPPER || '' }}
-          # Give test threads the stack the product gives itself. main.rs runs
-          # the owner thread and every tokio worker at
-          # CODEWHALE_MAIN_STACK_BYTES (32 MiB) because the engine and
-          # runtime-thread futures are genuinely deep. `#[tokio::test]` builds
-          # its own runtime and never sees that, so tests ran the same code on
-          # ~2 MiB (~1 MiB on Windows) — a configuration that never ships.
-          # That gap is what aborted the whole Windows test binary with
-          # STATUS_STACK_OVERFLOW in start_turn_accepts_dynamic_tools_and_
-          # environment_id, masking every other Windows result (78afd8d3d4
-          # Box::pin'd that one frame; the mismatch itself remained). std reads
-          # this for any thread spawned without an explicit size, which covers
-          # both libtest's per-test threads and tokio's workers. Test threads
-          # hold 16 MiB: the engine-only chains they run measured a ~2.5 MiB
-          # debug high-water, while the full ~16.5 MiB UI-loop chain that
-          # forced 32 MiB lives in spawned binaries, which size their own
-          # stacks explicitly and never read this variable.
-          RUST_MIN_STACK: '16777216'
-      - name: Retain sealed terminal failure diagnostics
-        if: failure()
-        uses: actions/upload-artifact@043fb46d1a93c77aae656e7c1c64a875d1fc6a0a # v7
-        with:
-          name: terminal-failures-${{ runner.os }}-${{ github.sha }}-${{ github.run_attempt }}
-          path: ${{ runner.temp }}/pty-failures/
-          include-hidden-files: true
-          if-no-files-found: ignore
-          retention-days: 7
       - name: Shared-process workspace qualification
-        if: github.event_name == 'workflow_dispatch' && inputs.workspace_test_mode == 'shared-process-twice' && matrix.os == 'ubuntu-latest'
+        if: github.event_name == 'workflow_dispatch' && inputs.workspace_test_mode == 'shared-process-twice'
         shell: bash
         env:
           CARGO_BUILD_JOBS: '1'
@@ -1128,83 +1008,193 @@ jobs:
           cat "$evidence/results.txt"
           exit "$failed"
       - name: Retain shared-process qualification evidence
-        if: always() && github.event_name == 'workflow_dispatch' && inputs.workspace_test_mode == 'shared-process-twice' && matrix.os == 'ubuntu-latest'
+        if: always() && github.event_name == 'workflow_dispatch' && inputs.workspace_test_mode == 'shared-process-twice'
         uses: actions/upload-artifact@v7
         with:
           name: shared-process-workspace-${{ github.sha }}-${{ github.run_attempt }}
           path: ${{ runner.temp }}/shared-process-workspace/
           if-no-files-found: error
           retention-days: 30
-      - name: Export exact compiled-image Native containment receipt
-        # JUnit is from the same full invocation: no second compile/test run.
-        # The official entire CI run must be green before release can consume
-        # this artifact. Missing/changed inputs or silent test skips are red.
-        if: needs.changes.outputs.heavy == 'true' && (matrix.os != 'ubuntu-latest' || github.event_name == 'workflow_dispatch' || github.event_name == 'pull_request') && !(github.event_name == 'workflow_dispatch' && inputs.workspace_test_mode == 'shared-process-twice')
+
+  # One sharded pipeline per OS (ci-test-suite.yml). Shard counts follow the
+  # measured serial test time of each OS (2026-11-12, run 43216596942: Linux 19
+  # min, macOS 24 min, Windows 25 min of test execution, on top of 30 to 65
+  # minutes of single-job compilation), so a shard is five to eight minutes of
+  # tests plus setup. Hosted macOS is the scarce pool, so it gets three shards.
+  # The caller names ("Sharded tests (...)") never equal the required
+  # "Test (windows-latest)" context, which test-windows-pr still produces on
+  # every pull request.
+  test-ubuntu:
+    name: Sharded tests (ubuntu-latest)
+    needs: changes
+    if: >-
+      needs.changes.outputs.heavy == 'true' &&
+      needs.changes.outputs.scoped != 'true' &&
+      (github.event_name == 'pull_request' || github.event_name == 'workflow_dispatch') &&
+      !(github.event_name == 'workflow_dispatch' && inputs.workspace_test_mode == 'shared-process-twice')
+    uses: ./.github/workflows/ci-test-suite.yml
+    with:
+      os: ubuntu-latest
+      shards: 4
+      shard_list: '[1,2,3,4]'
+      build_jobs: '2'
+      debug: line-tables-only
+
+  test-macos:
+    name: Sharded tests (macos-latest)
+    needs: changes
+    if: >-
+      needs.changes.outputs.heavy == 'true' &&
+      needs.changes.outputs.scoped != 'true' &&
+      (github.event_name == 'pull_request' || github.event_name == 'workflow_dispatch') &&
+      !(github.event_name == 'workflow_dispatch' && inputs.workspace_test_mode == 'shared-process-twice')
+    uses: ./.github/workflows/ci-test-suite.yml
+    with:
+      os: macos-latest
+      shards: 3
+      shard_list: '[1,2,3]'
+      build_jobs: '1'
+      debug: '0'
+
+  test-windows:
+    name: Sharded tests (windows-latest)
+    needs: changes
+    if: >-
+      needs.changes.outputs.heavy == 'true' &&
+      needs.changes.outputs.scoped != 'true' &&
+      (github.event_name == 'pull_request' || github.event_name == 'workflow_dispatch') &&
+      !(github.event_name == 'workflow_dispatch' && inputs.workspace_test_mode == 'shared-process-twice')
+    uses: ./.github/workflows/ci-test-suite.yml
+    with:
+      os: windows-latest
+      shards: 4
+      shard_list: '[1,2,3,4]'
+      build_jobs: '2'
+      debug: '0'
+
+  portable-boundaries:
+    name: Portable boundaries
+    needs: changes
+    if: >-
+      needs.changes.outputs.heavy == 'true' &&
+      needs.changes.outputs.scoped != 'true' &&
+      (github.event_name == 'pull_request' || github.event_name == 'workflow_dispatch')
+    timeout-minutes: 45
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v7
+      - uses: dtolnay/rust-toolchain@stable
+      - uses: mozilla-actions/[email protected]
+        id: sccache
+        continue-on-error: true
+        # Every run reads the GitHub Actions cache main wrote; only main writes (see Enable sccache).
+      - name: Enable sccache
+        if: steps.sccache.outcome == 'success'
+        shell: bash
+        run: |
+          echo "SCCACHE_GHA_ENABLED=true" >> "${GITHUB_ENV}"
+          if [ "${GITHUB_REF}" = refs/heads/main ]; then rw=READ_WRITE; else rw=READ_ONLY; fi
+          echo "SCCACHE_GHA_RW_MODE=${rw}" >> "${GITHUB_ENV}"
+          echo "RUSTC_WRAPPER=sccache" >> "${GITHUB_ENV}"
+          echo "SCCACHE_IGNORE_SERVER_IO_ERROR=1" >> "${GITHUB_ENV}"
+      - name: Install Linux system dependencies
+        run: |
+          bash scripts/ci-apt-install.sh libdbus-1-dev pkg-config bubblewrap apparmor-profiles
+          sh scripts/prepare-linux-test-sandbox.sh
+      - uses: Swatinem/rust-cache@v2
+        with:
+          cache-bin: false
+          save-if: ${{ github.ref == 'refs/heads/main' }}
+      - uses: taiki-e/install-action@nextest
+      - uses: actions/[email protected]
+        with:
+          node-version: 22
+      - uses: oven-sh/setup-bun@0c5077e51419868618aeaa5fe8019c62421857d6 # v2.2
+        with:
+          bun-version: 1.4.0
+      - name: Check portable config policy boundary
+        shell: bash
+        run: |
+          python3 -B scripts/test_check_command_config_proof.py
+          sh scripts/check-portable-config.sh
+      - name: Check portable core navigation boundary
+        shell: bash
+        run: |
+          python3 -B scripts/test_check_command_core_navigation_proof.py
+          sh scripts/check-portable-core-navigation.sh
+
+  scoped-test:
+    # workflow_dispatch with `crate` and/or `filter`: one package or one nextest
+    # filterset on the chosen OSes (default ubuntu-latest) instead of the full
+    # suite. `changes` skips every other job for such a run. The build uses the
+    # same flags as the full build; only `-p` and `-E` narrow what runs.
+    name: Scoped test (${{ matrix.os }})
+    needs: changes
+    if: needs.changes.outputs.scoped == 'true'
+    timeout-minutes: 90
+    runs-on: ${{ matrix.os }}
+    strategy:
+      fail-fast: false
+      matrix:
+        os: ${{ fromJSON(needs.changes.outputs.scoped_os) }}
+    env:
+      CARGO_BUILD_JOBS: ${{ matrix.os == 'macos-latest' && '1' || '2' }}
+      CARGO_PROFILE_DEV_DEBUG: ${{ matrix.os == 'ubuntu-latest' && 'line-tables-only' || '0' }}
+      CARGO_PROFILE_TEST_DEBUG: ${{ matrix.os == 'ubuntu-latest' && 'line-tables-only' || '0' }}
+      CODEWHALE_EXT_HOST_TESTS: '1'
+      CODEWHALE_EXT_HOST_BUN_TESTS: '1'
+      RUST_MIN_STACK: '16777216'
+      SCOPE_CRATE: ${{ inputs.crate }}
+      SCOPE_FILTER: ${{ inputs.filter }}
+    steps:
+      - uses: actions/checkout@v7
+      - name: Point terminal failure diagnostics at the runner temp directory
+        shell: bash
+        run: echo "QA_PTY_DIAGNOSTICS_DIR=${RUNNER_TEMP}/pty-failures" >> "${GITHUB_ENV}"
+      - uses: dtolnay/rust-toolchain@stable
+      - name: Install Linux system dependencies
+        if: matrix.os == 'ubuntu-latest'
+        run: |
+          bash scripts/ci-apt-install.sh libdbus-1-dev pkg-config bubblewrap apparmor-profiles
+          sh scripts/prepare-linux-test-sandbox.sh
+      - uses: Swatinem/rust-cache@v2
+        with:
+          cache-bin: false
+          save-if: false
+      - uses: taiki-e/install-action@nextest
+      - uses: actions/[email protected]
+        with:
+          node-version: 22
+      - uses: oven-sh/setup-bun@0c5077e51419868618aeaa5fe8019c62421857d6 # v2.2
+        with:
+          bun-version: 1.4.0
+      - name: Build matching compiled-host input for required Native containment receipts
+        shell: bash
+        run: bash scripts/ci/compiled-host-test-input.sh
+      - name: Build canonical executable for acceptance tests
+        # Acceptance tests in the TUI, engine and CLI crates, and any run that
+        # names no crate, look for the codewhale binary next to the test
+        # binaries; a narrow run on another crate does not need it.
+        if: env.SCOPE_CRATE == '' || env.SCOPE_CRATE == 'codewhale-tui' || env.SCOPE_CRATE == 'codewhale-engine' || env.SCOPE_CRATE == 'codewhale-cli'
+        shell: bash
+        run: cargo build -p codewhale-cli --bin codewhale --all-features --locked
+      - name: Run the scoped tests
         shell: bash
         run: |
           set -euo pipefail
-          test "$(git rev-parse HEAD)" = "$GITHUB_SHA"
-          git diff --exit-code
-          git diff --cached --exit-code
-          python3 scripts/compiled-host-native-receipt.py \
-            --junit target/nextest/ci/junit.xml \
-            --compiled-image "$CODEWHALE_COMPILED_HOST_TEST_BINARY" \
-            --bundle crates/engine/extension-host/dist/codewhale-extension-host.mjs \
-            --bun "$CODEWHALE_COMPILED_HOST_TEST_BUN" --source-sha "$GITHUB_SHA" \
-            --image-platform "$CODEWHALE_COMPILED_HOST_TEST_PLATFORM" --image-arch "$CODEWHALE_COMPILED_HOST_TEST_ARCH" \
-            --host-sha256 "$CODEWHALE_COMPILED_HOST_TEST_SHA256" \
-            --runtime-sha256 "$CODEWHALE_COMPILED_HOST_TEST_RUNTIME_SHA256" \
-            --bundle-sha256 "$CODEWHALE_COMPILED_HOST_TEST_BUNDLE_SHA256" \
-            --output "$RUNNER_TEMP/native-compiled-host"
-      - name: Retain exact compiled image and Native receipt
-        if: needs.changes.outputs.heavy == 'true' && (matrix.os != 'ubuntu-latest' || github.event_name == 'workflow_dispatch' || github.event_name == 'pull_request') && !(github.event_name == 'workflow_dispatch' && inputs.workspace_test_mode == 'shared-process-twice')
+          args=(--all-features --locked --profile ci --no-tests=fail)
+          if [ -n "${SCOPE_CRATE}" ]; then args+=(-p "${SCOPE_CRATE}"); else args+=(--workspace); fi
+          if [ -n "${SCOPE_FILTER}" ]; then args+=(-E "${SCOPE_FILTER}"); fi
+          sh scripts/with-hermetic-test-home.sh cargo nextest run "${args[@]}"
+      - name: Retain sealed terminal failure diagnostics
+        if: failure()
         uses: actions/upload-artifact@043fb46d1a93c77aae656e7c1c64a875d1fc6a0a # v7
         with:
-          name: native-compiled-host-${{ runner.os }}-${{ runner.arch }}
-          path: ${{ runner.temp }}/native-compiled-host/
-          if-no-files-found: error
-          retention-days: 30
-      - name: Run doctests
-        # The shared-process command already includes every workspace doctest.
-        if: needs.changes.outputs.heavy == 'true' && (matrix.os != 'ubuntu-latest' || github.event_name == 'workflow_dispatch' || github.event_name == 'pull_request') && !(github.event_name == 'workflow_dispatch' && inputs.workspace_test_mode == 'shared-process-twice')
-        shell: bash
-        run: sh scripts/with-hermetic-test-home.sh cargo test --workspace --all-features --locked --doc
-        env:
-          RUSTC_WRAPPER: ${{ matrix.os != 'windows-latest' && env.RUSTC_WRAPPER || '' }}
-          RUST_MIN_STACK: '16777216'
-      # The Ubuntu lint lane validates non-RSS backlog fields. Run the same
-      # source-bound measurement on macOS so loss or growth of RSS evidence
-      # fails closed instead of becoming an unsupported-field skip.
-      # Only on the self-hosted Mac: there it reuses the warm build. Every
-      # hosted Mac leg (fork PRs, or any PR with CW_SELF_HOSTED_MAC off) runs
-      # it in the separate `macos-budget` job with its own timeout, because
-      # on a cold hosted Mac it pushed Test past 90 minutes.
-      - name: Check persistence-backlog RSS budget
-        if: needs.changes.outputs.heavy == 'true' && matrix.os == 'macos-latest' && needs.changes.outputs.trusted == 'true' && vars.CW_SELF_HOSTED_MAC == 'true'
-        run: python3 scripts/check-persistence-backlog-budget.py
-      - name: Lockfile drift guard
-        if: needs.changes.outputs.heavy == 'true' && (matrix.os != 'ubuntu-latest' || github.event_name == 'workflow_dispatch' || github.event_name == 'pull_request')
-        run: git diff --exit-code -- Cargo.lock
-      - name: Run Offline Eval Harness
-        # The eval harness is OS-independent prompt/composition checking;
-        # running it once (on the faster macOS leg, warm from the test build)
-        # instead of once per desktop OS keeps the coverage while taking
-        # ~2min off the Windows critical path. Self-hosted Mac only; hosted
-        # Mac legs run it in `macos-budget` (see the RSS step above).
-        if: needs.changes.outputs.heavy == 'true' && matrix.os == 'macos-latest' && needs.changes.outputs.trusted == 'true' && vars.CW_SELF_HOSTED_MAC == 'true'
-        run: cargo run -p codewhale-cli --bin codewhale --all-features -- eval
-      - name: sccache stats
-        if: needs.changes.outputs.heavy == 'true' && (matrix.os != 'ubuntu-latest' || github.event_name == 'workflow_dispatch' || github.event_name == 'pull_request') && steps.sccache.outcome == 'success'
-        continue-on-error: true
-        shell: bash
-        run: sccache --show-stats
-      - name: Linux test location (CNB)
-        if: needs.changes.outputs.heavy == 'true' && matrix.os == 'ubuntu-latest' && github.event_name != 'workflow_dispatch' && github.event_name != 'pull_request'
-        run: echo "Linux workspace tests run on CNB for non-PR release/main pushes; pull requests run directly on Ubuntu."
-      - name: Release this machine's build lock (self-hosted)
-        if: always() && runner.environment == 'self-hosted'
-        shell: bash
-        run: rm -f "$RUNNER_TEMP/cw-build-lock.hold"
+          name: terminal-failures-scoped-${{ runner.os }}-${{ github.sha }}-${{ github.run_attempt }}
+          path: ${{ runner.temp }}/pty-failures/
+          include-hidden-files: true
+          if-no-files-found: ignore
+          retention-days: 7
 
   test-windows-pr:
     name: Test (windows-latest)
@@ -1239,7 +1229,7 @@ jobs:
             echo "## Test (windows-latest) on a pull request"
             echo
             echo "Compiled every workspace target for Windows with warnings denied and ran the installer PATH helper tests."
-            echo "The Windows and macOS test suites run in final validation, before a release, not on pull requests."
+            echo "The full Windows test suite runs beside this check as Sharded tests (windows-latest); it is not a required check."
           } >> "$GITHUB_STEP_SUMMARY"
 
   macos-budget:
@@ -1252,15 +1242,15 @@ jobs:
     # macOS leg runs on the self-hosted Mac (trusted event and
     # CW_SELF_HOSTED_MAC == 'true'), it runs these two steps on the warm
     # build instead. The name keeps "(fork PR)" so check contexts stay stable.
-    # Hosted macOS allows only five concurrent jobs per account, and this job
-    # doubled every pull request's claim on them, queueing the required Test
-    # (macos-latest) leg for hours. So pull requests skip it and every push to
-    # main still runs it; a regression shows up on the merge that caused it.
+    # Hosted macOS allows only five concurrent jobs per account, and the macOS
+    # test build and shards already take most of them, so pull requests skip
+    # this job and every dispatch (main-landing issues one per push to main)
+    # still runs it; a regression shows up on the merge that caused it. The
+    # sharded macOS suite never runs on the self-hosted Mac, so these gates no
+    # longer depend on CW_SELF_HOSTED_MAC.
     name: macOS budget and eval (fork PR)
     needs: changes
-    # The explicit Ubuntu-only qualification has no warm Test macOS leg, so
-    # retain these gates on the hosted fallback even when self-hosting is on.
-    if: needs.changes.outputs.heavy == 'true' && github.event_name != 'pull_request' && ((github.event_name == 'workflow_dispatch' && inputs.workspace_test_mode == 'shared-process-twice') || !(needs.changes.outputs.trusted == 'true' && vars.CW_SELF_HOSTED_MAC == 'true'))
+    if: needs.changes.outputs.heavy == 'true' && needs.changes.outputs.scoped != 'true' && github.event_name != 'pull_request'
     # Cold hosted builds consumed 73 minutes before eval could start (F run
     # 37497498829). Match Test's compile allowance; RSS/eval limits stay intact.
     timeout-minutes: 210
@@ -1282,9 +1272,10 @@ jobs:
   npm-wrapper-smoke:
     name: npm wrapper smoke
     needs: changes
-    if: github.event_name == 'workflow_dispatch'
-    # Same ternary rationale as the Test job: light legs only echo, so keep
-    # them off macOS/Windows runners. On pull_request the matrix is
+    if: needs.changes.outputs.scoped != 'true' && (github.event_name == 'pull_request' || github.event_name == 'workflow_dispatch')
+    # Light legs only echo, so keep them off macOS/Windows runners. The
+    # ternary is safe: matrix.os is always a non-empty literal. On
+    # pull_request the matrix is
     # ubuntu-only, so the required "npm wrapper smoke (ubuntu-latest)"
     # context is unaffected. Heavy pull requests execute the Ubuntu smoke
     # here; their branches may not be mirrored to CNB.
@@ -1310,13 +1301,15 @@ jobs:
       - uses: mozilla-actions/[email protected]
         id: sccache
         continue-on-error: true
-        # Main-only GitHub Actions cache backend; see the Safety gate job.
-        if: needs.changes.outputs.heavy == 'true' && (matrix.os != 'ubuntu-latest' || github.event_name == 'workflow_dispatch' || github.event_name == 'pull_request') && github.ref == 'refs/heads/main'
+        # Every run reads the GitHub Actions cache main wrote; only main writes (see Enable sccache).
+        if: needs.changes.outputs.heavy == 'true' && (matrix.os != 'ubuntu-latest' || github.event_name == 'workflow_dispatch' || github.event_name == 'pull_request')
       - name: Enable sccache
-        if: needs.changes.outputs.heavy == 'true' && (matrix.os != 'ubuntu-latest' || github.event_name == 'workflow_dispatch' || github.event_name == 'pull_request') && steps.sccache.outcome == 'success'
+        if: needs.changes.outputs.heavy == 'true' && (matrix.os != 'ubuntu-latest' || github.event_name == 'workflow_dispatch' || github.event_name == 'pull_request') && steps.sccache.outcome == 'success' && matrix.os != 'windows-latest'
         shell: bash
         run: |
           echo "SCCACHE_GHA_ENABLED=true" >> "${GITHUB_ENV}"
+          if [ "${GITHUB_REF}" = refs/heads/main ]; then rw=READ_WRITE; else rw=READ_ONLY; fi
+          echo "SCCACHE_GHA_RW_MODE=${rw}" >> "${GITHUB_ENV}"
           echo "RUSTC_WRAPPER=sccache" >> "${GITHUB_ENV}"
           echo "SCCACHE_IGNORE_SERVER_IO_ERROR=1" >> "${GITHUB_ENV}"
       - uses: actions/[email protected]
@@ -1358,8 +1351,14 @@ jobs:
   mobile-smoke:
     name: Mobile runtime smoke
     needs: changes
-    # Manual (workflow_dispatch) only; the release candidate builds android-arm64.
-    if: github.event_name == 'workflow_dispatch' && needs.changes.outputs.heavy == 'true'
+    # Not a required PR context. Pull requests run it only when the mobile
+    # runtime surface changed (see the `mobile` filter above); every dispatch,
+    # which main-landing issues for each push to main, runs it unconditionally
+    # as the pre-release safety net.
+    if: >-
+      needs.changes.outputs.scoped != 'true' &&
+      needs.changes.outputs.heavy == 'true' &&
+      (github.event_name != 'pull_request' || needs.changes.outputs.mobile == 'true')
     timeout-minutes: 30
     runs-on: ubuntu-latest
     steps:
@@ -1367,14 +1366,15 @@ jobs:
       - uses: dtolnay/rust-toolchain@stable
       - uses: mozilla-actions/[email protected]
         id: sccache
-        # Main-only GitHub Actions cache backend; see the Safety gate job.
-        if: github.ref == 'refs/heads/main'
+        # Every run reads the GitHub Actions cache main wrote; only main writes (see Enable sccache).
         continue-on-error: true
       - name: Enable sccache
         if: steps.sccache.outcome == 'success'
         shell: bash
         run: |
           echo "SCCACHE_GHA_ENABLED=true" >> "${GITHUB_ENV}"
+          if [ "${GITHUB_REF}" = refs/heads/main ]; then rw=READ_WRITE; else rw=READ_ONLY; fi
+          echo "SCCACHE_GHA_RW_MODE=${rw}" >> "${GITHUB_ENV}"
           echo "RUSTC_WRAPPER=sccache" >> "${GITHUB_ENV}"
           echo "SCCACHE_IGNORE_SERVER_IO_ERROR=1" >> "${GITHUB_ENV}"
       - name: Install Linux system dependencies
@@ -1401,7 +1401,7 @@ jobs:
   actionlint:
     name: Workflow lint
     needs: changes
-    if: needs.changes.outputs.actions == 'true'
+    if: needs.changes.outputs.actions == 'true' && needs.changes.outputs.scoped != 'true'
     timeout-minutes: 15
     runs-on: ubuntu-latest
     steps:
@@ -1418,7 +1418,8 @@ jobs:
   # Check documentation builds without warnings
   docs:
     name: Documentation
-    if: github.event_name == 'schedule' || github.event_name == 'workflow_dispatch'
+    needs: changes
+    if: needs.changes.outputs.scoped != 'true' && (github.event_name == 'schedule' || github.event_name == 'workflow_dispatch')
     timeout-minutes: 60
     runs-on: ubuntu-latest
     steps:
@@ -1439,17 +1440,17 @@ jobs:
   cargo-deny:
     name: cargo-deny
     needs: changes
-    if: github.event_name == 'workflow_dispatch'
+    if: github.event_name == 'workflow_dispatch' && needs.changes.outputs.scoped != 'true'
     uses: ./.github/workflows/cargo-deny.yml
 
   security-audit:
     name: Security audit
     needs: changes
-    if: github.event_name == 'workflow_dispatch'
+    if: github.event_name == 'workflow_dispatch' && needs.changes.outputs.scoped != 'true'
     uses: ./.github/workflows/security-audit.yml
 
   ohos:
     name: OpenHarmony
     needs: changes
-    if: github.event_name == 'workflow_dispatch'
+    if: github.event_name == 'workflow_dispatch' && needs.changes.outputs.scoped != 'true'
     uses: ./.github/workflows/ohos.yml

.github/workflows/nightly.yml

@@ -73,7 +73,7 @@ jobs:
         id: sccache
         continue-on-error: true
       - name: Enable sccache
-        if: steps.sccache.outcome == 'success'
+        if: steps.sccache.outcome == 'success' && runner.os != 'Windows'
         shell: bash
         run: |
           {

.github/workflows/release-artifacts.yml

@@ -178,7 +178,7 @@ jobs:
         id: sccache
         continue-on-error: true
       - name: Enable sccache
-        if: steps.sccache.outcome == 'success'
+        if: steps.sccache.outcome == 'success' && runner.os != 'Windows'
         shell: bash
         run: |
           {

.mailmap

@@ -14,3 +14,17 @@ Hunter Bown <[email protected]> Cursor Agent <[email protected]>
 Hunter Bown <[email protected]> cursoragent <[email protected]>
 Hunter Bown <[email protected]> gemini-code-assist[bot] <gemini-code-assist[bot]@users.noreply.github.com>
 Hunter Bown <[email protected]> gemini-code-assist[bot] <956858+gemini-code-assist[bot]@users.noreply.github.com>
+ddlw <[email protected]> <[email protected]>
+dmitri-0 <[email protected]> <[email protected]>
+Paulo Aboim Pinto <[email protected]> <[email protected]>
+Paulo Aboim Pinto <[email protected]> <[email protected]>
+Paulo Aboim Pinto <[email protected]> <[email protected]>
+AiurArtanis <[email protected]> <[email protected]>
+Apeiron <[email protected]> <[email protected]>
+ChaceLyee2101 <[email protected]> <[email protected]>
+movic <[email protected]> <[email protected]>
+merchloubna70-dot <[email protected]> <[email protected]>
+merchloubna70-dot <[email protected]> <[email protected]>
+luola <[email protected]> <[email protected]>
+SONG ZHENRUI <[email protected]> <[email protected]>
+xiluoduyu <[email protected]> <[email protected]>

AGENTS.md

@@ -259,9 +259,16 @@ prefer merging the contributor's PR itself). When a contributor's work
 lands as our commit, the contributor is that commit's **author** (`git commit
 --author` with their GitHub-linked address), it carries `Harvested from PR #N by
 @handle`, and maintainers or agents go in `Co-authored-by`. GitHub's contributor
-list counts commit authors only: about 200 harvested commits authored by
-maintainer identities left their contributors off it, which is why
-`docs/CONTRIBUTORS.md` carries the full wall (`scripts/contributor-wall.py`).
+count takes commit authors and the `Co-authored-by` trailers in a commit's final
+trailer block, and only addresses linked to an account: a trailer cut off by a
+blank line, an unlinked email, or a noreply address carrying someone else's
+numeric id credits nobody or the wrong person. Harvested commits authored by
+maintainer identities without such a trailer left their contributors off it,
+which is why `docs/CONTRIBUTORS.md` carries the full wall
+(`scripts/contributor-wall.py`). Credit lost in history that cannot change is
+restored by a later commit with one trailer per person, naming each pull
+request; `scripts/check-contributor-credit.py` fails the release window when a
+harvest, a trailer or a noreply id would leave someone uncounted.
 A maintainer closes their PR by hand with a thank-you naming the commit;
 nothing closes a PR automatically. Canonical human identities come from
 `.github/AUTHOR_MAP`.

CHANGELOG.md

@@ -7,6 +7,11 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 ## [Unreleased]
 
+- **Renaming sessions with Chinese titles:** the session picker no longer rejects a Chinese, Japanese, Korean or emoji title of more than about 33 characters as "Title must be 1–100 characters". It now counts characters, as the message says and as `/rename` and the Runtime API already did, so any title up to 100 characters is accepted. Thanks [@luochen211](https://github.com/luochen211) (PR 7322).
+- **Non-ASCII file names:** `/diff`, the `@git` and `@diff` mentions and the Runtime API workspace diff now show paths with Chinese, emoji and other non-ASCII characters as written instead of as quoted octal escapes, and the turn context no longer lists the workspace root as an empty "Active paths" entry. Thanks [@luochen211](https://github.com/luochen211) (PR 7317).
+- **Session names after a crash, and keep-forever snapshot retention:** a session name saved right after a crash that cut off the previous session-index line is no longer lost, because the new entry now starts on its own line; and a very large `[snapshots] max_age_days` (such as `9223372036854775807` for "keep forever") no longer wraps around and deletes every undo snapshot at session start. Thanks [@luochen211](https://github.com/luochen211) (PR 7316).
+- **Blank environment variables:** exporting a variable empty (for example `CODEWHALE_MODEL=` from a compose `${VAR:-}` template) no longer hides its legacy alias. `DEEPSEEK_MODEL`, `DEEPSEEK_BASE_URL`, `KIMI_BASE_URL`, `DEEPSEEK_FORCE_HTTP1`, `DEEPSEEK_STREAM_OPEN_TIMEOUT_SECS` and the other provider alias variables now apply as they do when the primary is unset. Thanks [@luochen211](https://github.com/luochen211) (PR 7315).
+- **Qwen Code as an external agent:** Codewhale now finds an installed Qwen Code CLI (`qwen`, 0.6.1 or newer) and can run it as a session over ACP with `qwen --acp`. `GET /v1/agents/external` lists it, an older install is reported as needing `npm i -g @qwen-code/qwen-code@latest`, and sign-in stays inside Qwen Code (`/auth`). Thanks [@MannXo](https://github.com/MannXo) (#7293).
 - **README and contributor credit:** the repository front page now opens the way the website does: the line "The model can be anyone's. The agent should be yours.", one sentence on what Codewhale is, the install command and one picture, with reference detail moved into the linked guides; the 18 translated READMEs follow the same structure and say they are machine-drafted. The contributor wall and the website now count the same 433 people (407 GitHub accounts and the commit authors whose email is not linked to an account), where the website had counted accounts only. `DEPENDENCY_SECURITY.md` moved into `docs/dependency-maintenance.md`, and `.github/SUPPORT.md` says where to ask for help.
 - **Website Plugins page:** The website's Plugins page now describes what the released version does, and a labelled section says what 0.10.3 adds: one Plugins place, Standard and Creator mode, and the Permissions control.
 - **Plugin docs and plugin-creator:** Docs now say everything you add is a plugin. The bundled plugin-creator skill no longer tells an agent never to trust or enable a plugin (in Full Access it may, for one it wrote and installed from its project), points code and screen plugins at cordis-plugin-development, and an untouched copy from 0.10.2 now upgrades. The Chinese extension and Runtime API docs now say editor ACP sessions get plugin tools and configured MCP servers.
@@ -669,7 +674,7 @@ work is declared complete here. Items moved to 0.10.4 by name are marked below.
 - **Windows uninstall removes the receipt and update leftovers.** The installer's uninstaller also deletes `.codewhale-install.sha256` and `codewhale.old-*`/`codew.old-*` left by a self-update, so `bin` can be removed. Refs #7091.
 - **Windows `/undo` matches recorded file paths regardless of case.** A write recorded under a different drive-letter or directory case, or a `\\?\` prefix, is no longer dropped from the restore plan. Refs #6970.
 - **Windows: Fleet and metrics read live files again.** Replaying a Fleet ledger and reading audit or receipt records no longer fail with a sharing violation while their writer is still open; the same regular-file, link and identity checks apply. Plugin enablement now says which snapshot check refused (stage root, content identity or final check) instead of one generic message. Generated protocol and SDK exports keep LF line endings in Windows checkouts.
-- **Sessions in one folder no longer leave each other's sub-agents read-only or private:** a second session works beside another one, stale or suspended included, the older sub-agent store upgrades right away, and a store that cannot be opened is retried on the next sub-agent change instead of needing a restart. While another Codewhale session, a suspended one included, is open in a folder the desktop app or `serve` was asked to work in (other than the folder it started in), that Runtime still refuses sub-agents that would edit the folder directly; isolated worktrees are unaffected. A sub-agent store that cannot be opened shows its refusal once instead of repeating cleanup warnings on every sidebar refresh. Restart any session started with Codewhale 0.10.2 or older: once the folder upgrades it can no longer save sub-agent state (Refs #7232).
+- **A held coordination lock no longer makes a new session read-only.** When another session owns the delegated-agent store, the new session starts on private state and says so, instead of refusing every sub-agent with “Failed to clean up sub-agents”.
 - **A dev server can outlive the turn in an interactive session.** `persist:true` now works in the TUI and desktop when the thread has Full Access; restricted postures and Fleet/sub-agents still refuse it.
 - **`codewhale exec --add-dir <path>`** (repeatable) and `addDirs` on `POST /v1/threads` let the file tools read and write a sibling directory, a second checkout or a shared scratch root without turning on Full Access.
 - **Unsettled usage is persisted in production, not only in tests.** The interactive sink now records a legacy usage receipt when a provider response has no settlement, so children and turns stop reporting settlement failure.
@@ -706,6 +711,12 @@ work is declared complete here. Items moved to 0.10.4 by name are marked below.
   composer. Real Pinyin input on macOS and Windows is not yet confirmed.
   Thanks [@cmdcorp6534](https://github.com/cmdcorp6534) for the report
   (Refs #2323).
+- **Sub-agents beside an older session:** while a Codewhale 0.10.2 session
+  still uses a folder's older delegated-agent store, a 0.10.3 session in the
+  same folder keeps working sub-agents with private state that is not saved,
+  instead of refusing every sub-agent action. The older session's files are
+  left untouched, and the store upgrades when a session starts after the older
+  one exits (Refs #7232).
 - **Fleet members:** a replayed member start now matches only within its own
   session, so two sessions that reuse an operation key no longer receive each
   other's agent.
@@ -734,6 +745,12 @@ work is declared complete here. Items moved to 0.10.4 by name are marked below.
   bare self-package exports resolve inside the selected package directory.
   Invalid or incompatible ranges and paths outside the reviewed package still
   refuse. Real published-plugin activation remains unqualified.
+- **Coordination lock diagnostics (#7232):** stale lock-file contents no longer
+  identify an owner PID. A refused legacy admission keeps its reason in Work
+  and shows one notice per cause instead of repeating cleanup warnings on
+  every sidebar refresh. Legacy migration still requires the actual holder to
+  exit and the refused owning Runtime to restart; locks and retained state
+  remain protected.
 - **A local proxy that stops mid-session no longer strands the session (#7231):** the model, catalog, search, vision, finance, MCP-registry, profile and remote-control clients watch a loopback proxy port and go direct while it refuses connections, then resume proxying when it returns. Only a definite refusal from every resolved address changes the route; remote proxies, `NO_PROXY` matches, credentials and the existing precedence are unchanged. Set `CODEWHALE_LOCAL_PROXY_FALLBACK=0` to keep the old build-time snapshot.
 - **Proxy diagnostic privacy ([#7233](https://github.com/codewhale-hq/codewhale/issues/7233)):** malformed MCP proxy warnings omit configured values; local-proxy log labels mask userinfo and omit URL paths, queries and fragments. Routing is unchanged. All 23 existing proxy tests passed on macOS; runtime log capture and other platforms remain unqualified. Thanks [@SparkofSpike](https://github.com/SparkofSpike) for the report.
 - **Fork, undo and retry work after a workspace drift notice (#7234):** the history-boundary walks now treat a whole `<context_update>` envelope as engine-owned, so a session that received one can be cut at every saved turn. A person quoting a notice or a truncated envelope still counts as their own prompt.

CONTRIBUTING.md

@@ -165,7 +165,7 @@ before starting a change; preserve any local work when updating your branch.
 
 - Write tests for new functionality
 - Run the tests near your change (see [Fast local loop](#fast-local-loop));
-  your pull request runs the Linux workspace suite and a Windows build for
+  your pull request runs the full test suite on Linux, macOS and Windows for
   you (see [What happens after you open a pull request](#what-happens-after-you-open-a-pull-request))
 - Colocate unit tests beside the code they cover (standard Rust `#[cfg(test)]`
   modules), and add integration tests under the owning crate's `tests/`
@@ -174,8 +174,8 @@ before starting a change; preserve any local work when updating your branch.
 
 ### Pre-push verification
 
-These are the Rust commands your pull request runs on Linux (the `Lint` and
-`Test (ubuntu-latest)` checks). You do not need all of them before every
+These are the Rust commands your pull request runs on Linux (the `Lint` check
+and the Linux shards of `Sharded tests (ubuntu-latest)`). You do not need all of them before every
 push: run `cargo fmt`, then check and test the crates you touched (see
 [Fast local loop](#fast-local-loop)). Run the full set locally when a change
 spans many crates, or let the pull request run it for you:
@@ -215,7 +215,7 @@ cargo install cargo-nextest --locked      # once
 scripts/dev-test.sh tui
 scripts/dev-cargo.sh nextest run --workspace --all-features --locked
 
-# 4. The Linux suite your PR runs:
+# 4. The suite your PR runs on Linux (hosted CI runs it as parallel shards):
 cargo test --workspace --all-features --locked
 ```
 
@@ -340,6 +340,9 @@ maintainer to approve the first run for a brand-new GitHub account.
 | --- | --- | --- |
 | `Lint` (required) | `cargo fmt --check`, clippy on every target, and the fast drift checks (locales, READMEs, provider registry, dead-code and blocking-call budgets) | Run `cargo fmt --all` and the clippy command in [Pre-push verification](#pre-push-verification); the log names the file and the check |
 | `Test (windows-latest)` (required) | Compiles every workspace target for Windows with warnings denied, and the installer PATH helper tests | Usually a `cfg(windows)` path or an import only Windows sees; the log has the compiler error |
+| `Sharded tests (ubuntu-latest)`, `(macos-latest)`, `(windows-latest)` | The full workspace test suite, built once per OS and run as parallel shards of one nextest archive, plus the doctests; each OS ends with a `Suite check` that fails if the shards did not run every test exactly once | Open the failing `Shard k/N` job; the log names the test. The same test reproduces locally with `cargo nextest run -E 'test(name)'` |
+| `Safety gate`, `MSRV check (1.89)`, `npm wrapper smoke`, `Portable boundaries` | The hermetic safety and authorization tests, a build on the minimum Rust version, the npm wrapper install smoke and the portable-core boundary checks | The log names the failing test or the compiler error |
+| `Mobile runtime smoke` | The `codewhale serve --mobile` HTTP and event-stream smoke, only when the mobile runtime surface changed | The log shows the failing request |
 | `Version drift` | Version files, the `feat:` release-note rule, release helper tests and contributor credit | The credit step only reports: a maintainer adds your credit when the PR merges |
 
 Smaller checks cover the VS Code extension, integrations, the extension host
@@ -347,11 +350,15 @@ and plugin conversion, plus workflow lint when you change `.github/workflows`.
 A documentation-only change skips the Rust builds but still reports the
 required checks, so it can merge.
 
-Your pull request does not run the test suites, the safety gate, MSRV, the npm
-wrapper smoke, the slow runtime-contract and persistence-backlog measurements,
-the dependency audits or the OpenHarmony build. Maintainers run those once per
-release, on the exact commit being released. If that run finds a problem your
-change introduced, we fix it on `main`; we will not ask you to chase it.
+Only `Lint` and `Test (windows-latest)` are required. Every other job above is
+informational for a contributor's pull request: a red one tells you and the
+maintainers something, and never blocks the merge. Your pull request does not
+run the slow runtime-contract and persistence-backlog measurements, the macOS
+budget and eval job, the dependency audits or the OpenHarmony build; those run
+on `main` and for each release, on the exact commit. If one of them finds a
+problem your change introduced, we fix it on `main`; we will not ask you to
+chase it. A newer push to your branch cancels the older run of the same pull
+request, so the runners always work on your latest commit.
 
 A red check that your change did not cause is not your job to fix. We rerun it
 or fix it on `main`, and we never close a PR because CI is slow or flaky. Your

Cargo.lock

@@ -1179,7 +1179,6 @@ dependencies = [
  "agent-client-protocol",
  "agent-client-protocol-schema",
  "anyhow",
- "arboard",
  "async-stream",
  "async-trait",
  "axum",

crates/config/src/env_overrides.rs

@@ -140,48 +140,17 @@ impl EnvRuntimeOverrides {
         Self {
             provider,
             provider_source,
-            model: std::env::var("CODEWHALE_MODEL")
-                .or_else(|_| std::env::var("DEEPSEEK_MODEL"))
-                .or_else(|_| std::env::var("DEEPSEEK_DEFAULT_TEXT_MODEL"))
-                .ok()
-                .filter(|v| !v.trim().is_empty()),
-            volcengine_model: std::env::var("VOLCENGINE_MODEL")
-                .or_else(|_| std::env::var("VOLCENGINE_ARK_MODEL"))
-                .ok()
-                .filter(|v| !v.trim().is_empty()),
-            wanjie_ark_model: std::env::var("WANJIE_ARK_MODEL")
-                .or_else(|_| std::env::var("WANJIE_MODEL"))
-                .or_else(|_| std::env::var("WANJIE_MAAS_MODEL"))
-                .ok()
-                .filter(|v| !v.trim().is_empty()),
-            openrouter_model: std::env::var("OPENROUTER_MODEL")
-                .ok()
-                .filter(|v| !v.trim().is_empty()),
-            orcarouter_model: std::env::var("ORCAROUTER_MODEL")
-                .ok()
-                .filter(|v| !v.trim().is_empty()),
-            moonshot_model: std::env::var("MOONSHOT_MODEL")
-                .or_else(|_| std::env::var("KIMI_MODEL_NAME"))
-                .or_else(|_| std::env::var("KIMI_MODEL"))
-                .ok()
-                .filter(|v| !v.trim().is_empty()),
-            xiaomi_mimo_model: std::env::var("XIAOMI_MIMO_MODEL")
-                .or_else(|_| std::env::var("MIMO_MODEL"))
-                .ok()
-                .filter(|v| !v.trim().is_empty()),
-            xiaomi_mimo_mode: std::env::var("XIAOMI_MIMO_MODE")
-                .or_else(|_| std::env::var("MIMO_MODE"))
-                .ok()
-                .filter(|v| !v.trim().is_empty()),
-            novita_model: std::env::var("NOVITA_MODEL")
-                .ok()
-                .filter(|v| !v.trim().is_empty()),
-            fireworks_model: std::env::var("FIREWORKS_MODEL")
-                .ok()
-                .filter(|v| !v.trim().is_empty()),
-            arcee_model: std::env::var("ARCEE_MODEL")
-                .ok()
-                .filter(|v| !v.trim().is_empty()),
+            model: env_non_blank(&["CODEWHALE_MODEL", "DEEPSEEK_MODEL", "DEEPSEEK_DEFAULT_TEXT_MODEL"]),
+            volcengine_model: env_non_blank(&["VOLCENGINE_MODEL", "VOLCENGINE_ARK_MODEL"]),
+            wanjie_ark_model: env_non_blank(&["WANJIE_ARK_MODEL", "WANJIE_MODEL", "WANJIE_MAAS_MODEL"]),
+            openrouter_model: env_non_blank(&["OPENROUTER_MODEL"]),
+            orcarouter_model: env_non_blank(&["ORCAROUTER_MODEL"]),
+            moonshot_model: env_non_blank(&["MOONSHOT_MODEL", "KIMI_MODEL_NAME", "KIMI_MODEL"]),
+            xiaomi_mimo_model: env_non_blank(&["XIAOMI_MIMO_MODEL", "MIMO_MODEL"]),
+            xiaomi_mimo_mode: env_non_blank(&["XIAOMI_MIMO_MODE", "MIMO_MODE"]),
+            novita_model: env_non_blank(&["NOVITA_MODEL"]),
+            fireworks_model: env_non_blank(&["FIREWORKS_MODEL"]),
+            arcee_model: env_non_blank(&["ARCEE_MODEL"]),
             verbosity: env_non_blank(&["CODEWHALE_VERBOSITY", "DEEPSEEK_VERBOSITY"]),
             auth_mode: env_non_blank(&["CODEWHALE_AUTH_MODE", "DEEPSEEK_AUTH_MODE"]),
             log_level: env_non_blank(&["CODEWHALE_LOG_LEVEL", "DEEPSEEK_LOG_LEVEL"]),
@@ -216,9 +185,7 @@ impl EnvRuntimeOverrides {
                         None
                     }
                 }),
-            http_headers: std::env::var("CODEWHALE_HTTP_HEADERS")
-                .or_else(|_| std::env::var("DEEPSEEK_HTTP_HEADERS"))
-                .ok()
+            http_headers: env_non_blank(&["CODEWHALE_HTTP_HEADERS", "DEEPSEEK_HTTP_HEADERS"])
                 .and_then(|value| match parse_http_headers(&value) {
                     Ok(h) => Some(h),
                     Err(_) => {
@@ -227,290 +194,85 @@ impl EnvRuntimeOverrides {
                     }
                 })
                 .filter(|headers| !headers.is_empty()),
-            active_route_base_url: std::env::var("CODEWHALE_BASE_URL")
-                .or_else(|_| std::env::var("DEEPSEEK_BASE_URL"))
-                .ok()
-                .filter(|v| !v.trim().is_empty()),
-            deepseek_anthropic_base_url: std::env::var("DEEPSEEK_ANTHROPIC_BASE_URL")
-                .or_else(|_| std::env::var("DEEPSEEK_CLAUDE_BASE_URL"))
-                .ok()
-                .filter(|v| !v.trim().is_empty()),
-            nvidia_base_url: std::env::var("NVIDIA_NIM_BASE_URL")
-                .or_else(|_| std::env::var("NIM_BASE_URL"))
-                .or_else(|_| std::env::var("NVIDIA_BASE_URL"))
-                .ok()
-                .filter(|v| !v.trim().is_empty()),
-            openai_base_url: std::env::var("OPENAI_BASE_URL")
-                .ok()
-                .filter(|v| !v.trim().is_empty()),
-            atlascloud_base_url: std::env::var("ATLASCLOUD_BASE_URL")
-                .ok()
-                .filter(|v| !v.trim().is_empty()),
-            volcengine_base_url: std::env::var("VOLCENGINE_BASE_URL")
-                .or_else(|_| std::env::var("VOLCENGINE_ARK_BASE_URL"))
-                .or_else(|_| std::env::var("ARK_BASE_URL"))
-                .ok()
-                .filter(|v| !v.trim().is_empty()),
-            wanjie_ark_base_url: std::env::var("WANJIE_ARK_BASE_URL")
-                .or_else(|_| std::env::var("WANJIE_BASE_URL"))
-                .or_else(|_| std::env::var("WANJIE_MAAS_BASE_URL"))
-                .ok()
-                .filter(|v| !v.trim().is_empty()),
-            openrouter_base_url: std::env::var("OPENROUTER_BASE_URL")
-                .ok()
-                .filter(|v| !v.trim().is_empty()),
-            orcarouter_base_url: std::env::var("ORCAROUTER_BASE_URL")
-                .ok()
-                .filter(|v| !v.trim().is_empty()),
-            xiaomi_mimo_base_url: std::env::var("XIAOMI_MIMO_BASE_URL")
-                .or_else(|_| std::env::var("MIMO_BASE_URL"))
-                .ok()
-                .filter(|v| !v.trim().is_empty()),
-            novita_base_url: std::env::var("NOVITA_BASE_URL")
-                .ok()
-                .filter(|v| !v.trim().is_empty()),
-            fireworks_base_url: std::env::var("FIREWORKS_BASE_URL")
-                .ok()
-                .filter(|v| !v.trim().is_empty()),
-            siliconflow_base_url: std::env::var("SILICONFLOW_BASE_URL")
-                .ok()
-                .filter(|v| !v.trim().is_empty()),
-            siliconflow_model: std::env::var("SILICONFLOW_MODEL")
-                .ok()
-                .filter(|v| !v.trim().is_empty()),
-            arcee_base_url: std::env::var("ARCEE_BASE_URL")
-                .ok()
-                .filter(|v| !v.trim().is_empty()),
-            moonshot_base_url: std::env::var("MOONSHOT_BASE_URL")
-                .or_else(|_| std::env::var("KIMI_BASE_URL"))
-                .ok()
-                .filter(|v| !v.trim().is_empty()),
-            sglang_base_url: std::env::var("SGLANG_BASE_URL")
-                .ok()
-                .filter(|v| !v.trim().is_empty()),
-            vllm_base_url: std::env::var("VLLM_BASE_URL")
-                .ok()
-                .filter(|v| !v.trim().is_empty()),
-            ollama_base_url: std::env::var("OLLAMA_BASE_URL")
-                .ok()
-                .filter(|v| !v.trim().is_empty()),
-            ollama_cloud_base_url: std::env::var("OLLAMA_CLOUD_BASE_URL")
-                .ok()
-                .filter(|v| !v.trim().is_empty()),
-            ollama_cloud_model: std::env::var("OLLAMA_CLOUD_MODEL")
-                .ok()
-                .filter(|v| !v.trim().is_empty()),
-            huggingface_base_url: std::env::var("HUGGINGFACE_BASE_URL")
-                .or_else(|_| std::env::var("HF_BASE_URL"))
-                .ok()
-                .filter(|v| !v.trim().is_empty()),
-            huggingface_model: std::env::var("HUGGINGFACE_MODEL")
-                .or_else(|_| std::env::var("HF_MODEL"))
-                .ok()
-                .filter(|v| !v.trim().is_empty()),
-            modelscope_base_url: std::env::var("MODELSCOPE_BASE_URL")
-                .ok()
-                .filter(|v| !v.trim().is_empty()),
-            modelscope_model: std::env::var("MODELSCOPE_MODEL")
-                .ok()
-                .filter(|v| !v.trim().is_empty()),
-            together_base_url: std::env::var("TOGETHER_BASE_URL")
-                .ok()
-                .filter(|v| !v.trim().is_empty()),
-            together_model: std::env::var("TOGETHER_MODEL")
-                .ok()
-                .filter(|v| !v.trim().is_empty()),
-            qianfan_base_url: std::env::var("QIANFAN_BASE_URL")
-                .ok()
-                .filter(|v| !v.trim().is_empty())
-                .or_else(|| {
-                    std::env::var("BAIDU_QIANFAN_BASE_URL")
-                        .ok()
-                        .filter(|v| !v.trim().is_empty())
-                }),
-            qianfan_model: std::env::var("QIANFAN_MODEL")
-                .ok()
-                .filter(|v| !v.trim().is_empty())
-                .or_else(|| {
-                    std::env::var("BAIDU_QIANFAN_MODEL")
-                        .ok()
-                        .filter(|v| !v.trim().is_empty())
-                }),
-            openai_codex_base_url: std::env::var("OPENAI_CODEX_BASE_URL")
-                .or_else(|_| std::env::var("CODEX_BASE_URL"))
-                .ok()
-                .filter(|v| !v.trim().is_empty()),
-            openai_codex_model: std::env::var("OPENAI_CODEX_MODEL")
-                .or_else(|_| std::env::var("CODEX_MODEL"))
-                .ok()
-                .filter(|v| !v.trim().is_empty()),
-            anthropic_base_url: std::env::var("ANTHROPIC_BASE_URL")
-                .ok()
-                .filter(|v| !v.trim().is_empty()),
-            anthropic_model: std::env::var("ANTHROPIC_MODEL")
-                .ok()
-                .filter(|v| !v.trim().is_empty()),
-            openmodel_base_url: std::env::var("OPENMODEL_BASE_URL")
-                .ok()
-                .filter(|v| !v.trim().is_empty()),
-            openmodel_model: std::env::var("OPENMODEL_MODEL")
-                .ok()
-                .filter(|v| !v.trim().is_empty()),
-            zai_base_url: std::env::var("ZAI_BASE_URL")
-                .or_else(|_| std::env::var("Z_AI_BASE_URL"))
-                .or_else(|_| std::env::var("ZHIPU_BASE_URL"))
-                .or_else(|_| std::env::var("ZHIPUAI_BASE_URL"))
-                .or_else(|_| std::env::var("BIGMODEL_BASE_URL"))
-                .ok()
-                .filter(|v| !v.trim().is_empty()),
-            zai_model: std::env::var("ZAI_MODEL")
-                .or_else(|_| std::env::var("Z_AI_MODEL"))
-                .or_else(|_| std::env::var("ZHIPU_MODEL"))
-                .or_else(|_| std::env::var("ZHIPUAI_MODEL"))
-                .or_else(|_| std::env::var("BIGMODEL_MODEL"))
-                .or_else(|_| std::env::var("GLM_MODEL"))
-                .ok()
-                .filter(|v| !v.trim().is_empty()),
-            stepfun_base_url: std::env::var("STEPFUN_BASE_URL")
-                .or_else(|_| std::env::var("STEP_BASE_URL"))
-                .ok()
-                .filter(|v| !v.trim().is_empty()),
-            stepfun_model: std::env::var("STEPFUN_MODEL")
-                .or_else(|_| std::env::var("STEP_MODEL"))
-                .ok()
-                .filter(|v| !v.trim().is_empty()),
-            minimax_base_url: std::env::var("MINIMAX_BASE_URL")
-                .ok()
-                .filter(|v| !v.trim().is_empty()),
-            minimax_anthropic_base_url: std::env::var("MINIMAX_ANTHROPIC_BASE_URL")
-                .ok()
-                .filter(|v| !v.trim().is_empty()),
-            minimax_model: std::env::var("MINIMAX_MODEL")
-                .ok()
-                .filter(|v| !v.trim().is_empty()),
-            deepinfra_base_url: std::env::var("DEEPINFRA_BASE_URL")
-                .ok()
-                .filter(|v| !v.trim().is_empty()),
-            deepinfra_model: std::env::var("DEEPINFRA_MODEL")
-                .ok()
-                .filter(|v| !v.trim().is_empty()),
-            sakana_base_url: std::env::var("SAKANA_BASE_URL")
-                .ok()
-                .filter(|v| !v.trim().is_empty()),
-            sakana_model: std::env::var("SAKANA_MODEL")
-                .ok()
-                .filter(|v| !v.trim().is_empty()),
-            longcat_base_url: std::env::var("LONGCAT_BASE_URL")
-                .ok()
-                .filter(|v| !v.trim().is_empty()),
-            longcat_model: std::env::var("LONGCAT_MODEL")
-                .ok()
-                .filter(|v| !v.trim().is_empty()),
-            opencode_go_base_url: std::env::var("OPENCODE_GO_BASE_URL")
-                .ok()
-                .filter(|v| !v.trim().is_empty()),
-            opencode_go_model: std::env::var("OPENCODE_GO_MODEL")
-                .ok()
-                .filter(|v| !v.trim().is_empty()),
-            opencode_zen_base_url: std::env::var("OPENCODE_ZEN_BASE_URL")
-                .ok()
-                .filter(|v| !v.trim().is_empty()),
-            opencode_zen_model: std::env::var("OPENCODE_ZEN_MODEL")
-                .ok()
-                .filter(|v| !v.trim().is_empty()),
-            meta_base_url: std::env::var("META_MODEL_API_BASE_URL")
-                .ok()
-                .filter(|v| !v.trim().is_empty())
-                .or_else(|| {
-                    std::env::var("MODEL_API_BASE_URL")
-                        .ok()
-                        .filter(|v| !v.trim().is_empty())
-                }),
-            meta_model: std::env::var("META_MODEL_API_MODEL")
-                .ok()
-                .filter(|v| !v.trim().is_empty())
-                .or_else(|| {
-                    std::env::var("MODEL_API_MODEL")
-                        .ok()
-                        .filter(|v| !v.trim().is_empty())
-                }),
-            xai_base_url: std::env::var("XAI_BASE_URL")
-                .ok()
-                .filter(|v| !v.trim().is_empty()),
-            xai_model: std::env::var("XAI_MODEL")
-                .ok()
-                .filter(|v| !v.trim().is_empty()),
-            google_base_url: std::env::var("GOOGLE_BASE_URL")
-                .ok()
-                .filter(|v| !v.trim().is_empty())
-                .or_else(|| {
-                    std::env::var("GEMINI_BASE_URL")
-                        .ok()
-                        .filter(|v| !v.trim().is_empty())
-                }),
-            google_model: std::env::var("GOOGLE_MODEL")
-                .ok()
-                .filter(|v| !v.trim().is_empty())
-                .or_else(|| {
-                    std::env::var("GEMINI_MODEL")
-                        .ok()
-                        .filter(|v| !v.trim().is_empty())
-                }),
-            mistral_base_url: std::env::var("MISTRAL_BASE_URL")
-                .ok()
-                .filter(|v| !v.trim().is_empty()),
-            mistral_model: std::env::var("MISTRAL_MODEL")
-                .ok()
-                .filter(|v| !v.trim().is_empty()),
-            telecomjs_base_url: std::env::var("TELECOMJS_BASE_URL")
-                .ok()
-                .filter(|v| !v.trim().is_empty()),
-            telecomjs_model: std::env::var("TELECOMJS_MODEL")
-                .ok()
-                .filter(|v| !v.trim().is_empty()),
-            edenai_base_url: std::env::var("EDENAI_BASE_URL")
-                .ok()
-                .filter(|v| !v.trim().is_empty()),
-            edenai_model: std::env::var("EDENAI_MODEL")
-                .ok()
-                .filter(|v| !v.trim().is_empty()),
-            zenmux_base_url: std::env::var("ZENMUX_BASE_URL")
-                .ok()
-                .filter(|v| !v.trim().is_empty()),
-            zenmux_model: std::env::var("ZENMUX_MODEL")
-                .ok()
-                .filter(|v| !v.trim().is_empty()),
-            csdn_base_url: std::env::var("CSDN_BASE_URL")
-                .ok()
-                .filter(|v| !v.trim().is_empty()),
-            csdn_model: std::env::var("CSDN_MODEL")
-                .ok()
-                .filter(|v| !v.trim().is_empty()),
-            concentrate_base_url: std::env::var("CONCENTRATE_BASE_URL")
-                .ok()
-                .filter(|v| !v.trim().is_empty()),
-            concentrate_model: std::env::var("CONCENTRATE_MODEL")
-                .ok()
-                .filter(|v| !v.trim().is_empty()),
+            active_route_base_url: env_non_blank(&["CODEWHALE_BASE_URL", "DEEPSEEK_BASE_URL"]),
+            deepseek_anthropic_base_url: env_non_blank(&["DEEPSEEK_ANTHROPIC_BASE_URL", "DEEPSEEK_CLAUDE_BASE_URL"]),
+            nvidia_base_url: env_non_blank(&["NVIDIA_NIM_BASE_URL", "NIM_BASE_URL", "NVIDIA_BASE_URL"]),
+            openai_base_url: env_non_blank(&["OPENAI_BASE_URL"]),
+            atlascloud_base_url: env_non_blank(&["ATLASCLOUD_BASE_URL"]),
+            volcengine_base_url: env_non_blank(&["VOLCENGINE_BASE_URL", "VOLCENGINE_ARK_BASE_URL", "ARK_BASE_URL"]),
+            wanjie_ark_base_url: env_non_blank(&["WANJIE_ARK_BASE_URL", "WANJIE_BASE_URL", "WANJIE_MAAS_BASE_URL"]),
+            openrouter_base_url: env_non_blank(&["OPENROUTER_BASE_URL"]),
+            orcarouter_base_url: env_non_blank(&["ORCAROUTER_BASE_URL"]),
+            xiaomi_mimo_base_url: env_non_blank(&["XIAOMI_MIMO_BASE_URL", "MIMO_BASE_URL"]),
+            novita_base_url: env_non_blank(&["NOVITA_BASE_URL"]),
+            fireworks_base_url: env_non_blank(&["FIREWORKS_BASE_URL"]),
+            siliconflow_base_url: env_non_blank(&["SILICONFLOW_BASE_URL"]),
+            siliconflow_model: env_non_blank(&["SILICONFLOW_MODEL"]),
+            arcee_base_url: env_non_blank(&["ARCEE_BASE_URL"]),
+            moonshot_base_url: env_non_blank(&["MOONSHOT_BASE_URL", "KIMI_BASE_URL"]),
+            sglang_base_url: env_non_blank(&["SGLANG_BASE_URL"]),
+            vllm_base_url: env_non_blank(&["VLLM_BASE_URL"]),
+            ollama_base_url: env_non_blank(&["OLLAMA_BASE_URL"]),
+            ollama_cloud_base_url: env_non_blank(&["OLLAMA_CLOUD_BASE_URL"]),
+            ollama_cloud_model: env_non_blank(&["OLLAMA_CLOUD_MODEL"]),
+            huggingface_base_url: env_non_blank(&["HUGGINGFACE_BASE_URL", "HF_BASE_URL"]),
+            huggingface_model: env_non_blank(&["HUGGINGFACE_MODEL", "HF_MODEL"]),
+            modelscope_base_url: env_non_blank(&["MODELSCOPE_BASE_URL"]),
+            modelscope_model: env_non_blank(&["MODELSCOPE_MODEL"]),
+            together_base_url: env_non_blank(&["TOGETHER_BASE_URL"]),
+            together_model: env_non_blank(&["TOGETHER_MODEL"]),
+            qianfan_base_url: env_non_blank(&["QIANFAN_BASE_URL", "BAIDU_QIANFAN_BASE_URL"]),
+            qianfan_model: env_non_blank(&["QIANFAN_MODEL", "BAIDU_QIANFAN_MODEL"]),
+            openai_codex_base_url: env_non_blank(&["OPENAI_CODEX_BASE_URL", "CODEX_BASE_URL"]),
+            openai_codex_model: env_non_blank(&["OPENAI_CODEX_MODEL", "CODEX_MODEL"]),
+            anthropic_base_url: env_non_blank(&["ANTHROPIC_BASE_URL"]),
+            anthropic_model: env_non_blank(&["ANTHROPIC_MODEL"]),
+            openmodel_base_url: env_non_blank(&["OPENMODEL_BASE_URL"]),
+            openmodel_model: env_non_blank(&["OPENMODEL_MODEL"]),
+            zai_base_url: env_non_blank(&["ZAI_BASE_URL", "Z_AI_BASE_URL", "ZHIPU_BASE_URL", "ZHIPUAI_BASE_URL", "BIGMODEL_BASE_URL"]),
+            zai_model: env_non_blank(&["ZAI_MODEL", "Z_AI_MODEL", "ZHIPU_MODEL", "ZHIPUAI_MODEL", "BIGMODEL_MODEL", "GLM_MODEL"]),
+            stepfun_base_url: env_non_blank(&["STEPFUN_BASE_URL", "STEP_BASE_URL"]),
+            stepfun_model: env_non_blank(&["STEPFUN_MODEL", "STEP_MODEL"]),
+            minimax_base_url: env_non_blank(&["MINIMAX_BASE_URL"]),
+            minimax_anthropic_base_url: env_non_blank(&["MINIMAX_ANTHROPIC_BASE_URL"]),
+            minimax_model: env_non_blank(&["MINIMAX_MODEL"]),
+            deepinfra_base_url: env_non_blank(&["DEEPINFRA_BASE_URL"]),
+            deepinfra_model: env_non_blank(&["DEEPINFRA_MODEL"]),
+            sakana_base_url: env_non_blank(&["SAKANA_BASE_URL"]),
+            sakana_model: env_non_blank(&["SAKANA_MODEL"]),
+            longcat_base_url: env_non_blank(&["LONGCAT_BASE_URL"]),
+            longcat_model: env_non_blank(&["LONGCAT_MODEL"]),
+            opencode_go_base_url: env_non_blank(&["OPENCODE_GO_BASE_URL"]),
+            opencode_go_model: env_non_blank(&["OPENCODE_GO_MODEL"]),
+            opencode_zen_base_url: env_non_blank(&["OPENCODE_ZEN_BASE_URL"]),
+            opencode_zen_model: env_non_blank(&["OPENCODE_ZEN_MODEL"]),
+            meta_base_url: env_non_blank(&["META_MODEL_API_BASE_URL", "MODEL_API_BASE_URL"]),
+            meta_model: env_non_blank(&["META_MODEL_API_MODEL", "MODEL_API_MODEL"]),
+            xai_base_url: env_non_blank(&["XAI_BASE_URL"]),
+            xai_model: env_non_blank(&["XAI_MODEL"]),
+            google_base_url: env_non_blank(&["GOOGLE_BASE_URL", "GEMINI_BASE_URL"]),
+            google_model: env_non_blank(&["GOOGLE_MODEL", "GEMINI_MODEL"]),
+            mistral_base_url: env_non_blank(&["MISTRAL_BASE_URL"]),
+            mistral_model: env_non_blank(&["MISTRAL_MODEL"]),
+            telecomjs_base_url: env_non_blank(&["TELECOMJS_BASE_URL"]),
+            telecomjs_model: env_non_blank(&["TELECOMJS_MODEL"]),
+            edenai_base_url: env_non_blank(&["EDENAI_BASE_URL"]),
+            edenai_model: env_non_blank(&["EDENAI_MODEL"]),
+            zenmux_base_url: env_non_blank(&["ZENMUX_BASE_URL"]),
+            zenmux_model: env_non_blank(&["ZENMUX_MODEL"]),
+            csdn_base_url: env_non_blank(&["CSDN_BASE_URL"]),
+            csdn_model: env_non_blank(&["CSDN_MODEL"]),
+            concentrate_base_url: env_non_blank(&["CONCENTRATE_BASE_URL"]),
+            concentrate_model: env_non_blank(&["CONCENTRATE_MODEL"]),
             // `CODEWHALE_API_BASE` is a trust boundary, not a plain string:
             // an origin this route would refuse to send a bearer to is
             // dropped here rather than resolved into a route.
             codewhale_base_url: provider::codewhale_api_base_from_env(),
-            codewhale_model: std::env::var("CODEWHALE_MODEL")
-                .ok()
-                .filter(|v| !v.trim().is_empty()),
-            modelstudio_token_plan_base_url: std::env::var("MODELSTUDIO_TOKEN_PLAN_BASE_URL")
-                .ok()
-                .filter(|v| !v.trim().is_empty()),
-            modelstudio_token_plan_model: std::env::var("MODELSTUDIO_TOKEN_PLAN_MODEL")
-                .ok()
-                .filter(|v| !v.trim().is_empty()),
-            modelstudio_coding_plan_base_url: std::env::var("MODELSTUDIO_CODING_PLAN_BASE_URL")
-                .ok()
-                .filter(|v| !v.trim().is_empty()),
-            modelstudio_coding_plan_model: std::env::var("MODELSTUDIO_CODING_PLAN_MODEL")
-                .ok()
-                .filter(|v| !v.trim().is_empty()),
+            codewhale_model: env_non_blank(&["CODEWHALE_MODEL"]),
+            modelstudio_token_plan_base_url: env_non_blank(&["MODELSTUDIO_TOKEN_PLAN_BASE_URL"]),
+            modelstudio_token_plan_model: env_non_blank(&["MODELSTUDIO_TOKEN_PLAN_MODEL"]),
+            modelstudio_coding_plan_base_url: env_non_blank(&["MODELSTUDIO_CODING_PLAN_BASE_URL"]),
+            modelstudio_coding_plan_model: env_non_blank(&["MODELSTUDIO_CODING_PLAN_MODEL"]),
         }
     }
 

crates/config/src/settings_schema.rs

@@ -605,7 +605,7 @@ pub const SETTINGS_SCHEMA: &[SettingDef] = &[
     def(
         "show_thinking",
         SettingKind::Bool(SHOW_THINKING),
-        "true",
+        "false",
         ui(
             TAB_APPEARANCE,
             "display",
@@ -1546,12 +1546,6 @@ pub const SETTINGS_SCHEMA: &[SettingDef] = &[
         "false",
         None,
     ),
-    def(
-        "account_preferences_notice_shown",
-        SettingKind::String,
-        "",
-        None,
-    ),
     // Round 3 work-bar placement migration ran once (`top` → `bottom`).
     def(
         "work_surface_bottom_migrated",

crates/config/src/stream_timeouts.rs

@@ -27,10 +27,11 @@ pub fn resolve_stream_open_timeout(configured_secs: Option<u64>) -> std::time::D
             secs.clamp(MIN_STREAM_OPEN_TIMEOUT_SECS, MAX_STREAM_OPEN_TIMEOUT_SECS),
         ),
         _ => stream_open_timeout_from_env(
-            std::env::var("CODEWHALE_STREAM_OPEN_TIMEOUT_SECS")
-                .or_else(|_| std::env::var("DEEPSEEK_STREAM_OPEN_TIMEOUT_SECS"))
-                .ok()
-                .as_deref(),
+            crate::env_overrides::env_non_blank(&[
+                "CODEWHALE_STREAM_OPEN_TIMEOUT_SECS",
+                "DEEPSEEK_STREAM_OPEN_TIMEOUT_SECS",
+            ])
+            .as_deref(),
         ),
     }
 }
@@ -50,9 +51,7 @@ pub fn stream_open_timeout_from_env(value: Option<&str>) -> std::time::Duration
 /// `false`, ...) leaves HTTP/2 on.
 #[must_use]
 pub fn force_http1_from_env() -> bool {
-    std::env::var("CODEWHALE_FORCE_HTTP1")
-        .or_else(|_| std::env::var("DEEPSEEK_FORCE_HTTP1"))
-        .ok()
+    crate::env_overrides::env_non_blank(&["CODEWHALE_FORCE_HTTP1", "DEEPSEEK_FORCE_HTTP1"])
         .map(|v| v.trim().to_ascii_lowercase())
         .is_some_and(|v| matches!(v.as_str(), "1" | "true" | "yes" | "on"))
 }

crates/config/src/tests.rs

@@ -10767,6 +10767,60 @@ fn blank_or_unrecognized_env_overrides_do_not_shadow_config_or_legacy_vars() {
     ));
 }
 
+#[test]
+fn blank_primary_env_vars_fall_through_to_their_legacy_aliases() {
+    let _lock = env_lock();
+    let _env = EnvGuard::without_deepseek_runtime_overrides();
+    let names = [
+        "CODEWHALE_FORCE_HTTP1",
+        "DEEPSEEK_FORCE_HTTP1",
+        "CODEWHALE_STREAM_OPEN_TIMEOUT_SECS",
+        "DEEPSEEK_STREAM_OPEN_TIMEOUT_SECS",
+    ];
+    let saved: Vec<(&str, Option<OsString>)> = names
+        .iter()
+        .map(|name| (*name, env::var_os(name)))
+        .collect();
+    // SAFETY: env mutation is serialized by `env_lock` and restored below.
+    unsafe {
+        env::set_var("CODEWHALE_MODEL", "");
+        env::set_var("DEEPSEEK_MODEL", "legacy-model");
+        env::set_var("CODEWHALE_BASE_URL", " ");
+        env::set_var("DEEPSEEK_BASE_URL", "https://legacy.example/v1");
+        env::set_var("MOONSHOT_BASE_URL", "");
+        env::set_var("KIMI_BASE_URL", "https://kimi.example/v1");
+        env::set_var("CODEWHALE_FORCE_HTTP1", "");
+        env::set_var("DEEPSEEK_FORCE_HTTP1", "1");
+        env::set_var("CODEWHALE_STREAM_OPEN_TIMEOUT_SECS", "");
+        env::set_var("DEEPSEEK_STREAM_OPEN_TIMEOUT_SECS", "120");
+    }
+
+    let env_overrides = EnvRuntimeOverrides::load();
+    let force_http1 = crate::stream_timeouts::force_http1_from_env();
+    let open_timeout = crate::stream_timeouts::resolve_stream_open_timeout(None);
+
+    unsafe {
+        for (name, value) in saved {
+            match value {
+                Some(value) => env::set_var(name, value),
+                None => env::remove_var(name),
+            }
+        }
+    }
+
+    assert_eq!(env_overrides.model.as_deref(), Some("legacy-model"));
+    assert_eq!(
+        env_overrides.active_route_base_url.as_deref(),
+        Some("https://legacy.example/v1")
+    );
+    assert_eq!(
+        env_overrides.moonshot_base_url.as_deref(),
+        Some("https://kimi.example/v1")
+    );
+    assert!(force_http1);
+    assert_eq!(open_timeout, std::time::Duration::from_secs(120));
+}
+
 #[cfg(unix)]
 #[test]
 fn failed_legacy_state_copy_leaves_no_partial_primary() {

crates/dependencies/src/lib.rs

@@ -1183,7 +1183,7 @@ impl Git {
         overrides: Vec<(String, &'static str)>,
     ) -> anyhow::Result<Command> {
         let mut command = Self::review_base(workspace)?;
-        let mut count = 2;
+        let mut count = 3;
         for (key, value) in overrides {
             // A subsection may contain '='; `-c key=value` would then
             // override a different key. Separate env fields preserve it.
@@ -1195,7 +1195,7 @@ impl Git {
         Ok(command)
     }
 
-    /// Git with fsmonitor, hooks, lazy fetch and replace objects disabled,
+    /// Git with fsmonitor, hooks, lazy fetch, replace objects and path quoting disabled,
     /// running in `workspace`. [`Self::review_command`] adds filter overrides.
     pub fn review_base(workspace: &Path) -> anyhow::Result<Command> {
         use anyhow::Context;
@@ -1257,7 +1257,7 @@ impl Git {
             // Both phases must observe the same effective repository config.
             .env_remove("GIT_CONFIG")
             .env_remove("GIT_CONFIG_PARAMETERS")
-            .env("GIT_CONFIG_COUNT", "2")
+            .env("GIT_CONFIG_COUNT", "3")
             .env("GIT_CONFIG_KEY_0", "core.fsmonitor")
             .env("GIT_CONFIG_VALUE_0", "false")
             .env("GIT_CONFIG_KEY_1", "core.hooksPath")
@@ -1265,6 +1265,8 @@ impl Git {
                 "GIT_CONFIG_VALUE_1",
                 if cfg!(windows) { "NUL" } else { "/dev/null" },
             )
+            .env("GIT_CONFIG_KEY_2", "core.quotePath")
+            .env("GIT_CONFIG_VALUE_2", "false")
             .env("GIT_NO_LAZY_FETCH", "1")
             .env("GIT_NO_REPLACE_OBJECTS", "1");
         Ok(command)

crates/engine/Cargo.toml

@@ -219,9 +219,6 @@ libc = "0.2"
 windows-sys = { version = "0.61.2", features = ["Wdk_Storage_FileSystem", "Win32_System_IO", "Win32_Foundation", "Win32_Security", "Win32_Security_Authorization", "Win32_Security_Isolation", "Win32_Storage_FileSystem", "Win32_System_LibraryLoader", "Win32_System_Pipes", "Win32_System_Threading", "Win32_System_SystemServices", "Win32_System_WindowsProgramming"] }
 
 
-[target.'cfg(any(target_os = "macos", target_os = "windows", all(target_os = "linux", not(target_env = "ohos"))))'.dependencies]
-arboard = { version = "3.4", features = ["wayland-data-control"] }
-
 [target.'cfg(not(target_env = "ohos"))'.dependencies]
 portable-pty = "0.9"
 

crates/engine/src/core/authority/auto_review.rs

@@ -2152,10 +2152,7 @@ fn git_push_refspec_is_protected(refspec: &str) -> bool {
     // Tag-like names (`v1`, `v0.9.1`): git resolves branch-vs-tag on the
     // server, so treat them as publishes.
     let mut chars = lower.chars();
-    if chars.next() == Some('v') && chars.next().is_some_and(|ch| ch.is_ascii_digit()) {
-        return true;
-    }
-    false
+    chars.next() == Some('v') && chars.next().is_some_and(|ch| ch.is_ascii_digit())
 }
 
 fn git_tag_tokens_are_publish_like(tokens: &[&str]) -> bool {

crates/engine/src/core/engine.rs

@@ -314,8 +314,8 @@ pub struct EngineConfig {
     pub session_id: Option<String>,
     /// Optional host-owned root for delegated-agent runtime state.
     ///
-    /// When unset, the worker ledger and complete transcript artifacts
-    /// retain their historical location under
+    /// When unset, the worker ledger, complete transcript artifacts and
+    /// coordination lock retain their historical location under
     /// `workspace/.codewhale/state`. Embedders may set a session-scoped root
     /// to separate that control-plane state from the execution workspace.
     /// Child cwd and file authority still derive from `workspace`; hosts using

crates/engine/src/extension_host/dsh_workspace.rs

@@ -203,10 +203,13 @@ impl Drop for TestOverlay {
 
 fn overlay_granted() -> bool {
     #[cfg(test)]
-    if TEST_OVERLAY.with(std::cell::Cell::get) {
-        return true;
+    {
+        TEST_OVERLAY.with(std::cell::Cell::get)
+    }
+    #[cfg(not(test))]
+    {
+        false
     }
-    false
 }
 
 fn caller(call: &BrokerCall, tool_name: &'static str) -> Result<ToolContext, BrokerRefusal> {

crates/engine/src/fleet/agents.rs

@@ -187,6 +187,19 @@ pub const AGENTS: &[ExternalAgent] = &[
         adapter_binary: None,
         adapter_cli_env: None,
     },
+    ExternalAgent {
+        id: "qwen-code",
+        name: "Qwen Code",
+        binary: "qwen",
+        acp_args: Some(&["--acp"]),
+        website: "https://github.com/QwenLM/qwen-code",
+        min_version: Some("0.6.1"),
+        upgrade: Some("npm i -g @qwen-code/qwen-code@latest"),
+        login: None,
+        adapter: None,
+        adapter_binary: None,
+        adapter_cli_env: None,
+    },
 ];
 
 const FIXTURE_AGENT: ExternalAgent = ExternalAgent {

crates/engine/src/fleet/manager.rs

@@ -1748,11 +1748,7 @@ impl FleetManager {
                 return Err(start_error);
             }
         };
-        if !started {
-            return Ok(false);
-        }
-
-        Ok(true)
+        Ok(started)
     }
 
     fn start_leased_workers(

crates/engine/src/runtime_threads/tests.rs

@@ -11615,9 +11615,10 @@ async fn compact_thread_receipt_keeps_exact_named_custom_identity() -> Result<()
 
 #[tokio::test]
 async fn compact_thread_with_real_engine_reaches_terminal_status() -> Result<()> {
+    // A real engine takes the workspace coordination lock exclusively, and
     // nextest runs tests in parallel processes with this crate as their working
     // directory. A `"."` workspace would therefore be the crate directory
-    // itself, shared by every test process.
+    // itself, and two exclusive holders of it cannot both start.
     let workspace = tempfile::tempdir()?;
     let manager = RuntimeThreadManager::open(
         {

crates/engine/src/tools/subagent/coord.rs

@@ -1721,7 +1721,7 @@ mod tests {
             .expect_err("mutation must fail when its receipt cannot persist")
             .to_string();
         assert!(
-            error.contains("delegated durable store admission failed"),
+            error.contains("Runtime coordination was opened without its writer lock"),
             "{error}"
         );
         assert!(

crates/engine/src/tools/subagent/coord/ledger.rs

@@ -133,14 +133,9 @@ impl WriteScopeClaim {
                 return true;
             }
         }
-        if self
-            .contracts
+        self.contracts
             .iter()
             .any(|contract| other.contracts.iter().any(|other| other == contract))
-        {
-            return true;
-        }
-        false
     }
 
     #[must_use]
@@ -326,10 +321,9 @@ pub struct CoordinationDetailProjection {
     pub metrics: CoordinationDetailMetrics,
     pub bounded: bool,
     pub limit: usize,
-    /// Whether this manager's durable store is admitted. When false, durable
-    /// ledger writes are refused until admission is retried and the UI must
-    /// say so — a counter must never tick on a turn the engine has already
-    /// settled.
+    /// Whether this process currently holds the workspace coordination flock.
+    /// When false, durable ledger writes are skipped and the UI must say so —
+    /// a counter must never tick on a turn the engine has already settled.
     #[serde(default = "default_process_lock_held")]
     pub process_lock_held: bool,
     /// Human-readable reason when [`Self::process_lock_held`] is false.

crates/engine/src/tools/subagent/mod.rs

@@ -499,6 +499,7 @@ const SUBAGENT_TRANSCRIPT_ARTIFACT_DIR: &str = "subagent-transcripts";
 const SUBAGENT_STATE_SCHEMA_VERSION: u32 = 1;
 const SUBAGENT_STATE_FILE: &str = "subagents.v3.json";
 const LEGACY_SUBAGENT_STATE_FILE: &str = "subagents.v1.json";
+const SUBAGENT_STATE_LOCK_FILE: &str = "subagents.v1.lock";
 const SUBAGENT_RESTART_REASON: &str = "Interrupted by process restart";
 const SUBAGENT_SESSION_CLOSED_REASON: &str = "Interrupted: parent session closed";
 #[cfg(test)]
@@ -3932,6 +3933,12 @@ impl SubAgent {
     }
 }
 
+/// Manager for active sub-agents.
+struct CoordinationProcessLock {
+    release: Option<std::sync::mpsc::SyncSender<()>>,
+    thread: Option<std::thread::JoinHandle<()>>,
+}
+
 /// Marker prefix retained for presentation compatibility with same-process
 /// handover notes from earlier hosts. It does not control lock acquisition
 /// or retry.
@@ -3947,6 +3954,123 @@ pub const COORDINATION_LOCK_TIMEOUT_MARKER: &str =
 
 pub const SESSION_PRIVATE_COORDINATION: &str = "another Codewhale session still uses this folder's older delegated-agent store, so its upgrade waits; sub-agents in this session keep private state that other sessions do not see and that is not saved, and the store upgrades when a session starts after the other one exits";
 
+#[derive(Debug)]
+struct SubAgentCutoverBlocked(anyhow::Error);
+
+impl std::fmt::Display for SubAgentCutoverBlocked {
+    fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
+        write!(f, "{}", self.0)
+    }
+}
+
+impl std::error::Error for SubAgentCutoverBlocked {}
+
+impl CoordinationProcessLock {
+    fn acquire(state_root: &Path, exclusive: bool) -> Result<Self> {
+        let requested_root = normalize_subagent_workspace(state_root);
+        // Root and path for the lock. A non-home workspace keeps itself as the
+        // root, so a repository-supplied `.codewhale` link is walked and
+        // refused before `create_dir_all` could follow it into the target. A
+        // home workspace resolves its own junction once and still walks
+        // everything below it.
+        let (walk_root, lock_path) = codewhale_util::files::workspace_state_path(
+            &requested_root,
+            &Path::new("state").join(SUBAGENT_STATE_LOCK_FILE),
+        );
+        let lock_dir = lock_path
+            .parent()
+            .map_or_else(|| walk_root.clone(), Path::to_path_buf);
+        // Refuse a linked `.codewhale` or `state` before creating anything:
+        // `create_dir_all` follows links and would populate the target.
+        reject_root_relative_symlinks(&walk_root, &lock_path)?;
+        fs::create_dir_all(&lock_dir)?;
+        let ignore = lock_dir.join(".gitignore");
+        let ignored = reject_root_relative_symlinks(&walk_root, &ignore).and_then(|()| {
+            match fs::OpenOptions::new()
+                .write(true)
+                .create_new(true)
+                .open(&ignore)
+            {
+                Ok(mut file) => file.write_all(b"*\n").map_err(Into::into),
+                Err(error) if error.kind() == std::io::ErrorKind::AlreadyExists => Ok(()),
+                Err(error) => Err(error.into()),
+            }
+        });
+        if let Err(error) = ignored {
+            tracing::warn!(target: "subagent", %error, "could not write {}", ignore.display());
+        }
+        // Creating a missing root can change its canonical spelling on
+        // Windows (for example by adding a `\\?\` prefix). Re-resolve both
+        // sides before checking containment.
+        let state_root = normalize_subagent_workspace(state_root);
+        let (walk_root, lock_path) = codewhale_util::files::workspace_state_path(
+            &state_root,
+            &Path::new("state").join(SUBAGENT_STATE_LOCK_FILE),
+        );
+        reject_root_relative_symlinks(&walk_root, &lock_path)?;
+        let lock_file = std::fs::OpenOptions::new()
+            .create(true)
+            .truncate(false)
+            .read(true)
+            .write(true)
+            .open(&lock_path)?;
+        let (ready_tx, ready_rx) = std::sync::mpsc::sync_channel(1);
+        let (release_tx, release_rx) = std::sync::mpsc::sync_channel(0);
+        let thread = std::thread::spawn(move || {
+            let mut lock = fd_lock::RwLock::new(lock_file);
+            let acquired = if exclusive {
+                lock.try_write().map(|guard| {
+                    let _ = ready_tx.send(Ok::<(), String>(()));
+                    let _guard = guard;
+                    let _ = release_rx.recv();
+                })
+            } else {
+                lock.try_read().map(|guard| {
+                    let _ = ready_tx.send(Ok::<(), String>(()));
+                    let _guard = guard;
+                    let _ = release_rx.recv();
+                })
+            };
+            match acquired {
+                Ok(()) => {}
+                Err(error) => {
+                    let _ = ready_tx.send(Err(error.to_string()));
+                }
+            }
+        });
+        match ready_rx.recv_timeout(Duration::from_secs(5)) {
+            Ok(Ok(())) => Ok(Self {
+                release: Some(release_tx),
+                thread: Some(thread),
+            }),
+            Ok(Err(error)) => {
+                let _ = thread.join();
+                Err(anyhow!(
+                    "delegated coordination lock is unavailable for {}: {error}; close any other running or suspended Codewhale session using this workspace, then restart this Runtime if admission was refused",
+                    state_root.display()
+                ))
+            }
+            Err(error) => {
+                drop(release_tx);
+                let _ = thread.join();
+                Err(anyhow!(
+                    "{COORDINATION_LOCK_TIMEOUT_MARKER} for {}: {error}",
+                    state_root.display()
+                ))
+            }
+        }
+    }
+}
+
+impl Drop for CoordinationProcessLock {
+    fn drop(&mut self) {
+        self.release.take();
+        if let Some(thread) = self.thread.take() {
+            let _ = thread.join();
+        }
+    }
+}
+
 /// Only the captured Runtime owner admits roots. Held handles are not serialized;
 /// a restored attributed claim must be re-admitted before it can authorize work.
 struct AdmittedCoordinationRoot {
@@ -3982,12 +4106,19 @@ pub struct SubAgentManager {
     worker_origins: HashMap<String, WorkerOrigin>,
     worker_origin_schema: u32,
     workspace: PathBuf,
-    /// Root that owns the delegated-agent ledger and complete transcript
-    /// artifacts. Defaults to `workspace`; embedders may
+    /// Root that owns the delegated-agent ledger, complete transcript
+    /// artifacts and coordination lock. Defaults to `workspace`; embedders may
     /// isolate it without changing child execution cwd or file authority.
     state_root: PathBuf,
     state_path: Option<PathBuf>,
     state_unreachable_warned: bool,
+    coordination_process_lock: std::sync::Mutex<Option<Arc<CoordinationProcessLock>>>,
+    coordination_process_lock_required: bool,
+    coordination_exclusive: bool,
+    coordination_snapshot_read_only: bool,
+    store_cutover_blocked: bool,
+    session_private: bool,
+    workspace_writer_locks: HashMap<PathBuf, Arc<CoordinationProcessLock>>,
     /// Configured default per-child model-turn budget (`[subagents]
     /// default_max_steps`, #5324). `None` keeps the unbounded Fleet default;
     /// an explicit spawn `max_steps` still wins. Zero means unbounded.
@@ -4628,6 +4759,13 @@ impl SubAgentManager {
             state_root,
             state_path: None,
             state_unreachable_warned: false,
+            coordination_process_lock: std::sync::Mutex::new(None),
+            coordination_process_lock_required: false,
+            coordination_exclusive: false,
+            coordination_snapshot_read_only: false,
+            store_cutover_blocked: false,
+            session_private: false,
+            workspace_writer_locks: HashMap::new(),
             max_steps: None,
             wall_time: None,
             max_agents,
@@ -4693,6 +4831,22 @@ impl SubAgentManager {
         return Err("coordination workspace identity is unsupported on this platform".into());
         #[cfg(any(unix, windows))]
         {
+            if self.coordination_exclusive
+                && !self.coordination_snapshot_read_only
+                && normalize_subagent_workspace(&lexical)
+                    != normalize_subagent_workspace(&self.state_root)
+                && !self.workspace_writer_locks.contains_key(&canonical_root)
+            {
+                match CoordinationProcessLock::acquire(&lexical, true) {
+                    Ok(lock) => {
+                        self.workspace_writer_locks
+                            .insert(canonical_root.clone(), Arc::new(lock));
+                    }
+                    Err(error) => {
+                        tracing::warn!(target: "subagent", %error, "admitted coordination workspace without its writer lock; shared-workspace writers there stay refused until it is free");
+                    }
+                }
+            }
             let receipt = CoordinationClaimScope {
                 canonical_root,
                 platform: platform.into(),
@@ -4712,61 +4866,44 @@ impl SubAgentManager {
         }
     }
 
-    fn refuse_live_selected_store_writers(&self, origin: &WorkerOrigin) -> Result<(), String> {
-        let (Some(selected), Some(_)) = (&origin.selected_workspace, &origin.scope) else {
+    fn ensure_origin_writer_lock(&mut self, origin: &WorkerOrigin) -> Result<(), String> {
+        let (Some(selected), Some(scope)) = (&origin.selected_workspace, &origin.scope) else {
             return Ok(());
         };
-        if normalize_subagent_workspace(selected) == normalize_subagent_workspace(&self.state_root)
+        if !self.coordination_exclusive
+            || normalize_subagent_workspace(selected)
+                == normalize_subagent_workspace(&self.state_root)
+            || self
+                .workspace_writer_locks
+                .contains_key(&scope.canonical_root)
         {
             return Ok(());
         }
-        let path = default_state_path(selected).map_err(|error| error.to_string())?;
-        let unreadable = |error: std::io::Error| {
-            format!(
-                "cannot check {} for another open Codewhale session ({error}); shared-workspace writers there stay refused, or launch the builder in an isolated worktree",
-                selected.display()
-            )
-        };
-        let directory = path
-            .parent()
-            .ok_or_else(|| "delegated store has no parent".to_string())?;
-        let entries = match fs::read_dir(directory) {
-            Ok(entries) => entries,
-            Err(error) if error.kind() == std::io::ErrorKind::NotFound => return Ok(()),
-            Err(error) => return Err(unreadable(error)),
-        };
-        for entry in entries {
-            let name = entry.map_err(unreadable)?.file_name();
-            let Some(writer) = name
-                .to_str()
-                .and_then(|name| name.strip_prefix("subagents.v3.writer."))
-                .and_then(|name| name.strip_suffix(".lock"))
-            else {
-                continue;
-            };
-            if valid_store_writer(writer) && store_writer_is_live(&path, writer) {
-                return Err(format!(
-                    "another Codewhale session ({writer}) has {} open; shared-workspace writers there stay refused until it exits, or launch the builder in an isolated worktree",
-                    selected.display()
-                ));
-            }
+        let lock =
+            CoordinationProcessLock::acquire(selected, true).map_err(|error| error.to_string())?;
+        self.workspace_writer_locks
+            .insert(scope.canonical_root.clone(), Arc::new(lock));
+        Ok(())
+    }
+
+    fn ensure_worker_coordination_locks(
+        &mut self,
+        spec: &AgentWorkerSpec,
+        origin: &WorkerOrigin,
+    ) -> Result<(), String> {
+        let claim = worker_coordination_claim(spec)?;
+        if claim.as_ref().is_some_and(|(_, isolated)| *isolated) {
+            return Ok(());
+        }
+        self.ensure_coordination_process_lock()?;
+        if claim.is_some() {
+            self.ensure_origin_writer_lock(origin)?;
         }
         Ok(())
     }
 
-    fn admit_worker_shared_writes(
-        &self,
-        spec: &AgentWorkerSpec,
-        origin: &WorkerOrigin,
-    ) -> Result<(), String> {
-        match worker_coordination_claim(spec)? {
-            Some((_, false)) => self.refuse_live_selected_store_writers(origin),
-            _ => Ok(()),
-        }
-    }
-
     pub fn ensure_registered_worker_coordination_locks(
-        &self,
+        &mut self,
         worker_id: &str,
     ) -> Result<(), String> {
         let spec = self
@@ -4781,7 +4918,7 @@ impl SubAgentManager {
             .cloned()
             .ok_or_else(|| "worker origin is missing from the owner journal".to_string())?;
         self.validate_worker_origin(&origin)?;
-        self.admit_worker_shared_writes(&spec, &origin)
+        self.ensure_worker_coordination_locks(&spec, &origin)
     }
 
     fn admitted_coordination_root(&self, workspace: &Path) -> Option<&AdmittedCoordinationRoot> {
@@ -5179,6 +5316,7 @@ impl SubAgentManager {
                 .map_err(|error| error.to_string());
         }
 
+        self.ensure_coordination_process_lock()?;
         if let Some(root) = self.admitted_coordination_root(workspace) {
             root.validate()?;
         }
@@ -5257,6 +5395,7 @@ impl SubAgentManager {
                 .map_err(|error| error.to_string());
         }
 
+        self.ensure_coordination_process_lock()?;
         self.coordination
             .update_decision_status(decision_id, status, owner, expected_version)
     }
@@ -5303,6 +5442,7 @@ impl SubAgentManager {
                 .map_err(|error| error.to_string());
         }
 
+        self.ensure_coordination_process_lock()?;
         let expected_owner = self.nearest_common_fan_in_owner(&input_decisions)?;
         if owner != expected_owner {
             return Err(format!(
@@ -5574,13 +5714,19 @@ impl SubAgentManager {
             },
             bounded: true,
             limit,
-            process_lock_held: self.state_path.is_none() || self.store_authority.is_some(),
-            process_lock_note: (self.state_path.is_some() && self.store_authority.is_none()).then(
-                || {
-                    "delegated store admission failed; it is retried on the next delegated write"
-                        .to_string()
-                },
-            ),
+            process_lock_held: !self.session_private && {
+                // Retry acquire on each projection so a session that lost the
+                // flock at construction recovers once the previous owner exits.
+                let _ = self.coordination_process_lock_status();
+                self.holds_coordination_process_lock()
+            },
+            process_lock_note: if self.session_private {
+                Some(SESSION_PRIVATE_COORDINATION.to_string())
+            } else if self.holds_coordination_process_lock() {
+                None
+            } else {
+                self.coordination_process_lock_status().err()
+            },
         }
     }
 
@@ -5715,6 +5861,7 @@ impl SubAgentManager {
                 .and_then(|outcome| outcome);
         }
 
+        self.ensure_coordination_process_lock()?;
         let Some(existing) = self
             .coordination
             .write_claims
@@ -5801,6 +5948,7 @@ impl SubAgentManager {
                 .map_err(|error| error.to_string());
         }
 
+        self.ensure_coordination_process_lock()?;
         let active = self.active_coordination_owners();
         let released = self
             .coordination
@@ -5925,12 +6073,109 @@ impl SubAgentManager {
         match SubAgentStoreAuthority::open(&self.state_root, &path, &self.current_session_boot_id) {
             Ok(authority) => self.store_authority = Some(authority),
             Err(error) => {
-                tracing::warn!(target: "subagent", %error, "delegated store admission failed; it is retried on the next delegated write");
+                self.coordination_snapshot_read_only = true;
+                self.store_cutover_blocked = error
+                    .chain()
+                    .any(|cause| cause.is::<SubAgentCutoverBlocked>());
+                tracing::warn!(target: "subagent", %error, "delegated store admission failed; writable state remains refused");
             }
         }
         self
     }
 
+    fn into_session_private(mut self) -> Self {
+        self.state_path = None;
+        self.store_authority = None;
+        self.coordination_snapshot_read_only = false;
+        self.coordination_process_lock_required = false;
+        self.session_private = true;
+        tracing::warn!(target: "subagent", "{SESSION_PRIVATE_COORDINATION}");
+        self
+    }
+
+    fn require_coordination_process_lock(mut self) -> Self {
+        self.coordination_process_lock_required = true;
+        if self.store_authority.is_none()
+            && let Some(path) = self.state_path.as_ref()
+        {
+            match SubAgentStoreAuthority::open(
+                &self.state_root,
+                path,
+                &self.current_session_boot_id,
+            ) {
+                Ok(authority) => self.store_authority = Some(authority),
+                Err(error) => {
+                    self.coordination_snapshot_read_only = true;
+                    tracing::warn!(target: "subagent", %error, "delegated store migration or admission failed; no durable writes are admitted");
+                    return self;
+                }
+            }
+        }
+        let acquired =
+            CoordinationProcessLock::acquire(&self.state_root, self.coordination_exclusive)
+                .ok()
+                .map(Arc::new);
+        if acquired.is_none() {
+            return self.into_session_private();
+        }
+        *self
+            .coordination_process_lock
+            .get_mut()
+            .expect("coordination lock slot poisoned") = acquired;
+        self
+    }
+
+    /// Terminal sessions retry failed acquisition when the previous owner
+    /// exits. A Runtime loaded without its writer lock must restart rather
+    /// than publish its stale snapshot. The flock is held on an open fd —
+    /// deleting the lock file does not release it (#5036).
+    fn ensure_coordination_process_lock(&self) -> Result<(), String> {
+        if self.coordination_snapshot_read_only {
+            return Err(
+                "Runtime coordination was opened without its writer lock; stop and restart this Runtime process after the other session exits"
+                    .into(),
+            );
+        }
+        if !self.coordination_process_lock_required {
+            return Ok(());
+        }
+        let mut slot = self
+            .coordination_process_lock
+            .lock()
+            .expect("coordination lock slot poisoned");
+        if slot.is_some() {
+            return Ok(());
+        }
+        match CoordinationProcessLock::acquire(&self.state_root, self.coordination_exclusive) {
+            Ok(lock) => {
+                *slot = Some(Arc::new(lock));
+                Ok(())
+            }
+            Err(error) => Err(error.to_string()),
+        }
+    }
+
+    /// Whether this process currently owns the workspace coordination flock.
+    /// Read-only inspection of the in-process slot — does not attempt acquire.
+    #[must_use]
+    pub fn holds_coordination_process_lock(&self) -> bool {
+        if !self.coordination_process_lock_required {
+            // Tests and non-durable managers do not require the flock.
+            return true;
+        }
+        self.coordination_process_lock
+            .lock()
+            .expect("coordination lock slot poisoned")
+            .is_some()
+    }
+
+    /// Probe lock ownership for UI/inspect surfaces. Retries acquire when the
+    /// slot is empty so a terminal session can recover after the previous
+    /// owner exits (#2.6 / #5036); read-only Runtime snapshots require restart.
+    pub fn coordination_process_lock_status(&self) -> Result<(), String> {
+        self.ensure_coordination_process_lock()
+    }
+
     #[must_use]
     pub fn with_running_heartbeat_timeout(mut self, timeout: Duration) -> Self {
         self.running_heartbeat_timeout = if timeout.is_zero() {
@@ -5978,31 +6223,18 @@ impl SubAgentManager {
                 .is_some_and(|authority| authority.owned.lock().contains(id))
     }
 
-    fn admitted_store_authority(&mut self) -> Result<Option<Arc<SubAgentStoreAuthority>>> {
-        let Some(path) = self.state_path.as_ref() else {
-            return Ok(None);
-        };
-        if let Some(authority) = self.store_authority.as_ref() {
-            return Ok(Some(authority.clone()));
-        }
-        let authority =
-            SubAgentStoreAuthority::open(&self.state_root, path, &self.current_session_boot_id)
-                .map_err(|error| {
-                    anyhow!(
-                        "delegated durable store admission failed ({error:#}); it is retried on the next delegated write"
-                    )
-                })?;
-        self.store_authority = Some(authority.clone());
-        Ok(Some(authority))
-    }
-
     pub(crate) fn with_store_transaction<T>(
         &mut self,
         operation: impl FnOnce(&mut Self) -> Result<T>,
     ) -> Result<T> {
-        let Some(authority) = self.admitted_store_authority()? else {
+        if self.state_path.is_none() {
             return operation(self);
-        };
+        }
+        self.ensure_coordination_process_lock()
+            .map_err(anyhow::Error::msg)?;
+        let authority = self.store_authority.clone().ok_or_else(|| {
+            anyhow!("delegated durable store admission failed; inspect before retrying")
+        })?;
         if authority.is_current_transaction() {
             return operation(self);
         }
@@ -6203,6 +6435,8 @@ impl SubAgentManager {
         worker_id: &str,
         payload: PersistedSubAgentState,
     ) -> Result<Option<AgentWorkerRecord>> {
+        self.ensure_coordination_process_lock()
+            .map_err(anyhow::Error::msg)?;
         if self.state_path.is_none() {
             return Ok(None);
         }
@@ -6300,6 +6534,8 @@ impl SubAgentManager {
     /// completes.  Callers may `.join()` for synchronous semantics or drop it
     /// for fire-and-forget.
     fn persist_state(&self) -> Result<std::thread::JoinHandle<()>> {
+        self.ensure_coordination_process_lock()
+            .map_err(anyhow::Error::msg)?;
         let Some((_, payload)) = self.build_persist_payload()? else {
             return Ok(std::thread::spawn(|| {}));
         };
@@ -6307,7 +6543,16 @@ impl SubAgentManager {
             .store_authority
             .clone()
             .ok_or_else(|| anyhow!("delegated store authority is unavailable"))?;
+        let writer_locks = self
+            .coordination_process_lock
+            .lock()
+            .expect("coordination lock slot poisoned")
+            .iter()
+            .cloned()
+            .chain(self.workspace_writer_locks.values().cloned())
+            .collect::<Vec<_>>();
         Ok(std::thread::spawn(move || {
+            let _writer_locks = writer_locks;
             if let Err(error) = authority.publish(&payload, None, None) {
                 tracing::warn!(target: "subagent", %error, "failed to persist sub-agent checkpoint");
             }
@@ -6315,6 +6560,8 @@ impl SubAgentManager {
     }
 
     fn persist_state_synchronously(&self) -> Result<()> {
+        self.ensure_coordination_process_lock()
+            .map_err(anyhow::Error::msg)?;
         let Some((_, payload)) = self.build_persist_payload()? else {
             return Ok(());
         };
@@ -6377,6 +6624,8 @@ impl SubAgentManager {
     /// Unlike `persist_state`, this performs disk I/O **synchronously** to
     /// guarantee data is flushed before the process exits.
     pub fn flush_pending_persist(&mut self) -> Result<()> {
+        self.ensure_coordination_process_lock()
+            .map_err(anyhow::Error::msg)?;
         if self.persist_pending {
             self.persist_state_synchronously()?;
             self.last_persist_at = Some(Instant::now());
@@ -6599,7 +6848,8 @@ impl SubAgentManager {
     }
 
     pub fn finalize_session_close_for_session(&mut self, active_session_id: &str) -> Result<usize> {
-        if (self.state_path.is_none() || self.store_authority.is_some())
+        if !self.coordination_snapshot_read_only
+            && (self.state_path.is_none() || self.store_authority.is_some())
             && !self.agents.values().any(|agent| {
                 self.owns_store_record(&agent.id)
                     && self.agent_is_owned_by_session(agent, active_session_id)
@@ -6858,9 +7108,10 @@ impl SubAgentManager {
         let Some((claim, isolated_worktree)) = worker_coordination_claim(spec)? else {
             return Ok(());
         };
-        if !isolated_worktree {
-            self.refuse_live_selected_store_writers(origin)?;
-        }
+        // Isolated-worktree builders mutate their own checkout, not the
+        // shared workspace, so they must not contend for the per-workspace
+        // coordination process lock (#5036).
+        self.ensure_worker_coordination_locks(spec, origin)?;
         let claim = self.namespace_write_claim(&origin_workspace, isolated_worktree, claim)?;
         let scope = if isolated_worktree {
             None
@@ -6940,7 +7191,11 @@ impl SubAgentManager {
         let previous_origins = self.worker_origins.clone();
         let previous_schema = self.worker_origin_schema;
         let claim = worker_coordination_claim(&spec)?;
-        self.admit_worker_shared_writes(&spec, &origin)?;
+        // Same isolation rule as `preflight_worker_coordination` (#5036).
+        let isolated_worktree = claim
+            .as_ref()
+            .is_some_and(|(_, isolated_worktree)| *isolated_worktree);
+        self.ensure_worker_coordination_locks(&spec, &origin)?;
         let previous_worker_records = self.worker_records.clone();
         let previous_coordination = self.coordination.clone();
         let persisted_claim = claim
@@ -6992,6 +7247,18 @@ impl SubAgentManager {
         let worker_id_for_log = spec.worker_id.clone();
         self.register_worker_for_session(spec, &origin.owner_session_id, None);
         self.install_worker_origin(&worker_id_for_log, origin);
+        if isolated_worktree && self.ensure_coordination_process_lock().is_err() {
+            // This process does not own the durable coordination state and an
+            // isolated-worktree builder must not be blocked on it: register
+            // in memory only and leave the durable ledger to the lock owner.
+            // The persist-layer lock check stays intact (#5036).
+            tracing::warn!(
+                target: "subagent",
+                worker = %worker_id_for_log,
+                "registered isolated-worktree worker in memory only; delegated coordination lock is owned elsewhere"
+            );
+            return Ok(());
+        }
         if let Err(error) = self.persist_state_synchronously() {
             self.worker_records = previous_worker_records;
             self.coordination = previous_coordination;
@@ -7417,6 +7684,7 @@ impl SubAgentManager {
                 .map_err(|error| error.to_string());
         }
 
+        self.ensure_coordination_process_lock()?;
         let worker_id = spec.worker_id.clone();
         let previous = self
             .worker_records
@@ -10059,9 +10327,20 @@ impl SubAgentManager {
                 ));
             }
         }
-        if write_capable && !options.isolated_worktree {
-            self.refuse_live_selected_store_writers(&origin)
-                .map_err(anyhow::Error::msg)?;
+        if write_capable {
+            // Isolated-worktree children mutate their own checkout, so they
+            // do not contend for the shared-workspace process lock (#5036).
+            if !options.isolated_worktree {
+                self.ensure_coordination_process_lock()
+                    .map_err(anyhow::Error::msg)?;
+                self.ensure_origin_writer_lock(&origin)
+                    .map_err(anyhow::Error::msg)?;
+            }
+            if self.coordination_process_lock_required && self.state_path.is_none() {
+                return Err(anyhow!(
+                    "write-capable sub-agent launch requires a durable coordination state path"
+                ));
+            }
         }
         let runtime_origin_turn_id = if runtime.context.runtime.active_thread_id.as_deref()
             == Some(runtime.context.state_namespace.as_str())
@@ -10295,9 +10574,7 @@ impl SubAgentManager {
                 self.resume_targets.insert(source.clone(), agent_id.clone());
             }
             self.agents.insert(agent_id.clone(), agent);
-            let persist_result = self
-                .admitted_store_authority()
-                .and_then(|_| self.persist_state_synchronously());
+            let persist_result = self.persist_state_synchronously();
             agent = self
                 .agents
                 .remove(&agent_id)
@@ -10610,6 +10887,8 @@ impl SubAgentManager {
         identity: &ChildActivationIdentity,
     ) -> Result<impl FnOnce() -> Result<()> + Send + 'static + use<>> {
         self.exact_activation(identity)?;
+        self.ensure_coordination_process_lock()
+            .map_err(anyhow::Error::msg)?;
         let (path, payload) = self
             .build_persist_payload()?
             .ok_or_else(|| anyhow!("child activation has no durable worker store"))?;
@@ -10634,7 +10913,16 @@ impl SubAgentManager {
             .clone()
             .ok_or_else(|| anyhow!("child activation has no store authority"))?;
         let identity = identity.clone();
+        let writer_locks = self
+            .coordination_process_lock
+            .lock()
+            .expect("coordination lock slot poisoned")
+            .iter()
+            .cloned()
+            .chain(self.workspace_writer_locks.values().cloned())
+            .collect::<Vec<_>>();
         Ok(move || {
+            let _writer_locks = writer_locks;
             store_authority.publish(&payload, None, None)?;
             let persisted = read_durable_subagent_store(&state_root, &path)?.state;
             let worker = persisted
@@ -11221,10 +11509,13 @@ impl SubAgentManager {
             }
         }
         // Deliberately NOT here: a pass that terminalized every Running agent
-        // without a live task handle because another session shares the
-        // workspace (#2.6). Two Codewhale sessions in one workspace is ordinary
-        // usage, and liveness is decided by the heartbeat above — which is
-        // actual evidence (owner report, 2026-08-04).
+        // without a live task handle whenever this process lacked the
+        // coordination flock (#2.6). Two Codewhale sessions in one workspace is
+        // ordinary usage, and failure to append to a shared bookkeeping ledger
+        // is not evidence that live work has stopped. Liveness is decided by
+        // the heartbeat above — which is actual evidence — never by who owns
+        // the flock (owner report, 2026-08-04). See `docs/architecture/
+        // delegated-coordination.md` for what lock loss legitimately costs.
 
         // The retained lineage owns continuation lookup and descendant
         // control. Keep the agents projection while live work references it.
@@ -11354,12 +11645,11 @@ impl SubAgentManager {
     }
 
     /// #3803: whether the next sidebar refresh should attempt cleanup.
-    /// A manager whose durable store is not admitted only renders its refusal;
-    /// admission is retried on the next delegated write. Other managers use the
-    /// last attempt's interval to bound cleanup locking.
+    /// Read-only managers without a process lock only render their refusal.
+    /// Other managers use the last attempt's interval to bound cleanup locking.
     #[must_use]
     pub fn cleanup_due(&self, min_interval: Duration) -> bool {
-        (self.state_path.is_none() || self.store_authority.is_some())
+        (!self.coordination_snapshot_read_only || self.holds_coordination_process_lock())
             && self
                 .last_cleanup_at
                 .is_none_or(|last| last.elapsed() >= min_interval)
@@ -13500,6 +13790,22 @@ fn ensure_subagent_store_cutover(root: &Path, path: &Path) -> Result<()> {
         let _ = codewhale_config::private_directory::PrivateDirectory::inspect(&legacy)?;
         return Ok(());
     }
+    let _migration = if legacy.is_dir() {
+        None
+    } else {
+        Some(
+            CoordinationProcessLock::acquire(root, true)
+                .map_err(|error| anyhow::Error::new(SubAgentCutoverBlocked(error)))?,
+        )
+    };
+    if path.exists() {
+        read_durable_subagent_store(root, path)?;
+        anyhow::ensure!(
+            legacy.is_dir(),
+            "delegated migration did not retain its legacy-write tombstone"
+        );
+        return Ok(());
+    }
     let staging = parent.join("subagents.v1.cutover");
     let backup_name = "legacy.json";
     let original_name = "original.json";
@@ -13576,20 +13882,6 @@ fn ensure_subagent_store_cutover(root: &Path, path: &Path) -> Result<()> {
     write_json_atomic(root, path, &DurableSubAgentStore::from_legacy(state))
 }
 
-fn write_state_gitignore(root: &Path, path: &Path) -> Result<()> {
-    let ignore = checked_subagent_state_path(root, &path.with_file_name(".gitignore"))?;
-    match fs::OpenOptions::new()
-        .write(true)
-        .create_new(true)
-        .open(&ignore)
-    {
-        Ok(mut file) => file.write_all(b"*\n")?,
-        Err(error) if error.kind() == std::io::ErrorKind::AlreadyExists => {}
-        Err(error) => return Err(error.into()),
-    }
-    Ok(())
-}
-
 fn store_writer_is_live(path: &Path, writer: &str) -> bool {
     if !valid_store_writer(writer) {
         return true;
@@ -13662,9 +13954,6 @@ impl SubAgentStoreAuthority {
         let _opening = acquire_store_transaction(root, path)?;
         ensure_subagent_store_cutover(root, path)?;
         let image = read_durable_subagent_store(root, path)?;
-        if let Err(error) = write_state_gitignore(root, path) {
-            tracing::warn!(target: "subagent", %error, "could not ignore the delegated state directory");
-        }
         sync_subagent_store_directory(
             path.parent()
                 .ok_or_else(|| anyhow!("delegated store has no parent"))?,
@@ -14205,16 +14494,24 @@ pub fn new_shared_subagent_manager_with_timeout(
     )
 }
 
+pub fn new_shared_runtime_subagent_manager(
+    workspace: PathBuf,
+    workers: usize,
+) -> Result<SharedSubAgentManager> {
+    new_runtime_subagent_manager(workspace, workers, false)
+}
+
 pub fn new_exclusive_runtime_subagent_manager(
     workspace: PathBuf,
     workers: usize,
 ) -> Result<SharedSubAgentManager> {
-    new_shared_runtime_subagent_manager(workspace, workers)
+    new_runtime_subagent_manager(workspace, workers, true)
 }
 
-pub fn new_shared_runtime_subagent_manager(
+fn new_runtime_subagent_manager(
     workspace: PathBuf,
     workers: usize,
+    exclusive: bool,
 ) -> Result<SharedSubAgentManager> {
     type RuntimeManagerRegistry =
         parking_lot::Mutex<HashMap<PathBuf, (usize, std::sync::Weak<RwLock<SubAgentManager>>)>>;
@@ -14239,6 +14536,12 @@ pub fn new_shared_runtime_subagent_manager(
         .with_admission_limit(workers)
         .with_launch_concurrency(workers)
         .with_state_path(state_path);
+    manager.coordination_exclusive = exclusive;
+    manager = manager.require_coordination_process_lock();
+    if let Err(error) = manager.ensure_coordination_process_lock() {
+        manager.coordination_snapshot_read_only = true;
+        tracing::warn!(target: "subagent", %error, "Runtime coordination is read-only; stop and restart this Runtime process after the other session exits to load current state and regain delegated writes");
+    }
     manager.load_state()?;
     let manager = Arc::new(RwLock::new(manager));
     managers.insert(key, (workers, Arc::downgrade(&manager)));
@@ -14248,7 +14551,7 @@ pub fn new_shared_runtime_subagent_manager(
 /// Create a shared sub-agent manager with an explicit control-plane state root.
 ///
 /// `workspace` remains the child execution and file-authority root. The worker
-/// ledger and complete transcript artifacts live under
+/// ledger, complete transcript artifacts and coordination lock live under
 /// `state_root/.codewhale/state`. Distinct state roots intentionally do not
 /// share write claims; an embedding host that runs them against the same
 /// workspace must provide any required cross-session write coordination.
@@ -14291,8 +14594,25 @@ pub fn new_shared_subagent_manager_with_state_root_and_timeout(
     if let Some(state_path) = state_path {
         manager = manager.with_state_path(state_path);
     }
+    manager = if manager.store_cutover_blocked {
+        manager.into_session_private()
+    } else {
+        manager.require_coordination_process_lock()
+    };
+    if let Err(error) = manager.ensure_coordination_process_lock() {
+        tracing::warn!(target: "subagent", %error, "this session cannot append to the shared coordination ledger");
+    }
+    // Loading is a read. It was previously gated behind the *write* flock,
+    // which left a second session in the same workspace blind to every
+    // decision and write claim already recorded — and holding a default,
+    // empty ledger that would be written straight over the real one the
+    // moment the first session exited and the flock became acquirable
+    // (owner report, 2026-08-04). Read unconditionally; the write gate on
+    // `persist_state`/`persist_state_synchronously` still checks this
+    // manager's coordinator admission before publishing.
     if let Err(err) = manager.load_state() {
-        tracing::warn!(target: "subagent", ?err, "failed to load sub-agent state");
+        manager.coordination_snapshot_read_only = true;
+        tracing::warn!(target: "subagent", ?err, "failed to load sub-agent state; durable writes remain refused");
     }
     Arc::new(RwLock::new(manager))
 }
@@ -22833,29 +23153,10 @@ mod owner_scoped_coordination_tests {
         let mut manager = authority.try_write().unwrap();
         admit(&mut manager, &original);
         admit(&mut manager, &selected);
-        let independent = SubAgentManager::new(selected.clone(), 2)
-            .with_state_path(default_state_path(&selected).unwrap());
-        assert!(independent.store_authority.is_some());
-        #[cfg(unix)]
-        {
-            use std::os::unix::fs::PermissionsExt;
-            let state_directory = default_state_path(&selected)
-                .unwrap()
-                .parent()
-                .unwrap()
-                .to_path_buf();
-            let mode = fs::metadata(&state_directory).unwrap().permissions().mode();
-            fs::set_permissions(&state_directory, fs::Permissions::from_mode(0o000)).unwrap();
-            let unreadable =
-                manager.preflight_worker_coordination(&writer("selected", &selected, "src"));
-            fs::set_permissions(&state_directory, fs::Permissions::from_mode(mode)).unwrap();
-            assert!(unreadable.is_err(), "an unlistable store must fail closed");
-        }
-        let refused = manager
-            .preflight_worker_coordination(&writer("selected", &selected, "src"))
-            .unwrap_err();
-        assert!(refused.contains("another Codewhale session"), "{refused}");
-        drop(independent);
+        let independent =
+            SubAgentManager::new(selected.clone(), 2).require_coordination_process_lock();
+        assert!(independent.session_private);
+        assert!(independent.ensure_coordination_process_lock().is_ok());
         manager
             .preflight_worker_coordination(&writer("original", &original, "src"))
             .unwrap();

crates/engine/src/tools/subagent/tests.rs

@@ -1223,18 +1223,21 @@ fn coordination_hot_paths_count_only_active_authoritative_owners() {
 }
 
 #[test]
-fn runtime_mount_and_parallel_sessions_both_append_durably() {
+fn runtime_mount_coordination_lock_is_shared_with_parallel_sessions() {
     let tmp = tempdir().expect("tempdir");
     let runtime_authority = new_shared_runtime_subagent_manager(tmp.path().to_path_buf(), 8)
         .expect("admit TUI runtime writer");
-    let session = SubAgentManager::new(tmp.path().to_path_buf(), 8)
-        .with_state_path(default_state_path(tmp.path()).expect("state path"));
-    assert!(session.store_authority.is_some());
+    let session =
+        SubAgentManager::new(tmp.path().to_path_buf(), 8).require_coordination_process_lock();
+    session
+        .ensure_coordination_process_lock()
+        .expect("a second session appends beside the runtime");
+    assert!(!session.coordination_snapshot_read_only);
     session
         .persist_state_synchronously()
         .expect("session append");
     let runtime = runtime_authority.try_write().expect("runtime lock");
-    assert!(runtime.store_authority.is_some());
+    assert!(!runtime.coordination_snapshot_read_only);
     runtime
         .persist_state_synchronously()
         .expect("runtime append");
@@ -1360,9 +1363,10 @@ fn headless_worker_registration_enforces_live_claims_and_projects_context() {
             .iter()
             .any(|record| record.owner_session_id == "owner-worker-c")
     );
-    let independent = SubAgentManager::new(tmp.path().to_path_buf(), 8)
-        .with_state_path(default_state_path(tmp.path()).expect("state path"));
-    assert!(independent.store_authority.is_some());
+    let independent =
+        SubAgentManager::new(tmp.path().to_path_buf(), 8).require_coordination_process_lock();
+    assert!(independent.session_private);
+    assert!(independent.ensure_coordination_process_lock().is_ok());
     drop(independent);
     drop(authority_a);
     drop(authority_b);
@@ -1571,21 +1575,25 @@ fn prior_session_owners_never_count_as_active_coordination_owners() {
 }
 
 #[test]
-fn parallel_sessions_admit_shared_and_isolated_writers_durably() {
+fn isolated_worktree_workers_skip_the_coordination_process_lock() {
     let tmp = tempdir().expect("tempdir");
     let workspace = tmp.path().canonicalize().expect("canonical workspace");
-    let state_path = default_state_path(&workspace).expect("state path");
 
-    let holder = SubAgentManager::new(workspace.clone(), 4).with_state_path(state_path.clone());
-    assert!(holder.store_authority.is_some());
+    let holder = SubAgentManager::new(workspace.clone(), 4).require_coordination_process_lock();
+    holder
+        .ensure_coordination_process_lock()
+        .expect("holder owns the lock");
 
-    let mut contender = SubAgentManager::new(workspace.clone(), 4).with_state_path(state_path);
-    assert!(contender.store_authority.is_some());
+    let mut contender =
+        SubAgentManager::new(workspace.clone(), 4).require_coordination_process_lock();
+    contender
+        .ensure_coordination_process_lock()
+        .expect("second manager now also owns shared lock (coexistence)");
 
     let shared = make_write_worker_spec("shared-writer", workspace.clone(), "src/shared");
     contender
         .preflight_worker_coordination(&shared)
-        .expect("shared-workspace writer proceeds beside another session");
+        .expect("shared-workspace writer proceeds with shared lock");
 
     let mut isolated = make_write_worker_spec("isolated-writer", workspace.clone(), "src/iso");
     isolated
@@ -1595,64 +1603,69 @@ fn parallel_sessions_admit_shared_and_isolated_writers_durably() {
         .worktree = true;
     contender
         .preflight_worker_coordination(&isolated)
-        .expect("isolated-worktree writer preflights beside another session");
+        .expect("isolated-worktree writer preflights without the lock");
     contender
         .register_worker_with_coordination(isolated)
-        .expect("isolated-worktree writer registers durably");
+        .expect("isolated-worktree writer registers without the lock");
+
+    // Once the previous owner exits, a manager that failed earlier must
+    // acquire the lock on retry instead of replaying a memoized failure.
     drop(holder);
+    contender
+        .ensure_coordination_process_lock()
+        .expect("lock acquisition retries after the holder exits");
     contender
         .preflight_worker_coordination(&shared)
-        .expect("shared-workspace writer proceeds after the other session exits");
+        .expect("shared-workspace writer proceeds once the lock is held");
 }
 
 #[test]
-fn coordination_detail_projection_reports_store_admission_and_its_retry() {
+fn coordination_detail_projection_reports_process_lock_ownership() {
     let tmp = tempdir().expect("tempdir");
     let workspace = tmp.path().canonicalize().expect("canonical workspace");
-    let state_path = default_state_path(&workspace).expect("state path");
 
-    let holder = SubAgentManager::new(workspace.clone(), 4).with_state_path(state_path.clone());
+    let holder = SubAgentManager::new(workspace.clone(), 4).require_coordination_process_lock();
+    holder
+        .ensure_coordination_process_lock()
+        .expect("holder owns the lock");
     let held = holder.coordination_detail_projection(None, 8);
-    assert!(held.process_lock_held);
+    assert!(
+        held.process_lock_held,
+        "holder projection must report lock held"
+    );
     assert!(held.process_lock_note.is_none());
 
-    let contender = SubAgentManager::new(workspace.clone(), 4).with_state_path(state_path.clone());
+    let contender = SubAgentManager::new(workspace, 4).require_coordination_process_lock();
     let shared = contender.coordination_detail_projection(None, 8);
-    assert!(shared.process_lock_held);
-    assert!(shared.process_lock_note.is_none());
-
-    let committed = std::fs::read(&state_path).expect("committed store");
-    std::fs::write(&state_path, b"{").expect("corrupt store");
-    let mut refused = SubAgentManager::new(workspace, 4).with_state_path(state_path.clone());
-    let projection = refused.coordination_detail_projection(None, 8);
-    assert!(!projection.process_lock_held);
-    assert!(projection.process_lock_note.is_some());
-    assert!(!refused.cleanup_due(Duration::ZERO));
-    assert!(refused.release_stale_write_claims(None).is_err());
-    assert!(!refused.cleanup_due(Duration::ZERO));
-
-    std::fs::write(&state_path, committed).expect("restore store");
-    refused
-        .release_stale_write_claims(None)
-        .expect("admission is retried once the store is readable");
+    // With shared locks both holders coexist — contender reports held and no note.
     assert!(
-        refused
-            .coordination_detail_projection(None, 8)
-            .process_lock_held
+        shared.process_lock_held,
+        "contender projection must report lock held with shared locks"
+    );
+    assert!(
+        shared.process_lock_note.is_none(),
+        "shared projection carries no note"
     );
-    assert!(refused.cleanup_due(Duration::ZERO));
 }
 
-/// A second Codewhale session in the same workspace is ordinary usage and
-/// must never mean "my agents are dead". Bookkeeping and liveness are separate
-/// concerns (owner report, 2026-08-04).
+/// A second Codewhale session in the same workspace is ordinary usage. Losing
+/// the coordination flock means "I cannot append to the shared ledger" — it
+/// must never mean "my agents are dead". Bookkeeping failure and liveness are
+/// separate concerns (owner report, 2026-08-04).
 #[test]
-fn a_second_session_keeps_its_running_agents_live() {
+fn a_second_session_without_the_lock_keeps_its_running_agents_live() {
     let tmp = tempdir().expect("tempdir");
     let workspace = tmp.path().canonicalize().expect("canonical workspace");
 
-    let _first = SubAgentManager::new(workspace.clone(), 4);
-    let mut contender = SubAgentManager::new(workspace, 4);
+    let holder = SubAgentManager::new(workspace.clone(), 4).require_coordination_process_lock();
+    holder
+        .ensure_coordination_process_lock()
+        .expect("holder owns the lock");
+
+    let mut contender = SubAgentManager::new(workspace, 4).require_coordination_process_lock();
+    contender
+        .ensure_coordination_process_lock()
+        .expect("contender now also owns shared lock");
 
     let (input_tx, _input_rx) = mpsc::unbounded_channel();
     let mut agent = SubAgent::new(
@@ -1667,6 +1680,9 @@ fn a_second_session_keeps_its_running_agents_live() {
         PathBuf::from("."),
         contender.session_boot_id().to_string(),
     );
+    // Fresh, well inside the heartbeat window: nothing about this agent is
+    // stale. The only thing "wrong" with it is that this process cannot write
+    // the shared ledger.
     agent.last_activity_at = Instant::now();
     let agent_id = agent.id.clone();
     contender.agents.insert(agent_id.clone(), agent);
@@ -1676,7 +1692,7 @@ fn a_second_session_keeps_its_running_agents_live() {
         .expect("durable worker operation");
     assert_eq!(
         cancelled, 0,
-        "another session alone must not terminalize anything"
+        "lock loss alone must not terminalize anything"
     );
     let snap = contender
         .get_result(&agent_id)
@@ -1684,21 +1700,26 @@ fn a_second_session_keeps_its_running_agents_live() {
     assert_eq!(
         snap.status,
         SubAgentStatus::Running,
-        "a live agent stays Running beside another session: {:?}",
+        "a live agent stays Running when this session cannot write the ledger: {:?}",
         snap.status
     );
 }
 
-/// A second session must load the existing ledger before it persists any of
-/// its own state. That preserves prior decisions instead of replacing them
-/// with an empty ledger.
+/// A second session shares the workspace coordination read-lock and must load
+/// the existing ledger before it persists any of its own state. That preserves
+/// prior decisions instead of replacing them with an empty ledger.
 #[test]
-fn a_second_session_loads_the_workspace_ledger() {
+fn a_second_session_with_the_shared_lock_loads_the_workspace_ledger() {
     let tmp = tempdir().expect("tempdir");
     let workspace = tmp.path().canonicalize().expect("canonical workspace");
     let state_path = default_state_path(&workspace).expect("state path");
 
-    let mut first = SubAgentManager::new(workspace.clone(), 4).with_state_path(state_path.clone());
+    let mut first = SubAgentManager::new(workspace.clone(), 4)
+        .with_state_path(state_path.clone())
+        .require_coordination_process_lock();
+    first
+        .ensure_coordination_process_lock()
+        .expect("first session owns the lock");
     first
         .record_coordination_decision(DecisionRecord {
             decision_id: "shared-decision".to_string(),
@@ -1716,15 +1737,20 @@ fn a_second_session_loads_the_workspace_ledger() {
         .persist_state_synchronously()
         .expect("first session persists the ledger");
 
+    // Second session in the same workspace while the first still holds the
+    // shared flock. It coexists and must see the workspace ledger.
     let second = new_shared_subagent_manager(workspace.clone(), 4);
     {
         let guard = second.blocking_read();
-        assert!(guard.store_authority.is_some());
+        assert!(
+            guard.holds_coordination_process_lock(),
+            "test premise: the second session joins the shared flock"
+        );
         let ledger = guard.coordination_snapshot();
         assert_eq!(
             ledger.decisions.len(),
             1,
-            "second session must load the workspace ledger"
+            "second session must load the workspace ledger it cannot write"
         );
         assert_eq!(ledger.decisions[0].decision_id, "shared-decision");
     }
@@ -1733,9 +1759,12 @@ fn a_second_session_loads_the_workspace_ledger() {
     drop(first);
     {
         let guard = second.blocking_write();
+        guard
+            .ensure_coordination_process_lock()
+            .expect("shared lock remains available after the first session exits");
         guard
             .persist_state_synchronously()
-            .expect("second session persists after the first exits");
+            .expect("second session persists once it owns the lock");
     }
     let mut replayed = SubAgentManager::new(workspace, 4).with_state_path(state_path);
     replayed.load_state().expect("reload the workspace ledger");
@@ -1747,15 +1776,23 @@ fn a_second_session_loads_the_workspace_ledger() {
 }
 
 /// The complement of the test above: an orphan still terminalizes, but only on
-/// heartbeat evidence. Another session is not part of the decision either way.
+/// heartbeat evidence. Lock ownership is not part of the decision either way.
 #[test]
-fn cleanup_terminalizes_orphans_on_heartbeat_evidence_not_on_other_sessions() {
+fn cleanup_terminalizes_orphans_on_heartbeat_evidence_not_on_lock_ownership() {
     let tmp = tempdir().expect("tempdir");
     let workspace = tmp.path().canonicalize().expect("canonical workspace");
 
-    let _first = SubAgentManager::new(workspace.clone(), 4);
-    let mut contender =
-        SubAgentManager::new(workspace, 4).with_running_heartbeat_timeout(Duration::from_secs(30));
+    let holder = SubAgentManager::new(workspace.clone(), 4).require_coordination_process_lock();
+    holder
+        .ensure_coordination_process_lock()
+        .expect("holder owns the lock");
+
+    let mut contender = SubAgentManager::new(workspace, 4)
+        .with_running_heartbeat_timeout(Duration::from_secs(30))
+        .require_coordination_process_lock();
+    contender
+        .ensure_coordination_process_lock()
+        .expect("contender now also owns shared lock");
 
     let insert_orphan = |manager: &mut SubAgentManager, name: &str, idle: Duration| {
         let (input_tx, _input_rx) = mpsc::unbounded_channel();
@@ -1795,7 +1832,7 @@ fn cleanup_terminalizes_orphans_on_heartbeat_evidence_not_on_other_sessions() {
             .expect("fresh orphan still listed")
             .status,
         SubAgentStatus::Running,
-        "a fresh orphan is not terminalized just because another session is open"
+        "a fresh orphan is not terminalized just because the flock is elsewhere"
     );
     assert!(
         matches!(
@@ -11314,7 +11351,7 @@ async fn write_capable_agent_does_not_launch_when_durable_registration_fails() {
         .expect_err("writer must fail before spawn when its durable claim cannot commit")
         .to_string();
     assert!(
-        error.contains("durably register") || error.contains("store admission failed"),
+        error.contains("durably register") || error.contains("writer lock"),
         "{error}"
     );
 
@@ -12193,10 +12230,7 @@ fn persist_state_rejects_state_path_outside_state_root() {
         .persist_state()
         .expect_err("outside state path should fail");
 
-    assert!(
-        format!("{err:#}").contains("must stay within state root"),
-        "{err:#}"
-    );
+    assert!(format!("{err:#}").contains("without its writer lock"));
 }
 
 /// A user who keeps application state on another volume makes the workspace's
@@ -12371,7 +12405,9 @@ fn explicit_state_roots_isolate_managers_for_the_same_execution_workspace() {
         let mut manager = manager.try_write().expect("manager write lock");
         assert_eq!(manager.workspace.as_path(), workspace.as_path());
         assert_eq!(manager.state_root.as_path(), expected_root.as_path());
-        assert!(manager.store_authority.is_some());
+        manager
+            .ensure_coordination_process_lock()
+            .expect("independent state-root coordination lock");
         manager.register_worker(make_worker_spec(worker_id, workspace.clone()));
         manager
             .persist_state_synchronously()
@@ -12426,21 +12462,31 @@ fn persist_state_rejects_symlinked_state_directory() {
     );
 }
 
+/// A linked `.codewhale` must be refused before anything is created: the
+/// coordination lock used to `create_dir_all` through the link first.
 #[cfg(unix)]
 #[test]
-fn a_linked_codewhale_admission_writes_no_gitignore_through_it() {
+fn coordination_lock_refuses_linked_codewhale_without_writing_through_it() {
     let tmp = tempdir().expect("tempdir");
-    let root = tmp.path().canonicalize().expect("canonical root");
-    let workspace = root.join("workspace");
-    let outside = root.join("outside");
+    let workspace = tmp.path().join("workspace");
+    let outside = tmp.path().join("outside");
     std::fs::create_dir_all(&workspace).expect("mkdir workspace");
     std::fs::create_dir_all(&outside).expect("mkdir outside");
     std::os::unix::fs::symlink(&outside, workspace.join(".codewhale")).expect("link .codewhale");
-    let manager = new_shared_subagent_manager(workspace, 4);
-    assert!(manager.blocking_read().store_authority.is_none());
+
+    let Err(err) = CoordinationProcessLock::acquire(&workspace, false) else {
+        panic!("a linked .codewhale must be refused");
+    };
+    let message = format!("{err:#}");
     assert!(
-        !outside.join("state").join(".gitignore").exists(),
-        "a refused admission writes no .gitignore through the link"
+        message.contains("must not traverse symlinks")
+            || message.contains("must stay within state root"),
+        "{message}"
+    );
+    assert_eq!(
+        std::fs::read_dir(&outside).expect("read outside").count(),
+        0,
+        "nothing may be created through the link"
     );
 }
 
@@ -20828,66 +20874,24 @@ fn write_json_atomic_survives_concurrent_writers() {
     assert!(leftover.is_empty(), "temp files leaked: {leftover:?}");
 }
 
-fn two_process_decision(id: &str) -> DecisionRecord {
-    DecisionRecord {
-        decision_id: id.to_string(),
-        subject: id.to_string(),
-        status: DecisionStatus::Accepted,
-        owner: "root".to_string(),
-        scope: vec!["src".to_string()],
-        constraints: Vec::new(),
-        evidence_handles: Vec::new(),
-        version: 1,
-        sequence: 0,
-    }
-}
-
-fn wait_for_two_process_file(path: &Path, child: &mut std::process::Child, what: &str) {
-    let deadline = Instant::now() + Duration::from_secs(30);
-    while !path.exists() && Instant::now() < deadline {
-        std::thread::sleep(Duration::from_millis(10));
-    }
-    if !path.exists() {
-        let _ = child.kill();
-        panic!("{what} never became ready");
-    }
-}
-
 #[test]
-fn a_stale_session_never_makes_a_new_session_read_only() {
-    const ROLE_ENV: &str = "CODEWHALE_TEST_COORDINATION_ROLE";
-    const WORKSPACE_ENV: &str = "CODEWHALE_TEST_COORDINATION_WORKSPACE";
+fn coordination_process_lock_rejects_second_process() {
+    const ROLE_ENV: &str = "CODEWHALE_TEST_COORDINATION_LOCK_ROLE";
+    const WORKSPACE_ENV: &str = "CODEWHALE_TEST_COORDINATION_LOCK_WORKSPACE";
     const TEST_NAME: &str =
-        "tools::subagent::tests::a_stale_session_never_makes_a_new_session_read_only";
+        "tools::subagent::tests::coordination_process_lock_rejects_second_process";
 
     if let Some(role) = std::env::var_os(ROLE_ENV) {
         let workspace = PathBuf::from(std::env::var_os(WORKSPACE_ENV).expect("workspace env"));
-        let wait_for_release = || {
-            let deadline = Instant::now() + Duration::from_secs(60);
+        if role == "legacy-holder" {
+            let _legacy_session = CoordinationProcessLock::acquire(&workspace, false)
+                .expect("an older session holds the shared lock");
+            std::fs::write(workspace.join("holder.ready"), b"ready").unwrap();
+            let deadline = Instant::now() + Duration::from_secs(10);
             while !workspace.join("holder.release").exists() && Instant::now() < deadline {
                 std::thread::sleep(Duration::from_millis(10));
             }
             assert!(workspace.join("holder.release").exists(), "release timeout");
-        };
-        if role == "legacy-holder" {
-            let lock_file = std::fs::OpenOptions::new()
-                .create(true)
-                .truncate(false)
-                .read(true)
-                .write(true)
-                .open(
-                    workspace
-                        .join(".codewhale")
-                        .join("state")
-                        .join("subagents.v1.lock"),
-                )
-                .expect("open legacy lock");
-            let legacy = fd_lock::RwLock::new(lock_file);
-            let _guard = legacy
-                .try_read()
-                .expect("an older session holds the legacy lock");
-            std::fs::write(workspace.join("holder.ready"), b"ready").unwrap();
-            wait_for_release();
             return;
         }
         let manager = new_shared_subagent_manager_with_timeout(
@@ -20897,82 +20901,78 @@ fn a_stale_session_never_makes_a_new_session_read_only() {
             Duration::from_secs(30),
             4,
         );
-        let mut guard = manager.try_write().unwrap();
-        assert!(guard.store_authority.is_some(), "{role:?} store admitted");
-        assert!(
-            guard
-                .coordination_detail_projection(None, 8)
-                .process_lock_held
-        );
-        let name = role.to_string_lossy().into_owned();
-        guard
-            .record_coordination_decision(two_process_decision(&format!("{name}-decision")))
-            .expect("record a decision beside the other session");
-        guard.register_worker(make_worker_spec(
-            &format!("{name}-worker"),
-            workspace.clone(),
-        ));
-        guard
-            .persist_state_synchronously()
-            .expect("publish beside the other session");
         if role == "holder" {
+            manager
+                .try_read()
+                .unwrap()
+                .ensure_coordination_process_lock()
+                .expect("holder owns lock");
             std::fs::write(workspace.join("holder.ready"), b"ready").unwrap();
-            wait_for_release();
+            let deadline = Instant::now() + Duration::from_secs(10);
+            while !workspace.join("holder.release").exists() && Instant::now() < deadline {
+                std::thread::sleep(Duration::from_millis(10));
+            }
+            assert!(workspace.join("holder.release").exists(), "release timeout");
         } else {
-            assert!(
-                guard
-                    .coordination_snapshot()
-                    .decisions
-                    .iter()
-                    .any(|decision| decision.decision_id == "holder-decision"),
-                "the new session merges the open session's ledger"
-            );
+            manager
+                .try_read()
+                .unwrap()
+                .ensure_coordination_process_lock()
+                .expect("second process now also succeeds with shared lock");
         }
         return;
     }
 
+    let dir = tempdir().expect("tempdir");
+    let workspace = dir.path().canonicalize().expect("canonical workspace");
     let test_binary = std::env::current_exe().expect("test binary");
-    let spawn = |role: &str, workspace: &Path| {
-        std::process::Command::new(&test_binary)
-            .arg("--exact")
-            .arg(TEST_NAME)
-            .arg("--nocapture")
-            .env(ROLE_ENV, role)
-            .env(WORKSPACE_ENV, workspace)
-            .stdout(std::process::Stdio::piped())
-            .stderr(std::process::Stdio::piped())
-            .spawn()
-            .expect("spawn session process")
-    };
-    let finish = |child: std::process::Child, what: &str| {
-        let output = child.wait_with_output().expect("session output");
-        assert!(
-            output.status.success(),
-            "{what} failed:\nstdout:\n{}\nstderr:\n{}",
+    let mut holder = std::process::Command::new(&test_binary)
+        .arg("--exact")
+        .arg(TEST_NAME)
+        .arg("--nocapture")
+        .env(ROLE_ENV, "holder")
+        .env(WORKSPACE_ENV, &workspace)
+        .stdout(std::process::Stdio::piped())
+        .stderr(std::process::Stdio::piped())
+        .spawn()
+        .expect("spawn lock holder");
+    let deadline = Instant::now() + Duration::from_secs(5);
+    while !workspace.join("holder.ready").exists() && Instant::now() < deadline {
+        std::thread::sleep(Duration::from_millis(10));
+    }
+    if !workspace.join("holder.ready").exists() {
+        let _ = holder.kill();
+        let output = holder.wait_with_output().expect("holder output");
+        panic!(
+            "holder never acquired lock:\nstdout:\n{}\nstderr:\n{}",
             String::from_utf8_lossy(&output.stdout),
             String::from_utf8_lossy(&output.stderr)
         );
-    };
-
-    let dir = tempdir().expect("tempdir");
-    let workspace = dir.path().canonicalize().expect("canonical workspace");
-    let mut holder = spawn("holder", &workspace);
-    wait_for_two_process_file(&workspace.join("holder.ready"), &mut holder, "open session");
-    finish(spawn("contender", &workspace), "new session");
-    std::fs::write(workspace.join("holder.release"), b"release").unwrap();
-    finish(holder, "open session");
-    let workers = load_persisted_agent_worker_records_with_state_root(&workspace, &workspace)
-        .expect("read merged store");
-    for id in ["holder-worker", "contender-worker"] {
-        assert!(
-            workers.iter().any(|record| record.spec.worker_id == id),
-            "{id} survives the other session's publication"
-        );
     }
-    let mut replayed = SubAgentManager::new(workspace.clone(), 4)
-        .with_state_path(default_state_path(&workspace).unwrap());
-    replayed.load_state().expect("reload merged ledger");
-    assert_eq!(replayed.coordination.decisions.len(), 2);
+
+    let contender = std::process::Command::new(&test_binary)
+        .arg("--exact")
+        .arg(TEST_NAME)
+        .arg("--nocapture")
+        .env(ROLE_ENV, "contender")
+        .env(WORKSPACE_ENV, &workspace)
+        .output()
+        .expect("spawn lock contender");
+    assert!(
+        contender.status.success(),
+        "contender failed:\nstdout:\n{}\nstderr:\n{}",
+        String::from_utf8_lossy(&contender.stdout),
+        String::from_utf8_lossy(&contender.stderr)
+    );
+
+    std::fs::write(workspace.join("holder.release"), b"release").unwrap();
+    let output = holder.wait_with_output().expect("holder output");
+    assert!(
+        output.status.success(),
+        "holder failed:\nstdout:\n{}\nstderr:\n{}",
+        String::from_utf8_lossy(&output.stdout),
+        String::from_utf8_lossy(&output.stderr)
+    );
 
     let legacy_dir = tempdir().expect("tempdir");
     let legacy_workspace = legacy_dir
@@ -20988,42 +20988,75 @@ fn a_stale_session_never_makes_a_new_session_read_only() {
     )
     .unwrap();
     let legacy_bytes = std::fs::read(&legacy_file).unwrap();
-    let mut legacy_holder = spawn("legacy-holder", &legacy_workspace);
-    wait_for_two_process_file(
-        &legacy_workspace.join("holder.ready"),
-        &mut legacy_holder,
-        "older session",
-    );
-    let upgraded = new_shared_subagent_manager(legacy_workspace.clone(), 4);
+    let mut legacy_holder = std::process::Command::new(&test_binary)
+        .arg("--exact")
+        .arg(TEST_NAME)
+        .arg("--nocapture")
+        .env(ROLE_ENV, "legacy-holder")
+        .env(WORKSPACE_ENV, &legacy_workspace)
+        .stdout(std::process::Stdio::piped())
+        .stderr(std::process::Stdio::piped())
+        .spawn()
+        .expect("spawn legacy lock holder");
+    let deadline = Instant::now() + Duration::from_secs(5);
+    while !legacy_workspace.join("holder.ready").exists() && Instant::now() < deadline {
+        std::thread::sleep(Duration::from_millis(10));
+    }
+    if !legacy_workspace.join("holder.ready").exists() {
+        let _ = legacy_holder.kill();
+        let output = legacy_holder
+            .wait_with_output()
+            .expect("legacy holder output");
+        panic!(
+            "legacy holder never acquired lock:\nstdout:\n{}\nstderr:\n{}",
+            String::from_utf8_lossy(&output.stdout),
+            String::from_utf8_lossy(&output.stderr)
+        );
+    }
+    let private = new_shared_subagent_manager(legacy_workspace.clone(), 4);
     {
-        let mut guard = upgraded.blocking_write();
+        let guard = private.blocking_read();
         assert!(
-            guard.store_authority.is_some(),
-            "a stale older session must not leave the new one read-only"
+            guard.ensure_coordination_process_lock().is_ok(),
+            "a session beside an older one keeps working sub-agents"
         );
         let projection = guard.coordination_detail_projection(None, 8);
+        assert!(!projection.process_lock_held);
         assert!(
-            projection.process_lock_held,
+            projection
+                .process_lock_note
+                .as_deref()
+                .is_some_and(|note| note.starts_with(SESSION_PRIVATE_COORDINATION)),
             "{:?}",
             projection.process_lock_note
         );
-        assert!(projection.process_lock_note.is_none());
-        guard
-            .record_coordination_decision(two_process_decision("upgraded-decision"))
-            .expect("the new session writes while the older one is still open");
-        assert!(guard.cleanup_due(Duration::from_secs(3600)));
     }
-    assert!(
-        state_dir.join(SUBAGENT_STATE_FILE).exists() && legacy_file.is_dir(),
-        "the store upgrades without waiting for the older session to exit"
-    );
     assert_eq!(
-        std::fs::read(legacy_file.join("legacy.json")).unwrap(),
+        std::fs::read(&legacy_file).unwrap(),
         legacy_bytes,
-        "the older store is retained byte for byte"
+        "the older session's store stays untouched"
+    );
+    assert!(
+        !state_dir.join(SUBAGENT_STATE_FILE).exists(),
+        "no forced cutover while the older session runs"
     );
     std::fs::write(legacy_workspace.join("holder.release"), b"release").unwrap();
-    finish(legacy_holder, "older session");
+    let output = legacy_holder
+        .wait_with_output()
+        .expect("legacy holder output");
+    assert!(
+        output.status.success(),
+        "legacy holder failed:\nstdout:\n{}\nstderr:\n{}",
+        String::from_utf8_lossy(&output.stdout),
+        String::from_utf8_lossy(&output.stderr)
+    );
+    let upgraded = new_shared_subagent_manager(legacy_workspace, 4);
+    assert!(
+        state_dir.join(SUBAGENT_STATE_FILE).exists() && legacy_file.is_dir(),
+        "the next session upgrades the store once the older one exits"
+    );
+    assert!(upgraded.blocking_read().holds_coordination_process_lock());
+    drop(private);
 }
 
 // === agent(action="wait") + peek throttling (#4097) ===
@@ -24138,32 +24171,35 @@ fn validate_resume_from_source(
 }
 
 /// Owner report 2026-08-04: a model/provider switch spawns the new engine's
-/// manager while the old engine's manager is still open in this same process.
-/// Both stay admitted, and the new one keeps writing after the old one drops.
+/// manager while the old engine still holds the workspace coordination flock
+/// in this same process. That losing acquisition must classify itself as a
+/// same-process handover (not "another Codewhale process"), and must
+/// self-heal via the projection retry once the previous owner drops.
 #[test]
-fn a_same_process_engine_handover_keeps_both_managers_admitted() {
+fn coordination_lock_loss_to_own_process_reads_as_handover_and_self_heals() {
     let tmp = tempdir().expect("tempdir");
-    let state_path = default_state_path(tmp.path()).expect("state path");
     let first =
-        SubAgentManager::new(tmp.path().to_path_buf(), 1).with_state_path(state_path.clone());
-    let second = SubAgentManager::new(tmp.path().to_path_buf(), 1).with_state_path(state_path);
+        SubAgentManager::new(tmp.path().to_path_buf(), 1).require_coordination_process_lock();
     assert!(
-        first
-            .coordination_detail_projection(None, 8)
-            .process_lock_held
+        first.holds_coordination_process_lock(),
+        "first manager must own the flock"
     );
+
+    let second =
+        SubAgentManager::new(tmp.path().to_path_buf(), 1).require_coordination_process_lock();
     assert!(
-        second
-            .coordination_detail_projection(None, 8)
-            .process_lock_held
+        second.holds_coordination_process_lock(),
+        "second manager now also holds shared flock (coexistence)"
     );
-    first
-        .persist_state_synchronously()
-        .expect("old engine publishes");
+    let note = second.coordination_process_lock_status();
+    assert!(note.is_ok(), "shared lock should not error: {note:?}");
+
     drop(first);
-    second
-        .persist_state_synchronously()
-        .expect("new engine publishes after the old one drops");
+    assert!(
+        second.coordination_process_lock_status().is_ok(),
+        "retry must acquire once the previous engine dropped the flock"
+    );
+    assert!(second.holds_coordination_process_lock());
 }
 
 // -- resume-from-checkpoint tests (checkpoint-based continuation) --

crates/localization/locales/ca.json

@@ -25,7 +25,6 @@
   "TerminalDockLive": "actiu",
   "ExtensionPromptUnavailable": "Les instruccions de les extensions no estan disponibles; consulta /plugin show.",
   "ProfileConstitutionUnavailableLocal": "No s'han pogut carregar les preferències del compte de Codewhale; aquest torn fa servir la configuració local. La clau del proveïdor continua funcionant; executa codewhale login per recuperar les preferències del compte.",
-  "AccountPreferencesNotLoaded": "Compte no carregat. Executa {command}. La clau del proveïdor continua funcionant.",
   "ComposerOversizedSubmitHeld": "No s’ha enviat: el missatge supera els {limit} caràcters i no s’ha pogut desar en un fitxer de text enganxat ({error}). El text complet continua al camp d’entrada: escurça’l o habilita l’escriptura a .codewhale/pastes i torna a prémer Enter.",
   "ComposerOversizedSubmitHeldShort": "No s’ha enviat: supera els {limit} caràcters; no s’ha desat el fitxer d’enganxat.",
   "CmdPluginOwnerReport": "Amfitrió d’extensions:\n  Estat: {state}\n  Eines actives ({count}): {tools}",
@@ -1668,7 +1667,7 @@
   "PhaseIdle": "inactiu",
   "PhaseDraft": "esborrany",
   "PhaseWorking": "treballant",
-  "PhaseReasoning": "pensant",
+  "PhaseReasoning": "raonant",
   "PhaseReading": "llegint",
   "PhaseUsingTool": "usant una eina",
   "PhaseSubagents": "subagents en execució",

crates/localization/locales/de.json

@@ -25,7 +25,6 @@
   "TerminalDockLive": "aktiv",
   "ExtensionPromptUnavailable": "Anweisungen von Erweiterungen sind nicht verfügbar; prüfe /plugin show.",
   "ProfileConstitutionUnavailableLocal": "Die Codewhale-Kontoeinstellungen konnten nicht geladen werden; dieser Durchlauf nutzt deine lokalen Einstellungen. Dein Anbieter-Schlüssel funktioniert weiter; führe codewhale login aus, um die Kontoeinstellungen wiederherzustellen.",
-  "AccountPreferencesNotLoaded": "Konto nicht geladen. Führe {command} aus. Dein Anbieter-Schlüssel funktioniert weiter.",
   "ComposerOversizedSubmitHeld": "Nicht gesendet: Diese Nachricht enthält mehr als {limit} Zeichen und konnte nicht als Einfügedatei gespeichert werden ({error}). Der vollständige Text steht noch im Eingabefeld: Kürze ihn oder mache .codewhale/pastes beschreibbar und drücke dann erneut Enter.",
   "ComposerOversizedSubmitHeldShort": "Nicht gesendet: über {limit} Zeichen; Einfügedatei nicht gespeichert.",
   "CmdPluginOwnerReport": "Erweiterungshost:\n  Status: {state}\n  Aktive Werkzeuge ({count}): {tools}",
@@ -1668,7 +1667,7 @@
   "PhaseIdle": "ruhend",
   "PhaseDraft": "entwurf",
   "PhaseWorking": "arbeitet",
-  "PhaseReasoning": "denkt nach",
+  "PhaseReasoning": "überlegt",
   "PhaseReading": "liest",
   "PhaseUsingTool": "nutzt Tool",
   "PhaseSubagents": "Subagenten laufen",

crates/localization/locales/en.json

@@ -25,7 +25,6 @@
   "TerminalDockLive": "live",
   "ExtensionPromptUnavailable": "Extension instructions are unavailable; inspect /plugin show.",
   "ProfileConstitutionUnavailableLocal": "Codewhale account preferences could not be loaded, so this turn uses your local settings. Your provider key still works; run codewhale login to restore account preferences.",
-  "AccountPreferencesNotLoaded": "Account not loaded. Run {command}. Your provider key still works.",
   "ComposerOversizedSubmitHeld": "Not sent: this message is over {limit} characters and could not be saved as a paste file ({error}). The full text is still in the composer: shorten it, or make .codewhale/pastes writable, then press Enter again.",
   "ComposerOversizedSubmitHeldShort": "Not sent: over {limit} characters; paste file not saved.",
   "CmdPluginOwnerReport": "Extension host:\n  State: {state}\n  Live tools ({count}): {tools}",
@@ -1691,7 +1690,7 @@
   "PhaseIdle": "idle",
   "PhaseDraft": "draft",
   "PhaseWorking": "working",
-  "PhaseReasoning": "thinking",
+  "PhaseReasoning": "reasoning",
   "PhaseReading": "reading",
   "PhaseUsingTool": "using tool",
   "PhaseSubagents": "agents underway",

crates/localization/locales/es-419.json

@@ -25,7 +25,6 @@
   "TerminalDockLive": "activo",
   "ExtensionPromptUnavailable": "Las instrucciones de las extensiones no están disponibles; revisa /plugin show.",
   "ProfileConstitutionUnavailableLocal": "No se pudieron cargar las preferencias de la cuenta de Codewhale; este turno usa tu configuración local. Tu clave del proveedor sigue funcionando; ejecuta codewhale login para restaurar las preferencias de la cuenta.",
-  "AccountPreferencesNotLoaded": "Cuenta no cargada. Ejecuta {command}. Tu clave del proveedor sigue funcionando.",
   "ComposerOversizedSubmitHeld": "No se envió: este mensaje supera los {limit} caracteres y no se pudo guardar en un archivo de texto pegado ({error}). El texto completo sigue en el campo de entrada: acórtalo o habilita la escritura en .codewhale/pastes y vuelve a presionar Enter.",
   "ComposerOversizedSubmitHeldShort": "No se envió: supera los {limit} caracteres; no se guardó el archivo pegado.",
   "CmdPluginOwnerReport": "Host de extensiones:\n  Estado: {state}\n  Herramientas activas ({count}): {tools}",
@@ -1691,7 +1690,7 @@
   "PhaseIdle": "inactivo",
   "PhaseDraft": "borrador",
   "PhaseWorking": "trabajando",
-  "PhaseReasoning": "pensando",
+  "PhaseReasoning": "razonando",
   "PhaseReading": "leyendo",
   "PhaseUsingTool": "usando herramienta",
   "PhaseSubagents": "subagentes en curso",

crates/localization/locales/fr.json

@@ -25,7 +25,6 @@
   "TerminalDockLive": "actif",
   "ExtensionPromptUnavailable": "Les instructions des extensions sont indisponibles ; consultez /plugin show.",
   "ProfileConstitutionUnavailableLocal": "Les préférences du compte Codewhale n'ont pas pu être chargées ; ce tour utilise vos réglages locaux. Votre clé de fournisseur fonctionne toujours ; lancez codewhale login pour rétablir les préférences du compte.",
-  "AccountPreferencesNotLoaded": "Compte non chargé. Lancez {command}. Votre clé de fournisseur fonctionne toujours.",
   "ComposerOversizedSubmitHeld": "Non envoyé : ce message dépasse {limit} caractères et n’a pas pu être enregistré dans un fichier de texte collé ({error}). Le texte intégral reste dans la zone de saisie : raccourcissez-le ou autorisez l’écriture dans .codewhale/pastes, puis appuyez à nouveau sur Enter.",
   "ComposerOversizedSubmitHeldShort": "Non envoyé : plus de {limit} caractères ; fichier collé non enregistré.",
   "CmdPluginOwnerReport": "Hôte des extensions :\n  État : {state}\n  Outils actifs ({count}) : {tools}",
@@ -1668,7 +1667,7 @@
   "PhaseIdle": "inactif",
   "PhaseDraft": "brouillon",
   "PhaseWorking": "travail en cours",
-  "PhaseReasoning": "réflexion",
+  "PhaseReasoning": "raisonnement",
   "PhaseReading": "lecture",
   "PhaseUsingTool": "utilisation d'outil",
   "PhaseSubagents": "sous-agents en cours",

crates/localization/locales/hi.json

@@ -25,7 +25,6 @@
   "TerminalDockLive": "सक्रिय",
   "ExtensionPromptUnavailable": "एक्सटेंशन के निर्देश उपलब्ध नहीं हैं; /plugin show से जाँचें।",
   "ProfileConstitutionUnavailableLocal": "Codewhale खाते की प्राथमिकताएँ लोड नहीं हो सकीं, इसलिए यह टर्न आपकी स्थानीय सेटिंग्स का उपयोग करता है। आपकी प्रदाता कुंजी काम करती रहेगी; खाते की प्राथमिकताएँ वापस लाने के लिए codewhale login चलाएँ।",
-  "AccountPreferencesNotLoaded": "खाता लोड नहीं हुआ। {command} चलाएँ। आपकी प्रदाता कुंजी काम करती रहेगी।",
   "ComposerOversizedSubmitHeld": "नहीं भेजा गया: इस संदेश में {limit} से अधिक वर्ण हैं और इसे पेस्ट फ़ाइल के रूप में सहेजा नहीं जा सका ({error})। पूरा पाठ अभी भी इनपुट बॉक्स में है: इसे छोटा करें या .codewhale/pastes में लिखने की अनुमति दें, फिर Enter दबाएँ।",
   "ComposerOversizedSubmitHeldShort": "नहीं भेजा गया: {limit} से अधिक वर्ण; पेस्ट फ़ाइल सहेजी नहीं गई।",
   "CmdPluginOwnerReport": "एक्सटेंशन होस्ट:\n  स्थिति: {state}\n  सक्रिय टूल ({count}): {tools}",
@@ -1668,7 +1667,7 @@
   "PhaseIdle": "निष्क्रिय",
   "PhaseDraft": "मसौदा",
   "PhaseWorking": "काम चल रहा",
-  "PhaseReasoning": "सोच रहा",
+  "PhaseReasoning": "तर्क चल रहा",
   "PhaseReading": "पढ़ रहा",
   "PhaseUsingTool": "टूल इस्तेमाल हो रहा",
   "PhaseSubagents": "सबएजेंट चल रहे हैं",

crates/localization/locales/id.json

@@ -25,7 +25,6 @@
   "TerminalDockLive": "aktif",
   "ExtensionPromptUnavailable": "Instruksi ekstensi tidak tersedia; periksa /plugin show.",
   "ProfileConstitutionUnavailableLocal": "Preferensi akun Codewhale tidak dapat dimuat, jadi giliran ini memakai pengaturan lokal. Kunci penyedia tetap berfungsi; jalankan codewhale login untuk memulihkan preferensi akun.",
-  "AccountPreferencesNotLoaded": "Akun tidak dimuat. Jalankan {command}. Kunci penyedia tetap berfungsi.",
   "ComposerOversizedSubmitHeld": "Belum terkirim: pesan ini melebihi {limit} karakter dan tidak dapat disimpan sebagai berkas teks tempelan ({error}). Teks lengkap masih ada di kolom input: persingkat teks atau izinkan penulisan ke .codewhale/pastes, lalu tekan Enter lagi.",
   "ComposerOversizedSubmitHeldShort": "Belum terkirim: melebihi {limit} karakter; berkas tempelan tidak tersimpan.",
   "CmdPluginOwnerReport": "Host ekstensi:\n  Status: {state}\n  Alat aktif ({count}): {tools}",
@@ -1668,7 +1667,7 @@
   "PhaseIdle": "diam",
   "PhaseDraft": "draf",
   "PhaseWorking": "bekerja",
-  "PhaseReasoning": "berpikir",
+  "PhaseReasoning": "menalar",
   "PhaseReading": "membaca",
   "PhaseUsingTool": "memakai tool",
   "PhaseSubagents": "subagen berjalan",

crates/localization/locales/ja.json

@@ -25,7 +25,6 @@
   "TerminalDockLive": "稼働中",
   "ExtensionPromptUnavailable": "拡張機能の指示を利用できません。/plugin show で確認してください。",
   "ProfileConstitutionUnavailableLocal": "Codewhale アカウントの設定を読み込めなかったため、このターンはローカル設定を使います。プロバイダーキーはそのまま使えます。アカウント設定を戻すには codewhale login を実行してください。",
-  "AccountPreferencesNotLoaded": "アカウントを読み込めませんでした。{command} を実行してください。プロバイダーキーはそのまま使えます。",
   "ComposerOversizedSubmitHeld": "未送信：メッセージが {limit} 文字を超えており、貼り付けファイルに保存できませんでした（{error}）。全文は入力欄に残っています。短くするか .codewhale/pastes を書き込み可能にしてから、もう一度 Enter を押してください。",
   "ComposerOversizedSubmitHeldShort": "未送信：{limit} 文字超。貼り付けファイルは未保存です。",
   "CmdPluginOwnerReport": "拡張機能ホスト:\n  状態: {state}\n  有効なツール ({count}): {tools}",

crates/localization/locales/ko.json

@@ -25,7 +25,6 @@
   "TerminalDockLive": "실행 중",
   "ExtensionPromptUnavailable": "확장 기능의 지침을 사용할 수 없습니다. /plugin show에서 확인하세요.",
   "ProfileConstitutionUnavailableLocal": "Codewhale 계정 설정을 불러오지 못해 이번 턴은 로컬 설정을 사용합니다. 공급자 키는 계속 사용할 수 있으며, 계정 설정을 복원하려면 codewhale login을 실행하세요.",
-  "AccountPreferencesNotLoaded": "계정을 불러오지 못했습니다. {command}을 실행하세요. 공급자 키는 계속 사용할 수 있습니다.",
   "ComposerOversizedSubmitHeld": "전송하지 못했습니다. 메시지가 {limit}자를 초과하며 붙여넣기 파일로 저장할 수 없습니다({error}). 전체 텍스트는 입력란에 그대로 있습니다. 내용을 줄이거나 .codewhale/pastes에 쓰기 권한을 부여한 뒤 Enter를 다시 누르세요.",
   "ComposerOversizedSubmitHeldShort": "전송 안 됨: {limit}자 초과, 붙여넣기 파일 저장 실패.",
   "CmdPluginOwnerReport": "확장 호스트:\n  상태: {state}\n  활성 도구 ({count}): {tools}",
@@ -1691,7 +1690,7 @@
   "PhaseIdle": "유휴",
   "PhaseDraft": "초안",
   "PhaseWorking": "작업 중",
-  "PhaseReasoning": "생각 중",
+  "PhaseReasoning": "추론 중",
   "PhaseReading": "읽는 중",
   "PhaseUsingTool": "도구 사용 중",
   "PhaseSubagents": "서브에이전트 실행 중",

crates/localization/locales/pt-BR.json

@@ -25,7 +25,6 @@
   "TerminalDockLive": "ativo",
   "ExtensionPromptUnavailable": "As instruções das extensões não estão disponíveis; confira /plugin show.",
   "ProfileConstitutionUnavailableLocal": "Não foi possível carregar as preferências da conta Codewhale; este turno usa suas configurações locais. Sua chave do provedor continua funcionando; execute codewhale login para restaurar as preferências da conta.",
-  "AccountPreferencesNotLoaded": "Conta não carregada. Execute {command}. Sua chave do provedor continua funcionando.",
   "ComposerOversizedSubmitHeld": "Não enviado: esta mensagem excede {limit} caracteres e não pôde ser salva em um arquivo de texto colado ({error}). O texto completo continua no campo de entrada: reduza-o ou permita a gravação em .codewhale/pastes e pressione Enter novamente.",
   "ComposerOversizedSubmitHeldShort": "Não enviado: acima de {limit} caracteres; arquivo colado não salvo.",
   "CmdPluginOwnerReport": "Host de extensões:\n  Estado: {state}\n  Ferramentas ativas ({count}): {tools}",
@@ -1691,7 +1690,7 @@
   "PhaseIdle": "ocioso",
   "PhaseDraft": "rascunho",
   "PhaseWorking": "trabalhando",
-  "PhaseReasoning": "pensando",
+  "PhaseReasoning": "raciocinando",
   "PhaseReading": "lendo",
   "PhaseUsingTool": "usando ferramenta",
   "PhaseSubagents": "subagentes em andamento",

crates/localization/locales/ru.json

@@ -25,7 +25,6 @@
   "TerminalDockLive": "активен",
   "ExtensionPromptUnavailable": "Инструкции расширений недоступны; проверьте /plugin show.",
   "ProfileConstitutionUnavailableLocal": "Не удалось загрузить настройки аккаунта Codewhale, поэтому этот ход использует локальные настройки. Ключ провайдера продолжает работать; выполните codewhale login, чтобы вернуть настройки аккаунта.",
-  "AccountPreferencesNotLoaded": "Аккаунт не загружен. Выполните {command}. Ключ провайдера продолжает работать.",
   "ComposerOversizedSubmitHeld": "Не отправлено: сообщение длиннее {limit} символов, и его не удалось сохранить в файл вставленного текста ({error}). Полный текст остался в поле ввода: сократите его или разрешите запись в .codewhale/pastes, затем снова нажмите Enter.",
   "ComposerOversizedSubmitHeldShort": "Не отправлено: больше {limit} символов; файл вставки не сохранён.",
   "CmdPluginOwnerReport": "Хост расширений:\n  Состояние: {state}\n  Активные инструменты ({count}): {tools}",

crates/localization/locales/uk.json

@@ -25,7 +25,6 @@
   "TerminalDockLive": "активний",
   "ExtensionPromptUnavailable": "Інструкції розширень недоступні; перевірте /plugin show.",
   "ProfileConstitutionUnavailableLocal": "Не вдалося завантажити налаштування облікового запису Codewhale, тому цей хід використовує локальні налаштування. Ключ провайдера й далі працює; виконайте codewhale login, щоб повернути налаштування облікового запису.",
-  "AccountPreferencesNotLoaded": "Обліковий запис не завантажено. Виконайте {command}. Ключ провайдера й далі працює.",
   "ComposerOversizedSubmitHeld": "Не надіслано: повідомлення перевищує {limit} символів, і його не вдалося зберегти у файл вставленого тексту ({error}). Повний текст залишається в полі введення: скоротіть його або дозвольте запис у .codewhale/pastes, а потім знову натисніть Enter.",
   "ComposerOversizedSubmitHeldShort": "Не надіслано: понад {limit} символів; файл вставки не збережено.",
   "CmdPluginOwnerReport": "Хост розширень:\n  Стан: {state}\n  Активні інструменти ({count}): {tools}",

crates/localization/locales/vi.json

@@ -25,7 +25,6 @@
   "TerminalDockLive": "đang chạy",
   "ExtensionPromptUnavailable": "Không thể tải hướng dẫn của tiện ích mở rộng; kiểm tra /plugin show.",
   "ProfileConstitutionUnavailableLocal": "Không thể tải tùy chọn tài khoản Codewhale nên lượt này dùng cài đặt cục bộ. Khóa nhà cung cấp của bạn vẫn hoạt động; chạy codewhale login để khôi phục tùy chọn tài khoản.",
-  "AccountPreferencesNotLoaded": "Chưa tải được tài khoản. Chạy {command}. Khóa nhà cung cấp của bạn vẫn hoạt động.",
   "ComposerOversizedSubmitHeld": "Chưa gửi: tin nhắn vượt quá {limit} ký tự và không thể lưu thành tệp văn bản đã dán ({error}). Toàn bộ nội dung vẫn còn trong ô nhập: hãy rút ngắn hoặc cho phép ghi vào .codewhale/pastes, rồi nhấn Enter lần nữa.",
   "ComposerOversizedSubmitHeldShort": "Chưa gửi: vượt quá {limit} ký tự; chưa lưu tệp dán.",
   "CmdPluginOwnerReport": "Máy chủ tiện ích mở rộng:\n  Trạng thái: {state}\n  Công cụ đang hoạt động ({count}): {tools}",
@@ -1691,7 +1690,7 @@
   "PhaseIdle": "rảnh",
   "PhaseDraft": "nháp",
   "PhaseWorking": "đang xử lý",
-  "PhaseReasoning": "đang suy nghĩ",
+  "PhaseReasoning": "đang suy luận",
   "PhaseReading": "đang đọc",
   "PhaseUsingTool": "đang dùng công cụ",
   "PhaseSubagents": "đang chạy tác tử con",

crates/localization/locales/zh-Hans.json

@@ -25,7 +25,6 @@
   "TerminalDockLive": "运行中",
   "ExtensionPromptUnavailable": "扩展指令不可用；请通过 /plugin show 查看。",
   "ProfileConstitutionUnavailableLocal": "无法加载 Codewhale 账号偏好，本轮改用本地设置。你的提供商密钥仍可正常使用；运行 codewhale login 可恢复账号偏好。",
-  "AccountPreferencesNotLoaded": "账号未加载。请运行 {command}。你的提供商密钥仍可正常使用。",
   "ComposerOversizedSubmitHeld": "未发送：消息超过 {limit} 个字符，无法保存为粘贴文本文件（{error}）。完整文本仍保留在输入框中：请缩短内容，或允许写入 .codewhale/pastes，然后再次按 Enter。",
   "ComposerOversizedSubmitHeldShort": "未发送：超过 {limit} 个字符；粘贴文件未保存。",
   "CmdPluginOwnerReport": "扩展宿主：\n  状态：{state}\n  活动工具（{count}）：{tools}",
@@ -1691,7 +1690,7 @@
   "PhaseIdle": "空闲",
   "PhaseDraft": "草稿",
   "PhaseWorking": "工作中",
-  "PhaseReasoning": "思考中",
+  "PhaseReasoning": "推理中",
   "PhaseReading": "读取中",
   "PhaseUsingTool": "使用工具中",
   "PhaseSubagents": "智能体工作中",

crates/localization/locales/zh-Hant.json

@@ -25,7 +25,6 @@
   "TerminalDockLive": "執行中",
   "ExtensionPromptUnavailable": "擴充功能指令無法使用；請透過 /plugin show 查看。",
   "ProfileConstitutionUnavailableLocal": "無法載入 Codewhale 帳號偏好，本輪改用本機設定。你的服務商金鑰仍可正常使用；執行 codewhale login 可恢復帳號偏好。",
-  "AccountPreferencesNotLoaded": "帳號未載入。請執行 {command}。你的服務商金鑰仍可正常使用。",
   "ComposerOversizedSubmitHeld": "未傳送：訊息超過 {limit} 個字元，無法儲存為貼上文字檔案（{error}）。完整文字仍保留在輸入框中：請縮短內容，或允許寫入 .codewhale/pastes，然後再次按 Enter。",
   "ComposerOversizedSubmitHeldShort": "未傳送：超過 {limit} 個字元；貼上檔案未儲存。",
   "CmdPluginOwnerReport": "擴充功能主機：\n  狀態：{state}\n  活動工具（{count}）：{tools}",
@@ -1689,7 +1688,7 @@
   "PhaseIdle": "閒置",
   "PhaseDraft": "草稿",
   "PhaseWorking": "工作中",
-  "PhaseReasoning": "思考中",
+  "PhaseReasoning": "推理中",
   "PhaseReading": "讀取中",
   "PhaseUsingTool": "使用工具中",
   "PhaseSubagents": "代理程式工作中",

crates/localization/src/lib.rs

@@ -714,7 +714,6 @@ pub enum MessageId {
     ExtensionsInventorySkills,
     ExtensionPromptUnavailable,
     ProfileConstitutionUnavailableLocal,
-    AccountPreferencesNotLoaded,
     ExtensionsMarketplaceDetail,
     ExtensionsMarketplaceUnavailable,
     ExtensionsMcpEmpty,
@@ -3652,7 +3651,6 @@ pub const ALL_MESSAGE_IDS: &[MessageId] = &[
     MessageId::ExtensionsInventorySkills,
     MessageId::ExtensionPromptUnavailable,
     MessageId::ProfileConstitutionUnavailableLocal,
-    MessageId::AccountPreferencesNotLoaded,
     MessageId::ExtensionsMarketplaceDetail,
     MessageId::ExtensionsMarketplaceUnavailable,
     MessageId::ExtensionsMcpEmpty,

crates/runtime/src/settings.rs

@@ -492,8 +492,6 @@ pub struct Settings {
     /// One-time Operate-retirement toast has been shown. Same suppression
     /// contract as `yolo_deprecation_shown`.
     pub operate_deprecation_shown: bool,
-    #[serde(default, skip_serializing_if = "Option::is_none")]
-    pub account_preferences_notice_shown: Option<String>,
     /// Round 3 (2026-09-01) moved the work bar under the composer. Every
     /// settings.toml saved before that carries `work_surface_placement =
     /// "top"` — the old default, persisted verbatim by ordinary saves, not a
@@ -652,7 +650,9 @@ impl Default for Settings {
             mention_menu_limit: 128,
             mention_walk_depth: 10,
             mention_menu_behavior: "fuzzy".to_string(),
-            show_thinking: true,
+            // Reasoning is useful when explicitly requested, but it should
+            // never displace the actual conversation in the default TUI.
+            show_thinking: false,
             thinking_default_expanded: false,
             thinking_preview_lines: default_thinking_preview_lines(),
             thinking_highlight: true,
@@ -694,7 +694,6 @@ impl Default for Settings {
             feature_intro_shown: false,
             yolo_deprecation_shown: false,
             operate_deprecation_shown: false,
-            account_preferences_notice_shown: None,
             legacy_operate_default: false,
             work_surface_bottom_migrated: false,
             behavioral_tip_impressions: std::collections::BTreeMap::new(),
@@ -3748,36 +3747,6 @@ mod tests {
         assert_eq!(reloaded.permission_posture.as_deref(), Some("ask"));
     }
 
-    #[test]
-    fn thinking_shows_by_default_and_a_saved_choice_survives_reload_and_other_saves() {
-        let _g = config_path_test_guard();
-        let tmp = tempfile::tempdir().expect("tempdir");
-        let settings_path = tmp.path().join("settings.toml");
-        let _config_override =
-            EnvVarRestore::set("DEEPSEEK_CONFIG_PATH", tmp.path().join("config.toml"));
-
-        let absent = Settings::load().expect("load without a settings file");
-        assert!(absent.show_thinking);
-        assert!(!absent.thinking_default_expanded);
-        assert!(!absent.is_set("show_thinking"));
-
-        std::fs::write(&settings_path, "theme = \"light\"\n").expect("settings without the key");
-        let unnamed = Settings::load().expect("load settings without the key");
-        assert!(unnamed.show_thinking);
-        assert!(!unnamed.is_set("show_thinking"));
-
-        for saved in [false, true] {
-            std::fs::write(&settings_path, format!("show_thinking = {saved}\n"))
-                .expect("saved choice");
-            let loaded = Settings::load().expect("load the saved choice");
-            assert_eq!(loaded.show_thinking, saved);
-            assert!(loaded.is_set("show_thinking"));
-            Settings::transact(|settings| settings.set("locale", "en")).expect("another save");
-            let reloaded = Settings::load().expect("reload after another save");
-            assert_eq!(reloaded.show_thinking, saved);
-        }
-    }
-
     #[test]
     fn settings_load_keeps_top_placement_chosen_after_the_bottom_migration() {
         let _g = config_path_test_guard();
@@ -3917,7 +3886,7 @@ mod tests {
         let mut settings = Settings::default();
         // Density is calm by default; motion is an independent axis.
         assert!(settings.calm_mode);
-        settings.show_thinking = false;
+        assert!(!settings.show_thinking);
 
         let changed = settings.apply_preset("CALM").expect("calm preset applies");
         assert_eq!(
@@ -3947,8 +3916,8 @@ mod tests {
         assert!(settings.fancy_animations);
         assert_eq!(settings.transcript_spacing, "spacious");
         assert_eq!(settings.tool_collapse_mode, "compact");
-        assert!(settings.show_thinking);
-        assert!(!settings.thinking_default_expanded);
+        // Thinking is opt-in so the transcript stays focused on the chat.
+        assert!(!settings.show_thinking);
     }
 
     #[test]

crates/runtime/src/working_set.rs

@@ -1371,6 +1371,9 @@ fn relativize_candidate(
         .map(fs::Metadata::is_dir)
         .unwrap_or_else(|| candidate.ends_with('/'));
 
+    if rel_path.as_os_str().is_empty() {
+        return None;
+    }
     let rel_string = path_to_string(&rel_path)?;
     Some((rel_string, exists, is_dir))
 }
@@ -1655,6 +1658,19 @@ mod tests {
         assert!(!entry.is_dir);
     }
 
+    #[test]
+    fn workspace_root_is_not_an_active_path() {
+        let tmp = TempDir::new().expect("tempdir");
+        let root = tmp.path().canonicalize().expect("canonical root");
+        fs::write(root.join("笔记 📝.md"), "旧").expect("write");
+
+        let mut ws = WorkingSet::default();
+        let mention = format!("<file path=\"{}/笔记 📝.md\">", root.display());
+        ws.observe_user_message(&mention, &root);
+
+        assert!(!ws.entries.contains_key(""), "{:?}", ws.entries.keys());
+    }
+
     #[test]
     fn observe_tool_call_extracts_paths_from_input() {
         let tmp = TempDir::new().expect("tempdir");

crates/snapshot/src/repo.rs

@@ -1457,6 +1457,7 @@ impl SnapshotRepo {
                 .arg(&self.git_dir)
                 .arg("--work-tree")
                 .arg(&self.work_tree)
+                .args(["-c", "core.quotePath=false"])
                 .args(args);
             let output = run_bounded_git(&mut cmd, subcommand)?;
             if !output.status.success() {
@@ -1850,7 +1851,7 @@ impl SnapshotRepo {
             .duration_since(UNIX_EPOCH)
             .map_err(|e| io_other(format!("clock error: {e}")))?
             .as_secs() as i64;
-        let cutoff = now - max_age.as_secs() as i64;
+        let cutoff = now.saturating_sub(i64::try_from(max_age.as_secs()).unwrap_or(i64::MAX));
 
         let snapshots = self.list(usize::MAX)?;
         if snapshots.is_empty() {
@@ -3948,6 +3949,22 @@ mod tests {
         assert_eq!(list[0].label, "turn:1");
     }
 
+    #[test]
+    fn prune_with_an_age_beyond_i64_keeps_every_snapshot() {
+        let tmp = tempdir().unwrap();
+        let (repo, _home) = make_repo(tmp.path());
+        std::fs::write(repo.work_tree().join("f.txt"), "v0").unwrap();
+        repo.snapshot("turn:0").unwrap();
+        std::thread::sleep(Duration::from_millis(1100));
+
+        let removed = repo
+            .prune_older_than(Duration::from_secs(u64::MAX))
+            .unwrap();
+
+        assert_eq!(removed, 0);
+        assert_eq!(repo.list(10).unwrap().len(), 1);
+    }
+
     /// The 2026-08-04 regression: with a cut in the MIDDLE of history,
     /// `prune_older_than` used to `update-ref HEAD <oldest survivor>`, which
     /// orphaned (and gc destroyed) the NEWEST snapshots while keeping the
@@ -4892,6 +4909,24 @@ mod tests {
         assert_eq!(none.patch, "");
     }
 
+    #[test]
+    fn work_tree_changes_since_names_non_ascii_paths_as_written() {
+        let tmp = tempdir().unwrap();
+        let (repo, _home) = make_repo(tmp.path());
+        let ws = repo.work_tree().to_path_buf();
+        std::fs::write(ws.join("笔记 📝.md"), b"old\n").unwrap();
+        let before = repo.snapshot("pre-turn:1").expect("snapshot");
+        git_output(&repo, &["config", "core.quotePath", "true"]);
+        std::fs::write(ws.join("笔记 📝.md"), b"new\n").unwrap();
+
+        let changes = repo
+            .work_tree_changes_since(&before, 1 << 20)
+            .expect("changes");
+        assert_eq!(changes.names, "笔记 📝.md\n");
+        assert!(changes.stat.contains("笔记 📝.md"), "{}", changes.stat);
+        assert!(changes.patch.contains("b/笔记 📝.md"), "{}", changes.patch);
+    }
+
     /// A tree the repo no longer holds — the receipt outlived the object —
     /// is reported as absent rather than as a diff failure, so a caller can
     /// tell pruning from a broken repo without reading git's stderr.

crates/state/src/lib.rs

@@ -11,7 +11,7 @@ mod runtime_aliases;
 
 use std::collections::HashMap;
 use std::fs::{self, OpenOptions};
-use std::io::{BufRead, BufReader, Write};
+use std::io::{BufRead, BufReader, Read, Seek, SeekFrom, Write};
 use std::path::{Path, PathBuf};
 use std::sync::{Arc, LazyLock, Mutex, MutexGuard};
 
@@ -1453,6 +1453,7 @@ impl StateStore {
         self.with_session_index_lock(|| {
             let mut file = OpenOptions::new()
                 .create(true)
+                .read(true)
                 .append(true)
                 .open(&self.session_index_path)
                 .with_context(|| {
@@ -1461,7 +1462,14 @@ impl StateStore {
                         self.session_index_path.display()
                     )
                 })?;
-            writeln!(file, "{encoded}").context("failed to append session index entry")?;
+            let mut last = *b"\n";
+            if file.metadata()?.len() > 0 {
+                file.seek(SeekFrom::End(-1))?;
+                file.read_exact(&mut last)?;
+            }
+            let separator = if last[0] == b'\n' { "" } else { "\n" };
+            writeln!(file, "{separator}{encoded}")
+                .context("failed to append session index entry")?;
             // Durability: without this a crash mid-write can leave a torn
             // final line. Reads tolerate one (see `session_index_map`), but
             // not losing the entry beats recovering from having lost it.
@@ -2922,6 +2930,33 @@ mod tests {
         );
     }
 
+    #[test]
+    fn session_index_append_after_an_unterminated_torn_line_is_kept() {
+        let store = temp_state_store("session-index-unterminated");
+        store
+            .append_thread_name("thread-1", Some("first".to_string()), 1, None)
+            .expect("append first entry");
+        {
+            let mut file = OpenOptions::new()
+                .append(true)
+                .open(&store.session_index_path)
+                .expect("open session index");
+            write!(file, "{{\"thread_id\":\"thread-2\",\"thread_na").expect("write torn line");
+        }
+
+        store
+            .append_thread_name("thread-3", Some("third".to_string()), 3, None)
+            .expect("append third entry");
+
+        assert_eq!(
+            store
+                .find_thread_name_by_id("thread-3")
+                .expect("lookup thread-3")
+                .as_deref(),
+            Some("third"),
+        );
+    }
+
     /// Hook fired by compaction between writing the snapshot and renaming it
     /// over the live index — the window a concurrent append can be lost in.
     /// Only the store whose path a test registered is affected, so tests

crates/tui/CHANGELOG.md

@@ -7,6 +7,11 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 ## [Unreleased]
 
+- **Renaming sessions with Chinese titles:** the session picker no longer rejects a Chinese, Japanese, Korean or emoji title of more than about 33 characters as "Title must be 1–100 characters". It now counts characters, as the message says and as `/rename` and the Runtime API already did, so any title up to 100 characters is accepted. Thanks [@luochen211](https://github.com/luochen211) (PR 7322).
+- **Non-ASCII file names:** `/diff`, the `@git` and `@diff` mentions and the Runtime API workspace diff now show paths with Chinese, emoji and other non-ASCII characters as written instead of as quoted octal escapes, and the turn context no longer lists the workspace root as an empty "Active paths" entry. Thanks [@luochen211](https://github.com/luochen211) (PR 7317).
+- **Session names after a crash, and keep-forever snapshot retention:** a session name saved right after a crash that cut off the previous session-index line is no longer lost, because the new entry now starts on its own line; and a very large `[snapshots] max_age_days` (such as `9223372036854775807` for "keep forever") no longer wraps around and deletes every undo snapshot at session start. Thanks [@luochen211](https://github.com/luochen211) (PR 7316).
+- **Blank environment variables:** exporting a variable empty (for example `CODEWHALE_MODEL=` from a compose `${VAR:-}` template) no longer hides its legacy alias. `DEEPSEEK_MODEL`, `DEEPSEEK_BASE_URL`, `KIMI_BASE_URL`, `DEEPSEEK_FORCE_HTTP1`, `DEEPSEEK_STREAM_OPEN_TIMEOUT_SECS` and the other provider alias variables now apply as they do when the primary is unset. Thanks [@luochen211](https://github.com/luochen211) (PR 7315).
+- **Qwen Code as an external agent:** Codewhale now finds an installed Qwen Code CLI (`qwen`, 0.6.1 or newer) and can run it as a session over ACP with `qwen --acp`. `GET /v1/agents/external` lists it, an older install is reported as needing `npm i -g @qwen-code/qwen-code@latest`, and sign-in stays inside Qwen Code (`/auth`). Thanks [@MannXo](https://github.com/MannXo) (#7293).
 - **README and contributor credit:** the repository front page now opens the way the website does: the line "The model can be anyone's. The agent should be yours.", one sentence on what Codewhale is, the install command and one picture, with reference detail moved into the linked guides; the 18 translated READMEs follow the same structure and say they are machine-drafted. The contributor wall and the website now count the same 433 people (407 GitHub accounts and the commit authors whose email is not linked to an account), where the website had counted accounts only. `DEPENDENCY_SECURITY.md` moved into `docs/dependency-maintenance.md`, and `.github/SUPPORT.md` says where to ask for help.
 - **Website Plugins page:** The website's Plugins page now describes what the released version does, and a labelled section says what 0.10.3 adds: one Plugins place, Standard and Creator mode, and the Permissions control.
 - **Plugin docs and plugin-creator:** Docs now say everything you add is a plugin. The bundled plugin-creator skill no longer tells an agent never to trust or enable a plugin (in Full Access it may, for one it wrote and installed from its project), points code and screen plugins at cordis-plugin-development, and an untouched copy from 0.10.2 now upgrades. The Chinese extension and Runtime API docs now say editor ACP sessions get plugin tools and configured MCP servers.
@@ -669,7 +674,7 @@ work is declared complete here. Items moved to 0.10.4 by name are marked below.
 - **Windows uninstall removes the receipt and update leftovers.** The installer's uninstaller also deletes `.codewhale-install.sha256` and `codewhale.old-*`/`codew.old-*` left by a self-update, so `bin` can be removed. Refs #7091.
 - **Windows `/undo` matches recorded file paths regardless of case.** A write recorded under a different drive-letter or directory case, or a `\\?\` prefix, is no longer dropped from the restore plan. Refs #6970.
 - **Windows: Fleet and metrics read live files again.** Replaying a Fleet ledger and reading audit or receipt records no longer fail with a sharing violation while their writer is still open; the same regular-file, link and identity checks apply. Plugin enablement now says which snapshot check refused (stage root, content identity or final check) instead of one generic message. Generated protocol and SDK exports keep LF line endings in Windows checkouts.
-- **Sessions in one folder no longer leave each other's sub-agents read-only or private:** a second session works beside another one, stale or suspended included, the older sub-agent store upgrades right away, and a store that cannot be opened is retried on the next sub-agent change instead of needing a restart. While another Codewhale session, a suspended one included, is open in a folder the desktop app or `serve` was asked to work in (other than the folder it started in), that Runtime still refuses sub-agents that would edit the folder directly; isolated worktrees are unaffected. A sub-agent store that cannot be opened shows its refusal once instead of repeating cleanup warnings on every sidebar refresh. Restart any session started with Codewhale 0.10.2 or older: once the folder upgrades it can no longer save sub-agent state (Refs #7232).
+- **A held coordination lock no longer makes a new session read-only.** When another session owns the delegated-agent store, the new session starts on private state and says so, instead of refusing every sub-agent with “Failed to clean up sub-agents”.
 - **A dev server can outlive the turn in an interactive session.** `persist:true` now works in the TUI and desktop when the thread has Full Access; restricted postures and Fleet/sub-agents still refuse it.
 - **`codewhale exec --add-dir <path>`** (repeatable) and `addDirs` on `POST /v1/threads` let the file tools read and write a sibling directory, a second checkout or a shared scratch root without turning on Full Access.
 - **Unsettled usage is persisted in production, not only in tests.** The interactive sink now records a legacy usage receipt when a provider response has no settlement, so children and turns stop reporting settlement failure.
@@ -706,6 +711,12 @@ work is declared complete here. Items moved to 0.10.4 by name are marked below.
   composer. Real Pinyin input on macOS and Windows is not yet confirmed.
   Thanks [@cmdcorp6534](https://github.com/cmdcorp6534) for the report
   (Refs #2323).
+- **Sub-agents beside an older session:** while a Codewhale 0.10.2 session
+  still uses a folder's older delegated-agent store, a 0.10.3 session in the
+  same folder keeps working sub-agents with private state that is not saved,
+  instead of refusing every sub-agent action. The older session's files are
+  left untouched, and the store upgrades when a session starts after the older
+  one exits (Refs #7232).
 - **Fleet members:** a replayed member start now matches only within its own
   session, so two sessions that reuse an operation key no longer receive each
   other's agent.
@@ -734,6 +745,12 @@ work is declared complete here. Items moved to 0.10.4 by name are marked below.
   bare self-package exports resolve inside the selected package directory.
   Invalid or incompatible ranges and paths outside the reviewed package still
   refuse. Real published-plugin activation remains unqualified.
+- **Coordination lock diagnostics (#7232):** stale lock-file contents no longer
+  identify an owner PID. A refused legacy admission keeps its reason in Work
+  and shows one notice per cause instead of repeating cleanup warnings on
+  every sidebar refresh. Legacy migration still requires the actual holder to
+  exit and the refused owning Runtime to restart; locks and retained state
+  remain protected.
 - **A local proxy that stops mid-session no longer strands the session (#7231):** the model, catalog, search, vision, finance, MCP-registry, profile and remote-control clients watch a loopback proxy port and go direct while it refuses connections, then resume proxying when it returns. Only a definite refusal from every resolved address changes the route; remote proxies, `NO_PROXY` matches, credentials and the existing precedence are unchanged. Set `CODEWHALE_LOCAL_PROXY_FALLBACK=0` to keep the old build-time snapshot.
 - **Proxy diagnostic privacy ([#7233](https://github.com/codewhale-hq/codewhale/issues/7233)):** malformed MCP proxy warnings omit configured values; local-proxy log labels mask userinfo and omit URL paths, queries and fragments. Routing is unchanged. All 23 existing proxy tests passed on macOS; runtime log capture and other platforms remain unqualified. Thanks [@SparkofSpike](https://github.com/SparkofSpike) for the report.
 - **Fork, undo and retry work after a workspace drift notice (#7234):** the history-boundary walks now treat a whole `<context_update>` envelope as engine-owned, so a session that received one can be cut at every saved turn. A person quoting a notice or a truncated envelope still counts as their own prompt.

crates/tui/src/commands/contract/debug_operations.rs

@@ -157,7 +157,8 @@ fn session_diff(app: &App) -> std::io::Result<Option<DebugDiffObservation>> {
 /// The workspace repository's own uncommitted changes: what `/diff` shows
 /// before the session has a restore point to compare against.
 fn workspace_git_diff(workspace: &std::path::Path) -> DebugDiffObservation {
-    let git = |args: &[&str]| Git::output(args, workspace);
+    let git =
+        |args: &[&str]| Git::output(&[&["-c", "core.quotePath=false"], args].concat(), workspace);
     // Outside a repository `git diff` prints nothing on stdout, which would
     // read as "no changes".
     match git(&["rev-parse", "--is-inside-work-tree"]) {

crates/tui/src/tui/app.rs

@@ -87,27 +87,6 @@ pub(crate) use types::{
     WORKFLOW_DRAFT_INSTRUCTION_PREFIX,
 };
 
-const ACCOUNT_PREFERENCES_NOTICE: &str = "account-preferences-notice";
-
-fn account_sign_in_marker(profile: Option<&str>) -> String {
-    use sha2::{Digest, Sha256};
-    let api_base = codewhale_engine::profile_constitution::runtime_account_api_base();
-    let session = codewhale_secrets::account::AccountSessionStore::new(
-        codewhale_secrets::Secrets::file_backed_read_only(),
-        profile,
-        &api_base,
-    )
-    .runtime_info_at(Utc::now())
-    .ok()
-    .and_then(|info| info.session_id)
-    .unwrap_or_default();
-    Sha256::digest(session.as_bytes())
-        .iter()
-        .take(8)
-        .map(|byte| format!("{byte:02x}"))
-        .collect()
-}
-
 // === Types ===
 
 /// One login owns one mailbox. A cancelled task can only write its abandoned
@@ -2469,7 +2448,6 @@ pub struct App {
     /// re-render per character. `None` means "no bump since the last
     /// finalize" so the first chunk of a block always renders immediately.
     pub thinking_revision_last_bump_at: Option<Instant>,
-    pub(crate) thinking_revision_owed: bool,
     /// Newline-gated streaming collector state.
     pub streaming_state: StreamingState,
     /// Live approximate output tokens for the current assistant stream.
@@ -3813,38 +3791,6 @@ impl App {
         );
     }
 
-    pub(crate) fn notify_account_preferences_not_loaded(&mut self) {
-        let sign_in = account_sign_in_marker(self.config_profile.as_deref());
-        self.notify_account_preferences_not_loaded_for(&sign_in);
-    }
-
-    fn notify_account_preferences_not_loaded_for(&mut self, sign_in: &str) {
-        let first_for_sign_in = Settings::transact_opt(|settings| {
-            if settings.account_preferences_notice_shown.as_deref() == Some(sign_in) {
-                return Ok(None);
-            }
-            settings.account_preferences_notice_shown = Some(sign_in.to_string());
-            Ok(Some(()))
-        });
-        if matches!(first_for_sign_in, Ok(None)) {
-            return;
-        }
-        let text = self
-            .tr(MessageId::AccountPreferencesNotLoaded)
-            .replace("{command}", "codewhale login");
-        self.push_status_toast_record(StatusToast::standing(
-            text,
-            StatusToastLevel::Warning,
-            ACCOUNT_PREFERENCES_NOTICE,
-        ));
-    }
-
-    pub(crate) fn dismiss_account_preferences_notice(&mut self) -> bool {
-        let before = self.status_toasts.len();
-        self.retire_event_notices(ACCOUNT_PREFERENCES_NOTICE);
-        self.status_toasts.len() != before
-    }
-
     /// One-release migration notice for the Shift+Tab/Ctrl+T rebinding: users
     /// pressing Shift+Tab expecting the old thinking cycle land here first.
     fn notify_keybinding_migration_once(&mut self) {

crates/tui/src/tui/app/init.rs

@@ -1142,7 +1142,6 @@ impl App {
             streaming_thinking_active_entry: None,
             pending_thinking_translation_cells: HashMap::new(),
             thinking_revision_last_bump_at: None,
-            thinking_revision_owed: false,
             streaming_state: StreamingState::new(),
             streaming_output_token_estimate: 0,
             reasoning_buffer: String::new(),

crates/tui/src/tui/app/tests.rs

@@ -184,46 +184,6 @@ fn feature_intro_shows_once_persists_then_is_idempotent() {
     let _ = std::fs::remove_dir_all(&tmp);
 }
 
-#[test]
-fn account_preferences_notice_shows_once_per_sign_in_and_can_be_dismissed() {
-    let _env_lock = lock_test_env();
-    let tmp = tempfile::tempdir().unwrap();
-    let config_path = tmp.path().join("config.toml");
-    let _env = EnvVarGuard::set(
-        "DEEPSEEK_CONFIG_PATH",
-        config_path.to_string_lossy().as_ref(),
-    );
-    let shown = |app: &App| {
-        let notice = app
-            .tr(MessageId::AccountPreferencesNotLoaded)
-            .replace("{command}", "codewhale login");
-        app.status_toasts
-            .iter()
-            .filter(|toast| toast.text == notice && toast.kind == StatusToastKind::Standing)
-            .count()
-    };
-
-    let mut app = App::new(test_options(false), &Config::default());
-    app.notify_account_preferences_not_loaded_for("first-sign-in");
-    app.notify_account_preferences_not_loaded_for("first-sign-in");
-    assert_eq!(shown(&app), 1);
-    assert_eq!(
-        Settings::load()
-            .expect("settings should load")
-            .account_preferences_notice_shown
-            .as_deref(),
-        Some("first-sign-in")
-    );
-    assert!(app.dismiss_account_preferences_notice());
-    assert_eq!(shown(&app), 0);
-
-    let mut later_session = App::new(test_options(false), &Config::default());
-    later_session.notify_account_preferences_not_loaded_for("first-sign-in");
-    assert_eq!(shown(&later_session), 0, "the same sign-in is told once");
-    later_session.notify_account_preferences_not_loaded_for("second-sign-in");
-    assert_eq!(shown(&later_session), 1, "a later sign-in is told again");
-}
-
 #[test]
 fn initial_input_scenario() {
     // Scenario consolidation of: initial_input_prefill_waits_for_manual_submit, initial_input_submit_marks_startup_dispatch

crates/tui/src/tui/context_menu.rs

@@ -284,8 +284,8 @@ impl ContextMenuView {
         if start > 0 {
             rows.push(MenuRow::MoreAbove(start));
         }
-        for idx in start..end {
-            if idx > start && self.entries[idx].section_start {
+        for (idx, entry) in self.entries.iter().enumerate().take(end).skip(start) {
+            if idx > start && entry.section_start {
                 rows.push(MenuRow::Divider);
             }
             rows.push(MenuRow::Entry(idx));

crates/tui/src/tui/git_mention.rs

@@ -403,6 +403,25 @@ mod tests {
         assert!(!payload.truncated);
     }
 
+    #[test]
+    fn mentions_name_non_ascii_paths_as_written() {
+        let dir = tempfile::tempdir().unwrap();
+        init_repo(dir.path());
+        Command::new("git")
+            .args(["config", "core.quotePath", "true"])
+            .current_dir(dir.path())
+            .output()
+            .unwrap();
+        std::fs::write(dir.path().join("笔记 📝.md"), "旧\n").unwrap();
+        commit_all(dir.path(), "initial");
+        std::fs::write(dir.path().join("笔记 📝.md"), "新\n").unwrap();
+
+        let status = resolve_git_mention(GitMentionKind::Status, dir.path());
+        let diff = resolve_git_mention(GitMentionKind::Diff, dir.path());
+        assert!(status.block.contains("笔记 📝.md"), "{}", status.block);
+        assert!(diff.block.contains("b/笔记 📝.md"), "{}", diff.block);
+    }
+
     #[test]
     fn diff_covers_staged_and_unstaged_changes() {
         let dir = tempfile::tempdir().unwrap();

crates/tui/src/tui/goldens/config_panel_120x32.txt

@@ -9,7 +9,7 @@
                     │ Background                         (default)      ✎     │current   terminal                       
                     │ Pet mode                           Off            [ ]   │saved     terminal                       
                     │ Quiet transcript                   On             [x]   │startup   terminal                       
-                    │ Model reasoning in chat            On             [x]   │source    settings.toml                  
+                    │ Model reasoning in chat            Off            [ ]   │source    settings.toml                  
                     │ Thinking Default Expanded          Off            [ ]   │scope     SAVED                          
                     │ Thinking Preview Lines             2              ✎     │apply     applies on save                
                     │ Thinking background highlight      On             [x]   │kind      choice                         

crates/tui/src/tui/goldens/config_panel_80x24.txt

@@ -9,7 +9,7 @@
    Background                         (default)                 ✎    SAVED   █  
    Pet mode                           Off                       [ ]  SAVED   █  
    Quiet transcript                   On                        [x]  SAVED   █  
-   Model reasoning in chat            On                        [x]  SAVED   █  
+   Model reasoning in chat            Off                       [ ]  SAVED   █  
    Thinking Default Expanded          Off                       [ ]  SAVED   █  
    Thinking Preview Lines             2                         ✎    SAVED   │  
    Thinking background highlight      On                        [x]  SAVED   │  

crates/tui/src/tui/session_picker.rs

@@ -22,8 +22,8 @@ use crate::tui::views::{
 };
 use crate::tui::views::{ModalKind, ModalView, ViewAction, ViewEvent};
 use codewhale_engine::session_manager::{
-    SavedSession, SessionListFilter, SessionManager, SessionMetadata, extract_title,
-    extract_user_prompt, strip_thinking_tags,
+    MAX_SESSION_TITLE_CHARS, SavedSession, SessionListFilter, SessionManager, SessionMetadata,
+    extract_title, extract_user_prompt, strip_thinking_tags,
 };
 use codewhale_engine::session_projection::{MAX_PROJECTED_SESSIONS, SessionQuery, SessionSortMode};
 use codewhale_localization::{Locale, MessageId, tr};
@@ -787,7 +787,7 @@ impl SessionPickerView {
             self.status = Some(tr(self.locale, MessageId::SessionsNoSelection).into_owned());
             return ViewAction::None;
         };
-        if new_title.is_empty() || new_title.len() > 100 {
+        if new_title.is_empty() || new_title.chars().count() > MAX_SESSION_TITLE_CHARS {
             self.status = Some(tr(self.locale, MessageId::SessionsTitleLength).into_owned());
             return ViewAction::None;
         }
@@ -2045,6 +2045,28 @@ mod tests {
         );
     }
 
+    #[test]
+    fn rename_selected_limits_title_by_characters() {
+        let _lock = crate::test_support::lock_test_env();
+        let tmp = tempfile::tempdir().expect("tempdir");
+        let _home = crate::test_support::EnvVarGuard::set("CODEWHALE_HOME", tmp.path());
+        let manager = SessionManager::default_location().expect("session manager");
+        let mut saved = saved_session_with_messages(vec![text_message("user", "hello")]);
+        saved.metadata.id = "session-cjk".to_string();
+        manager.save_session(&saved).expect("save session");
+        let mut view = picker_with(vec![saved.metadata.clone()], None);
+
+        let title = "中文会话标题".repeat(6);
+        assert!(matches!(
+            view.rename_selected(&title),
+            ViewAction::Emit(ViewEvent::SessionRenamed { .. })
+        ));
+        assert_eq!(
+            manager.load_session("session-cjk").unwrap().metadata.title,
+            title
+        );
+    }
+
     #[test]
     fn archive_toggle_persists_hides_the_row_and_emits_an_archive_receipt() {
         let _lock = crate::test_support::lock_test_env();

crates/tui/src/tui/streaming_thinking.rs

@@ -30,17 +30,15 @@ use crate::tui::history::HistoryCell;
 /// frame cap (see `frame_rate_limiter`) yet imperceptible as lag.
 ///
 /// Correctness: this only skips *intermediate* repaints. Appended content is
-/// never dropped — it lands in the cell immediately — the display beat renders
-/// a skipped bump once the window has passed, and finalize always forces a
-/// bump so the final reasoning text is fully rendered.
+/// never dropped — it lands in the cell immediately — and finalize always
+/// forces a bump so the final reasoning text is fully rendered.
 const THINKING_REVISION_THROTTLE: Duration = Duration::from_millis(100);
 
 /// Bump the active-cell revision for a streaming thinking mutation, but at
 /// most once per [`THINKING_REVISION_THROTTLE`] window. Returns whether a bump
-/// was actually emitted. Skipped bumps coalesce into the next one, into
-/// [`render_owed_revision`] on a later display beat, or into the forced
-/// finalize bump, so no content is ever lost — only redundant intermediate
-/// re-wraps are dropped.
+/// was actually emitted. Skipped bumps coalesce into the next one (or into the
+/// forced finalize bump), so no content is ever lost — only redundant
+/// intermediate re-wraps are dropped.
 fn bump_thinking_revision_throttled(app: &mut App, now: Instant) -> bool {
     let due = app
         .thinking_revision_last_bump_at
@@ -49,17 +47,9 @@ fn bump_thinking_revision_throttled(app: &mut App, now: Instant) -> bool {
         app.thinking_revision_last_bump_at = Some(now);
         app.bump_active_cell_revision();
     }
-    app.thinking_revision_owed = !due;
     due
 }
 
-pub(super) fn render_owed_revision(app: &mut App, now: Instant) -> bool {
-    if app.streaming_thinking_active_entry.is_none() {
-        app.thinking_revision_owed = false;
-    }
-    app.thinking_revision_owed && bump_thinking_revision_throttled(app, now)
-}
-
 /// Ensure an in-flight Thinking entry exists in `active_cell` and return its
 /// entry index. If no thinking entry is currently streaming, push a fresh one.
 /// P2.3: thinking shares the active cell with subsequent tool calls so the
@@ -86,8 +76,7 @@ pub(super) fn ensure_active_entry(app: &mut App) -> usize {
 /// committed to the cell immediately; the active-cell revision bump that
 /// triggers a re-wrap of the live tail is debounced to at most one per
 /// [`THINKING_REVISION_THROTTLE`] window (#1620). Skipped bumps coalesce into
-/// the next append, a later display beat or the forced finalize bump, so no
-/// content is ever lost.
+/// the next append or the forced finalize bump, so no content is ever lost.
 pub(super) fn append(app: &mut App, entry_idx: usize, text: &str) {
     append_at(app, entry_idx, text, Instant::now());
 }
@@ -319,7 +308,6 @@ pub(super) fn finalize_active_entry(app: &mut App, duration: Option<f32>, remain
     // the debounce window so the next thinking block's first chunk renders
     // immediately rather than being coalesced into a stale window.
     app.thinking_revision_last_bump_at = None;
-    app.thinking_revision_owed = false;
     app.bump_active_cell_revision();
     true
 }
@@ -438,34 +426,6 @@ mod tests {
         );
     }
 
-    #[test]
-    fn a_paused_stream_renders_the_reasoning_it_already_received() {
-        let mut app = test_app();
-        let entry = ensure_active_entry(&mut app);
-        app.thinking_revision_last_bump_at = None;
-        let t0 = Instant::now();
-        append_at(&mut app, entry, "first slice ", t0);
-        let shown = app.active_cell_revision;
-        append_at(&mut app, entry, "held back", t0 + Duration::from_millis(16));
-        assert_eq!(app.active_cell_revision, shown);
-
-        assert!(!render_owed_revision(
-            &mut app,
-            t0 + Duration::from_millis(32)
-        ));
-        assert_eq!(app.active_cell_revision, shown);
-        assert!(render_owed_revision(
-            &mut app,
-            t0 + THINKING_REVISION_THROTTLE
-        ));
-        assert_eq!(app.active_cell_revision, shown.wrapping_add(1));
-        assert!(!render_owed_revision(
-            &mut app,
-            t0 + THINKING_REVISION_THROTTLE * 3
-        ));
-        assert_eq!(app.active_cell_revision, shown.wrapping_add(1));
-    }
-
     /// #7123: a reasoning block that opens while the display beat still holds
     /// an uncommitted slice of prose must not drop that slice. `start_block`
     /// resets `StreamingState`, so the tail has to be delivered first.

crates/tui/src/tui/transcript/tests.rs

@@ -1441,8 +1441,8 @@ fn prose_cells_fill_full_width_on_ultrawide_by_default() {
     let mut cache = TranscriptViewCache::new();
     cache.ensure_filtered(&refs, &revisions, 220, options, &HashMap::new(), None, None);
 
-    for idx in 0..3 {
-        let width = max_line_width(&cache.per_cell[idx].lines);
+    for (idx, cell) in cache.per_cell.iter().enumerate().take(3) {
+        let width = max_line_width(&cell.lines);
         assert!(
             width > RETIRED_RAIL_MEASURE,
             "prose cell {idx} wrapped to {width} columns — still on the retired \
@@ -1495,8 +1495,8 @@ fn transcript_prose_measure_caps_prose_but_not_tools() {
     let mut cache = TranscriptViewCache::new();
     cache.ensure_filtered(&refs, &revisions, 220, options, &HashMap::new(), None, None);
 
-    for idx in 0..3 {
-        let width = max_line_width(&cache.per_cell[idx].lines);
+    for (idx, cell) in cache.per_cell.iter().enumerate().take(3) {
+        let width = max_line_width(&cell.lines);
         assert!(
             width > 105,
             "prose cell {idx} wrapped to {width} columns — still on the retired \

crates/tui/src/tui/ui/event_loop.rs

@@ -7232,16 +7232,14 @@ pub(crate) async fn run_event_loop(
                     continue;
                 }
                 // An idle operator can dismiss the persistent context warning
-                // or the account preferences notice without affecting vim or
-                // attachment handling. While a turn is active Esc retains its
-                // cancellation meaning.
+                // without affecting vim or attachment handling. While a turn
+                // is active Esc retains its cancellation meaning.
                 KeyCode::Esc
                     if !app.is_loading
                         && app.input.is_empty()
                         && !slash_menu_open
                         && !mention_menu_open
-                        && (app.dismiss_context_pressure_warning()
-                            || app.dismiss_account_preferences_notice()) =>
+                        && app.dismiss_context_pressure_warning() =>
                 {
                     continue;
                 }
@@ -9328,10 +9326,14 @@ pub(super) fn apply_engine_status(app: &mut App, message: String) -> bool {
         app.add_message(HistoryCell::System { content: message });
         return true;
     }
-    if app.tr(MessageId::ProfileConstitutionUnavailableLocal) == message.as_str() {
-        app.notify_account_preferences_not_loaded();
-        return false;
+    // The account-profile fallback notice is sent once per engine; keep it in
+    // the transcript so the next turn's status line cannot erase it.
+    let retain = app.tr(MessageId::ProfileConstitutionUnavailableLocal) == message.as_str();
+    if retain {
+        app.add_message(HistoryCell::System {
+            content: message.clone(),
+        });
     }
     app.status_message = Some(message);
-    false
+    retain
 }

crates/tui/src/tui/ui/frame.rs

@@ -1555,10 +1555,7 @@ pub(crate) fn commit_streaming_display_tick(
         updated = true;
     }
 
-    if streaming_thinking::render_owed_revision(app, now) {
-        updated = true;
-    }
-    if app.streaming_state.has_pending_stream_text(0) || app.thinking_revision_owed {
+    if app.streaming_state.has_pending_stream_text(0) {
         stream_display_clock.note_delta(now);
     }
 

crates/tui/src/tui/ui/terminal.rs

@@ -1110,7 +1110,7 @@ mod screen_mode_tests {
                         viewport: ratatui::Viewport::Inline(3),
                     },
                 )
-                .map_err(|err| io::Error::other(err.to_string()))
+                .map_err(|err| match err {})
             },
         )
         .expect("a successful probe must switch");

crates/tui/src/tui/ui/tests.rs

@@ -27343,7 +27343,7 @@ fn turn_inspector_switches_isolated_full_turn_pages_with_tagged_reasoning_and_ou
 
     let latest = pager.body_text();
     assert!(latest.contains("SECOND-PROMPT"), "{latest}");
-    assert!(latest.contains("[∿ thinking 1/1 · running]"), "{latest}");
+    assert!(latest.contains("[∿ reasoning 1/1 · running]"), "{latest}");
     assert!(latest.contains("SECOND-THOUGHT"), "{latest}");
     assert!(latest.contains("[◆ · running]"), "{latest}");
     assert!(latest.contains("SECOND-ANSWER"), "{latest}");
@@ -27355,7 +27355,7 @@ fn turn_inspector_switches_isolated_full_turn_pages_with_tagged_reasoning_and_ou
     for expected in [
         "FIRST-PROMPT",
         "keep this input line",
-        "[∿ thinking 1/1 · Done]",
+        "[∿ reasoning 1/1 · Done]",
         "FIRST-THOUGHT",
         "full reasoning tail",
         "[⚙ using tool]",
@@ -35488,41 +35488,23 @@ fn extension_prompt_origin_keeps_denials_and_other_policy_holds() {
 }
 
 #[test]
-fn account_profile_fallback_notice_is_a_standing_notice_outside_the_transcript() {
+fn account_profile_fallback_notice_stays_in_the_transcript() {
     let _home = SettingsHomeGuard::new();
     let mut app = App::new(create_test_options(), &Config::default());
-    let engine_notice = app
+    let notice = app
         .tr(MessageId::ProfileConstitutionUnavailableLocal)
         .to_string();
     let before = app.history.len();
-    let status_before = app.status_message.clone();
-    assert!(!super::event_loop::apply_engine_status(
+    assert!(super::event_loop::apply_engine_status(
         &mut app,
-        engine_notice
+        notice.clone()
     ));
-    assert_eq!(app.history.len(), before);
-    assert_eq!(app.status_message, status_before);
     assert!(!super::event_loop::apply_engine_status(
         &mut app,
         "Executing tools sequentially".into()
     ));
-    assert_eq!(app.history.len(), before);
-    let notice = app
-        .tr(MessageId::AccountPreferencesNotLoaded)
-        .replace("{command}", "codewhale login");
-    let standing = |app: &App| {
-        app.status_toasts
-            .iter()
-            .find(|toast| toast.text == notice)
-            .cloned()
-    };
-    app.sync_status_message_to_toasts();
-    let shown = standing(&app).expect("a later routine status does not erase the notice");
-    assert_eq!(shown.level, StatusToastLevel::Warning);
-    assert_eq!(shown.ttl_ms, None);
-    assert!(app.dismiss_account_preferences_notice());
-    assert!(standing(&app).is_none());
-    assert!(!app.dismiss_account_preferences_notice());
+    assert_eq!(app.history.len(), before + 1);
+    assert!(matches!(&app.history[before], HistoryCell::System { content } if *content == notice));
 }
 
 #[test]

crates/tui/tests/cucumber/main.rs

@@ -32,8 +32,6 @@ mod plugin_e2e_acceptance;
 mod screen_mode_inline_pty;
 #[cfg(all(unix, feature = "long-running-tests"))]
 mod search_text_pty;
-#[cfg(all(unix, feature = "long-running-tests"))]
-mod thinking_stream_pty;
 mod tool_lifecycle_acceptance;
 #[cfg(all(unix, feature = "long-running-tests"))]
 mod work_bar_keys_pty;

crates/tui/tests/cucumber/thinking_stream_pty.rs

@@ -1,585 +0,0 @@
-use std::io::{BufRead, BufReader, Read, Write};
-use std::net::{SocketAddr, TcpListener, TcpStream};
-use std::sync::atomic::{AtomicBool, AtomicUsize, Ordering};
-use std::sync::{Arc, Mutex, mpsc};
-use std::time::{Duration, Instant};
-
-use super::qa_harness::{
-    harness::{Harness, SealedWorkspace, make_sealed_workspace},
-    keys,
-};
-
-const WAIT: Duration = Duration::from_secs(20);
-const MODEL: &str = "deepseek-v4-pro";
-const PROMPTS: [&str; 2] = ["first slow fixture prompt", "second slow fixture prompt"];
-const ANSWER: &str = "Fixture answer after thinking.";
-const REASONING_LINES: usize = 6;
-const NOTICE: &str = "Account not loaded. Run codewhale login.";
-const RETIRED_TRANSCRIPT_NOTICE: &str = "could not be loaded";
-
-fn reasoning_line(index: usize) -> String {
-    format!("thinking fixture line {index} of {REASONING_LINES}")
-}
-
-struct PacedProvider {
-    address: SocketAddr,
-    gate: mpsc::Sender<()>,
-    stop: Arc<AtomicBool>,
-    turns: Arc<AtomicUsize>,
-    profile_requests: Arc<AtomicUsize>,
-    acceptor: Option<std::thread::JoinHandle<()>>,
-}
-
-struct Served {
-    gates: Mutex<mpsc::Receiver<()>>,
-    stop: Arc<AtomicBool>,
-    turns: Arc<AtomicUsize>,
-    profile_requests: Arc<AtomicUsize>,
-}
-
-impl PacedProvider {
-    fn start() -> Self {
-        let listener = TcpListener::bind("127.0.0.1:0").expect("loopback provider");
-        let address = listener.local_addr().expect("loopback address");
-        let (gate, gates) = mpsc::channel();
-        let stop = Arc::new(AtomicBool::new(false));
-        let turns = Arc::new(AtomicUsize::new(0));
-        let profile_requests = Arc::new(AtomicUsize::new(0));
-        let served = Arc::new(Served {
-            gates: Mutex::new(gates),
-            stop: Arc::clone(&stop),
-            turns: Arc::clone(&turns),
-            profile_requests: Arc::clone(&profile_requests),
-        });
-        let acceptor = std::thread::spawn(move || {
-            for stream in listener.incoming() {
-                if served.stop.load(Ordering::Acquire) {
-                    break;
-                }
-                let Ok(stream) = stream else { continue };
-                let served = Arc::clone(&served);
-                std::thread::spawn(move || {
-                    let _ = serve(stream, &served);
-                });
-            }
-        });
-        Self {
-            address,
-            gate,
-            stop,
-            turns,
-            profile_requests,
-            acceptor: Some(acceptor),
-        }
-    }
-
-    fn model_base(&self) -> String {
-        format!("http://{}/v1", self.address)
-    }
-
-    fn account_base(&self) -> String {
-        format!("http://{}", self.address)
-    }
-
-    fn release(&self, gates: usize) {
-        for _ in 0..gates {
-            self.gate.send(()).expect("provider gate");
-        }
-    }
-
-    fn turns(&self) -> usize {
-        self.turns.load(Ordering::Acquire)
-    }
-}
-
-impl Drop for PacedProvider {
-    fn drop(&mut self) {
-        self.stop.store(true, Ordering::Release);
-        let _ = TcpStream::connect(self.address);
-        if let Some(acceptor) = self.acceptor.take() {
-            let _ = acceptor.join();
-        }
-    }
-}
-
-fn serve(stream: TcpStream, served: &Served) -> std::io::Result<()> {
-    stream.set_read_timeout(Some(Duration::from_secs(10)))?;
-    let mut reader = BufReader::new(stream.try_clone()?);
-    let mut stream = stream;
-    let mut request_line = String::new();
-    if reader.read_line(&mut request_line)? == 0 {
-        return Ok(());
-    }
-    let mut content_length = 0usize;
-    loop {
-        let mut header = String::new();
-        if reader.read_line(&mut header)? == 0 {
-            break;
-        }
-        let header = header.trim_end();
-        if header.is_empty() {
-            break;
-        }
-        if let Some((name, value)) = header.split_once(':')
-            && name.eq_ignore_ascii_case("content-length")
-        {
-            content_length = value.trim().parse().unwrap_or(0);
-        }
-    }
-    let mut body = vec![0u8; content_length.min(8 * 1024 * 1024)];
-    reader.read_exact(&mut body)?;
-    let mut parts = request_line.split_whitespace();
-    let method = parts.next().unwrap_or_default();
-    let path = parts.next().unwrap_or_default();
-    if method == "GET" && path.ends_with("/models") {
-        return respond_json(
-            &mut stream,
-            "200 OK",
-            &serde_json::json!({"object":"list","data":[{"id":MODEL,"object":"model"}]}),
-        );
-    }
-    if path.starts_with("/api/") {
-        served.profile_requests.fetch_add(1, Ordering::AcqRel);
-        return respond_json(
-            &mut stream,
-            "401 Unauthorized",
-            &serde_json::json!({"error":"loopback fixture account is signed out"}),
-        );
-    }
-    if method != "POST" {
-        return respond_json(
-            &mut stream,
-            "404 Not Found",
-            &serde_json::json!({"error":"not served by the loopback fixture"}),
-        );
-    }
-    let request: serde_json::Value = serde_json::from_slice(&body).unwrap_or_default();
-    let streaming = request["stream"].as_bool().unwrap_or(false);
-    let with_tools = request["tools"]
-        .as_array()
-        .is_some_and(|tools| !tools.is_empty());
-    let text = String::from_utf8_lossy(&body);
-    let turn = PROMPTS
-        .iter()
-        .rposition(|prompt| text.contains(prompt))
-        .map(|index| index + 1);
-    let paced = streaming
-        && with_tools
-        && turn.is_some_and(|turn| {
-            served
-                .turns
-                .compare_exchange(turn - 1, turn, Ordering::AcqRel, Ordering::Acquire)
-                .is_ok()
-        });
-    if !streaming {
-        return respond_json(
-            &mut stream,
-            "200 OK",
-            &serde_json::json!({
-                "id":"loopback","object":"chat.completion","model":MODEL,
-                "choices":[{"index":0,"message":{"role":"assistant","content":"ok"},"finish_reason":"stop"}],
-                "usage":{"prompt_tokens":1,"completion_tokens":1,"total_tokens":2}
-            }),
-        );
-    }
-    stream.write_all(
-        b"HTTP/1.1 200 OK\r\ncontent-type: text/event-stream\r\ncache-control: no-cache\r\nconnection: close\r\n\r\n",
-    )?;
-    stream.flush()?;
-    if paced {
-        wait_for_gate(served)?;
-        for index in 1..=REASONING_LINES {
-            send_delta(
-                &mut stream,
-                serde_json::json!({"reasoning_content": format!("{}\n", reasoning_line(index))}),
-            )?;
-        }
-        wait_for_gate(served)?;
-        send_delta(&mut stream, serde_json::json!({"content": ANSWER}))?;
-        wait_for_gate(served)?;
-    } else {
-        send_delta(&mut stream, serde_json::json!({"content": "ok"}))?;
-    }
-    send_chunk(
-        &mut stream,
-        &serde_json::json!({
-            "id":"loopback","object":"chat.completion.chunk","model":MODEL,
-            "choices":[{"index":0,"delta":{},"finish_reason":"stop"}],
-            "usage":{"prompt_tokens":8,"completion_tokens":4,"total_tokens":12}
-        }),
-    )?;
-    stream.write_all(b"data: [DONE]\n\n")?;
-    stream.flush()
-}
-
-fn wait_for_gate(served: &Served) -> std::io::Result<()> {
-    let gates = served.gates.lock().expect("provider gates");
-    loop {
-        if served.stop.load(Ordering::Acquire) {
-            return Err(std::io::ErrorKind::Interrupted.into());
-        }
-        match gates.recv_timeout(Duration::from_millis(100)) {
-            Ok(()) => return Ok(()),
-            Err(mpsc::RecvTimeoutError::Timeout) => {}
-            Err(mpsc::RecvTimeoutError::Disconnected) => {
-                return Err(std::io::ErrorKind::BrokenPipe.into());
-            }
-        }
-    }
-}
-
-fn send_delta(stream: &mut TcpStream, delta: serde_json::Value) -> std::io::Result<()> {
-    send_chunk(
-        stream,
-        &serde_json::json!({
-            "id":"loopback","object":"chat.completion.chunk","model":MODEL,
-            "choices":[{"index":0,"delta":delta,"finish_reason":null}]
-        }),
-    )
-}
-
-fn send_chunk(stream: &mut TcpStream, chunk: &serde_json::Value) -> std::io::Result<()> {
-    stream.write_all(format!("data: {chunk}\n\n").as_bytes())?;
-    stream.flush()
-}
-
-fn respond_json(
-    stream: &mut TcpStream,
-    status: &str,
-    body: &serde_json::Value,
-) -> std::io::Result<()> {
-    let body = body.to_string();
-    stream.write_all(
-        format!(
-            "HTTP/1.1 {status}\r\ncontent-type: application/json\r\ncontent-length: {}\r\nconnection: close\r\n\r\n{body}",
-            body.len()
-        )
-        .as_bytes(),
-    )?;
-    stream.flush()
-}
-
-fn prepare(workspace: &SealedWorkspace, provider: &PacedProvider, settings: Option<&str>) {
-    let state = workspace.home().join(".codewhale");
-    std::fs::write(state.join(".onboarded"), "").unwrap();
-    std::fs::write(
-        state.join("config.toml"),
-        format!(
-            "provider = \"deepseek\"\n[providers.deepseek]\nbase_url = {:?}\napi_key = \"loopback-only\"\nmodel = {MODEL:?}\n[notifications]\nmethod = \"off\"\ncompletion_sound = \"off\"\n",
-            provider.model_base()
-        ),
-    )
-    .unwrap();
-    if let Some(settings) = settings {
-        std::fs::write(state.join("settings.toml"), settings).unwrap();
-    }
-    let trust = workspace.workspace().join(".deepseek");
-    std::fs::create_dir_all(&trust).unwrap();
-    std::fs::write(trust.join("trusted"), "").unwrap();
-}
-
-fn launch(
-    workspace: &SealedWorkspace,
-    provider: &PacedProvider,
-    (rows, cols): (u16, u16),
-    session: &str,
-    animated: bool,
-) -> Harness {
-    let mut tui = Harness::builder(Harness::codewhale_binary())
-        .cwd(workspace.workspace())
-        .clear_env()
-        .seal_home(workspace.home())
-        .env("PATH", "/usr/bin:/bin:/usr/sbin:/sbin")
-        .env("SHELL", "/bin/sh")
-        .env("CODEWHALE_CLOUD_API_BASE", provider.account_base())
-        .env("CODEWHALE_DISABLE_MODELS_DEV_FETCH", "1")
-        .env("CODEWHALE_NO_UPDATE_CHECK", "1")
-        .env("NO_ANIMATIONS", if animated { "0" } else { "1" })
-        .args([
-            "--workspace",
-            workspace.workspace().to_str().unwrap(),
-            "--no-project-config",
-            "--skip-onboarding",
-            session,
-        ])
-        .size(rows, cols)
-        .spawn()
-        .unwrap();
-    tui.wait_for_text("accept edits", WAIT)
-        .unwrap_or_else(|error| panic!("the dial chip never painted: {error}"));
-    tui
-}
-
-fn expect(tui: &mut Harness, what: &str, predicate: impl FnMut(&super::qa_harness::Frame) -> bool) {
-    if let Err(error) = tui.wait_for(predicate, WAIT) {
-        panic!("waiting for {what}: {error}");
-    }
-}
-
-fn settle(tui: &mut Harness) {
-    let deadline = Instant::now() + Duration::from_secs(2);
-    let mut quiet_since = Instant::now();
-    while Instant::now() < deadline && quiet_since.elapsed() < Duration::from_millis(300) {
-        if tui.pump() {
-            quiet_since = Instant::now();
-        }
-        std::thread::sleep(Duration::from_millis(20));
-    }
-}
-
-fn report(tui: &mut Harness, label: &str) {
-    let frame = tui.frame();
-    println!(
-        "--- {label} at {}x{} ---\n{}",
-        frame.cols(),
-        frame.rows(),
-        frame.debug_dump()
-    );
-}
-
-fn footer_row(tui: &mut Harness) -> (u16, String) {
-    let frame = tui.frame();
-    (0..frame.rows())
-        .rev()
-        .map(|row| (row, frame.row(row)))
-        .find(|(_, text)| text.contains("accept edits"))
-        .unwrap_or_else(|| panic!("no dial chip on screen:\n{}", frame.debug_dump()))
-}
-
-fn assert_status_and_water(tui: &mut Harness, label: &str) {
-    let (footer, status) = footer_row(tui);
-    let frame = tui.frame();
-    for retired in ["operate", "Operate", "Work bar"] {
-        assert!(
-            !frame.contains(retired),
-            "{label}: `{retired}` is on screen:\n{}",
-            frame.debug_dump()
-        );
-    }
-    assert!(status.contains("accept edits"), "{label}: {status}");
-    for row in 0..footer.saturating_sub(3) {
-        let text = frame.row(row);
-        let cells: Vec<char> = text.chars().collect();
-        let school = cells.windows(2).any(|pair| {
-            pair.iter()
-                .all(|cell| ('\u{2801}'..='\u{28FF}').contains(cell))
-        });
-        assert!(
-            !school && !text.contains(['˚', '°', '≈', '～']),
-            "{label}: ambient art on transcript row {row} `{text}`:\n{}",
-            frame.debug_dump()
-        );
-    }
-}
-
-fn thinking_rows(tui: &mut Harness) -> usize {
-    let frame = tui.frame();
-    (0..frame.rows())
-        .filter(|&row| frame.row(row).contains("Thinking…"))
-        .count()
-}
-
-fn stream_first_turn(tui: &mut Harness, provider: &PacedProvider, thinking_shown: bool) {
-    tui.type_line(PROMPTS[0]).unwrap();
-    expect(tui, "the first request and its interrupt hint", |frame| {
-        provider.turns() == 1 && frame.contains("Esc to interrupt")
-    });
-    std::thread::sleep(Duration::from_millis(900));
-    report(tui, "loading");
-    assert_status_and_water(tui, "loading");
-
-    provider.release(1);
-    let tail = reasoning_line(REASONING_LINES);
-    if thinking_shown {
-        expect(tui, "the live thinking block", |frame| {
-            frame.contains("Thinking…") && frame.contains(&tail)
-        });
-    } else {
-        expect(tui, "the thinking activity line", |frame| {
-            frame.contains("Thinking…")
-        });
-    }
-    std::thread::sleep(Duration::from_millis(1300));
-    report(tui, "thinking");
-    assert_status_and_water(tui, "thinking");
-    assert_eq!(thinking_rows(tui), 1, "one Thinking line");
-    let (_, status) = footer_row(tui);
-    assert!(
-        status.contains("thinking ") && !status.contains("reasoning"),
-        "the clock names the phase: {status}"
-    );
-    let frame = tui.frame();
-    assert!(!frame.contains("reasoning..."), "{}", frame.debug_dump());
-    assert_eq!(frame.contains(&tail), thinking_shown);
-    assert!(
-        !frame.contains(&reasoning_line(1)),
-        "a long trace stays inside its window:\n{}",
-        frame.debug_dump()
-    );
-
-    provider.release(1);
-    expect(tui, "the streamed answer", |frame| frame.contains(ANSWER));
-    report(tui, "answer");
-    let frame = tui.frame();
-    assert!(!frame.contains("Thinking…"), "{}", frame.debug_dump());
-    assert_eq!(frame.contains("Thought for "), thinking_shown);
-    assert!(!frame.contains(&tail), "{}", frame.debug_dump());
-
-    provider.release(1);
-    expect(tui, "the settled turn", |frame| {
-        !frame.contains("Esc to interrupt")
-    });
-    settle(tui);
-    report(tui, "settled");
-    assert_status_and_water(tui, "settled");
-}
-
-#[test]
-fn thinking_streams_stops_and_restores_at_the_shipped_widths() {
-    for size in [(24, 80), (32, 100), (32, 120), (40, 140)] {
-        let workspace = make_sealed_workspace().unwrap();
-        let provider = PacedProvider::start();
-        prepare(&workspace, &provider, None);
-        let mut tui = launch(&workspace, &provider, size, "--fresh", true);
-        stream_first_turn(&mut tui, &provider, true);
-
-        tui.type_line(PROMPTS[1]).unwrap();
-        expect(&mut tui, "the second request", |_| provider.turns() == 2);
-        provider.release(1);
-        let tail = reasoning_line(REASONING_LINES);
-        expect(&mut tui, "thinking before the stop", |frame| {
-            frame.contains("Thinking…") && frame.contains(&tail)
-        });
-        tui.send(keys::key::esc()).unwrap();
-        expect(&mut tui, "the stopped turn", |frame| {
-            !frame.contains("Esc to interrupt") && !frame.contains("Thinking…")
-        });
-        settle(&mut tui);
-        report(&mut tui, "stopped");
-        assert_status_and_water(&mut tui, "stopped");
-        provider.release(2);
-
-        tui.send(keys::key::ctrl('u')).unwrap();
-        settle(&mut tui);
-        tui.send(keys::key::ctrl_d()).unwrap();
-        tui.wait_for_exit(Duration::from_secs(10));
-        tui.shutdown();
-
-        let mut tui = launch(&workspace, &provider, size, "--continue", false);
-        expect(&mut tui, "the restored conversation", |frame| {
-            frame.contains(ANSWER)
-        });
-        settle(&mut tui);
-        report(&mut tui, "restored");
-        assert_status_and_water(&mut tui, "restored");
-        let frame = tui.frame();
-        assert!(!frame.contains("Thinking…"), "{}", frame.debug_dump());
-        tui.shutdown();
-    }
-}
-
-#[test]
-fn a_saved_choice_to_hide_thinking_survives_the_new_default() {
-    let workspace = make_sealed_workspace().unwrap();
-    let provider = PacedProvider::start();
-    prepare(&workspace, &provider, Some("show_thinking = false\n"));
-    let mut tui = launch(&workspace, &provider, (24, 80), "--fresh", false);
-    stream_first_turn(&mut tui, &provider, false);
-    let frame = tui.frame();
-    assert!(!frame.contains("Thought"), "{}", frame.debug_dump());
-    tui.shutdown();
-    let saved: toml::Value = toml::from_str(
-        &std::fs::read_to_string(workspace.home().join(".codewhale/settings.toml")).unwrap(),
-    )
-    .unwrap();
-    assert_eq!(saved["show_thinking"].as_bool(), Some(false));
-}
-
-fn sign_in(workspace: &SealedWorkspace, provider: &PacedProvider, session: &str) {
-    let path = workspace.home().join(".codewhale/secrets/secrets.json");
-    std::fs::create_dir_all(path.parent().unwrap()).unwrap();
-    let secrets =
-        codewhale_secrets::Secrets::new(Arc::new(codewhale_secrets::FileKeyringStore::new(path)));
-    let bundle = serde_json::from_value(serde_json::json!({
-        "tokenType": "Bearer",
-        "accessToken": "loopback-access",
-        "refreshToken": "loopback-refresh",
-        "session": {"id": session},
-        "user": {"id": "loopback-account"}
-    }))
-    .unwrap();
-    codewhale_secrets::account::AccountSessionStore::new(secrets, None, &provider.account_base())
-        .save(bundle)
-        .unwrap();
-}
-
-fn run_ungated_turn(tui: &mut Harness, provider: &PacedProvider) {
-    provider.release(3);
-    tui.type_line(PROMPTS[0]).unwrap();
-    expect(tui, "the answer", |frame| frame.contains(ANSWER));
-    expect(tui, "the settled turn", |frame| {
-        !frame.contains("Esc to interrupt")
-    });
-}
-
-#[test]
-fn account_preferences_notice_is_shown_once_per_sign_in_and_esc_dismisses_it() {
-    let workspace = make_sealed_workspace().unwrap();
-    let provider = PacedProvider::start();
-    prepare(&workspace, &provider, None);
-    sign_in(&workspace, &provider, "first-sign-in");
-
-    for cols in [80, 120] {
-        let workspace = make_sealed_workspace().unwrap();
-        let provider = PacedProvider::start();
-        prepare(&workspace, &provider, None);
-        sign_in(&workspace, &provider, "first-sign-in");
-        let mut tui = launch(&workspace, &provider, (32, cols), "--fresh", false);
-        run_ungated_turn(&mut tui, &provider);
-        expect(&mut tui, "the standing notice", |frame| {
-            frame.contains(NOTICE)
-        });
-        report(&mut tui, "account notice");
-        assert!(provider.profile_requests.load(Ordering::Acquire) > 0);
-        let (_, status) = footer_row(&mut tui);
-        assert!(
-            status.contains(NOTICE),
-            "the notice is footer chrome: {status}"
-        );
-        assert!(!tui.frame().contains(RETIRED_TRANSCRIPT_NOTICE));
-        tui.send(keys::key::esc()).unwrap();
-        expect(&mut tui, "the dismissed notice", |frame| {
-            !frame.contains(NOTICE)
-        });
-        report(&mut tui, "account notice dismissed");
-        tui.shutdown();
-    }
-
-    let mut tui = launch(&workspace, &provider, (32, 120), "--fresh", false);
-    run_ungated_turn(&mut tui, &provider);
-    expect(&mut tui, "the first notice", |frame| frame.contains(NOTICE));
-    tui.shutdown();
-
-    let mut tui = launch(&workspace, &provider, (32, 120), "--fresh", false);
-    provider.turns.store(0, Ordering::Release);
-    run_ungated_turn(&mut tui, &provider);
-    tui.wait_for_idle(Duration::from_secs(6), Duration::from_secs(30))
-        .unwrap();
-    report(&mut tui, "same sign-in, later session");
-    assert!(
-        !tui.frame().contains(NOTICE),
-        "the same sign-in is told once"
-    );
-    assert!(!tui.frame().contains(RETIRED_TRANSCRIPT_NOTICE));
-    tui.shutdown();
-
-    sign_in(&workspace, &provider, "second-sign-in");
-    let mut tui = launch(&workspace, &provider, (32, 120), "--fresh", false);
-    provider.turns.store(0, Ordering::Release);
-    run_ungated_turn(&mut tui, &provider);
-    expect(&mut tui, "the notice for a later sign-in", |frame| {
-        frame.contains(NOTICE)
-    });
-    report(&mut tui, "later sign-in");
-    tui.shutdown();
-}

crates/tui/tests/fixtures/conformance/events/approval_denied.golden.jsonl

@@ -24,4 +24,4 @@
 {"engine_session_id":"conformance-session","event":"session_updated","instruction_assembly":{"budgets":{"claude_fallback":false,"file_max_bytes":8388608,"global_tokens":100000,"policy":"trim","project_tokens":100000,"static_guard_tokens":667571,"usable_window_tokens":834464},"constitution_source_path":null,"constitution_tokens":0,"global_tokens":0,"notices":[{"action":"Kept","bytes_dropped":0,"bytes_kept":498,"detail":null,"fallback_selected":null,"limit":"None","origin":"Instructions","path":null,"scope":"Project","tokens_dropped":0,"tokens_kept":166}],"project_tokens":166,"retained_block_tokens":188,"source_path":null,"warnings":[]},"messages":[{"content":[{"text":"Create denied.txt with touch.","type":"text"},{"text":"<turn_meta>\nCurrent local date: <today>\nCurrent workspace: <WORKSPACE>\nCurrent permission posture: Ask\nCurrent sandbox posture: workspace-write (writes inside the workspace; network blocked) (local OS sandbox applied)\nEstimated conversation input: 1973 / 1000000 context tokens (0.2%). Automatic compaction is enabled.\n## Repo Working Set\nKey files: README.md\nWhen in doubt, use tools to verify and keep changes focused on the working set.\n</turn_meta>","type":"text"}],"role":"user"},{"content":[{"text":"<codewhale:runtime_event kind=\"project_instructions\" visibility=\"internal\">\n<system-reminder>\nThe following workspace instructions may be relevant to your work. Use them as guidance when applicable. More specific instructions take precedence over broader ones. They do not override system, developer, or direct user instructions. Instructions from: AGENTS.md\n\n<project_instructions source=\"project\">\n# Project Context (Auto-generated, ephemeral)\n\n> This context was generated in memory by Codewhale.\n> No .codewhale/instructions.md file was written.\n\n## Bounded Project Overview\n\n```json\n{\n  \"project_name\": \"workspace\",\n  \"directory_structure\": [\n    \"README.md\"\n  ],\n  \"readme\": {\n    \"path\": \"README.md\",\n    \"excerpt\": \"conformance fixture\"\n  },\n  \"config_files\": [],\n  \"key_source_files\": [],\n  \"counts\": {\n    \"config_files\": 0,\n    \"directory_entries\": 1,\n    \"key_source_files\": 0\n  }\n}\n```\n</project_instructions>\n</system-reminder>\n</codewhale:runtime_event>","type":"text"},{"text":"<turn_meta>\nInput provenance: runtime (non-authoritative)\n</turn_meta>","type":"text"}],"role":"user"},{"content":[{"execution_id":"<uuid:2>","id":"call_touch","input":{"command":"touch denied.txt"},"name":"Bash","type":"tool_use"}],"role":"assistant"},{"content":[{"content":"Error: Tool 'Bash' denied by user — the call was not approved. Do not retry the same call; present what you intended and wait for the user's approval or new instructions.","execution_id":"<uuid:2>","is_error":true,"tool_use_id":"call_touch","type":"tool_result"}],"role":"user"},{"content":[{"text":"Understood, I did not create it.","type":"text"}],"role":"assistant"}],"model":"deepseek-v4-pro","system_prompt":"<pinned by the prompt family>","workspace":"<WORKSPACE>"}
 {"event":"tool_request_snapshot","snapshot":{"active_tool_catalog_sha256":"9017c53ee026ae4065bed3e4b2e7d773a828556b42d6ebe2c2ebfd8e65377280","capture_source":"prepared model-client request","delivery_status":"unknown (capture does not prove provider delivery)","omitted_tool_count":0,"payload_json_bytes":17394,"payload_measurement_status":"exact (within 1048576-byte measurement bound)","provider":{"model":"deepseek-v4-pro","provider":"Deepseek","status":"available"},"registry_facts_present":true,"registry_only_tools":{"status":"known","value":{"count":40,"omitted":8,"rendered":[{"truncated":false,"value":"Git"},{"truncated":false,"value":"Run"},{"truncated":false,"value":"Web"},{"truncated":false,"value":"apply_patch"},{"truncated":false,"value":"automation"},{"truncated":false,"value":"context_archive_get"},{"truncated":false,"value":"context_archive_search"},{"truncated":false,"value":"diagnostics"},{"truncated":false,"value":"finance"},{"truncated":false,"value":"github"},{"truncated":false,"value":"handle_read"},{"truncated":false,"value":"harness"},{"truncated":false,"value":"image_ocr"},{"truncated":false,"value":"list_dir"},{"truncated":false,"value":"lsp"},{"truncated":false,"value":"note"},{"truncated":false,"value":"notify"},{"truncated":false,"value":"pandoc_convert"},{"truncated":false,"value":"project_map"},{"truncated":false,"value":"registry_sync"},{"truncated":false,"value":"request_plugin_install"},{"truncated":false,"value":"request_user_input"},{"truncated":false,"value":"retrieve_tool_result"},{"truncated":false,"value":"revert_turn"},{"truncated":false,"value":"review"},{"truncated":false,"value":"send_later"},{"truncated":false,"value":"session_get"},{"truncated":false,"value":"session_search"},{"truncated":false,"value":"start_mcp_server"},{"truncated":false,"value":"start_registry_mcp_server"},{"truncated":false,"value":"tasks"},{"truncated":false,"value":"terminal/cancel"}]}},"registry_tool_count":{"status":"known","value":65},"rendered_tool_count":15,"schema_version":1,"step":1,"terminal":{"automatic_compaction_attempts":0,"effective_max_steps":null,"emergency_compaction_attempts":0,"empty_stop_retries":0,"final_report_requested":false,"last_prepared_output_limit_tokens":65536,"last_provider_finish_reason":{"truncated":false,"value":"end_turn"},"last_reported_input_tokens":200,"last_response_tool_calls":0,"last_response_tool_calls_suppressed":0,"model_requests_started":2,"model_step_index":1,"permission_denial_rounds_without_progress":0,"permission_strategy_switches":0,"reason":"provider_no_tool_call","reasoning_only_reprompts":0,"route_context_window_tokens":1000000,"status":"completed","step_budget_source":"max_steps","stream_resumes":0,"transparent_stream_retries":0,"transport_retries":0},"tool_count":15,"tools":[{"allowed_callers":{"status":"known","value":{"count":1,"omitted":0,"rendered":[{"truncated":false,"value":"direct"}]}},"approval":{"status":"known","value":{"truncated":false,"value":"Required"}},"cache_control_type":{"reason":"request field absent","status":"unknown"},"capabilities":{"status":"known","value":{"count":3,"omitted":0,"rendered":[{"truncated":false,"value":"ExecutesCode"},{"truncated":false,"value":"Sandboxable"},{"truncated":false,"value":"RequiresApproval"}]}},"defer_loading":{"status":"known","value":false},"description":{"truncated":true,"value":"Execute a shell command in the workspace and return stdout and stderr (last 2000 lines or 50KB). Optional timeout is foreground seconds (default 30 seconds); a command still running then moves to the background by itself — never killed for time — returning output so far, last-output age, build phase, and a task_id. Call task_shell_wait with that task_id to read more or wait; do not sleep-poll with bash. Before a final answer, collect every still-relevant background task (wait only when genuinely blocked) an"},"input_examples":{"reason":"request field absent","status":"unknown"},"input_schema_json":{"truncated":false,"value":"{\"additionalProperties\":false,\"properties\":{\"command\":{\"description\":\"The command to execute. Actual execution shell: `/bin/bash`. Use Bash syntax: pipelines, redirections, $(command), and && / || are supported. Quote paths and variable expansions, such as \\\"$path\\\"; use single quotes for literal text. For literal multiline input, use a quoted heredoc delimiter (<<'EOF') with its closing delimiter on a separate line. Use only installed programs; do not assume GNU-specific flags on macOS/BSD. Example: printf '%s\\\\n' 'sample'.\",\"type\":\"string\"},\"description\":{\"description\":\"Optional few-word summary shown to the user beside the command.\",\"type\":\"string\"},\"justification\":{\"description\":\"Required with sandbox_permissions: one sentence explaining why this exact command needs wider access.\",\"type\":\"string\"},\"read_only\":{\"description\":\"Set true to run analysis code (including Python/SQLite) or a network fetch (curl) with mandatory native filesysteLine truncated
 {"event":"turn_complete","parent_route_usage":{"input_tokens":350,"output_tokens":21},"routed_usage_dropped_records":0,"status":"completed","tool_catalog":"<pinned by the prompt family>","turn_id":"<uuid:1>","usage":{"input_tokens":350,"output_tokens":21}}
-{"harness_summary":{"model_requests":2,"non_streaming_requests":0,"workspace_after":{".codewhale/state/.gitignore":"cdbcae15105d6b78",".codewhale/state/subagents.v1.json/legacy.json":"9064f16acf90984c",".codewhale/state/subagents.v3.json":"e874da2951e53d2b",".codewhale/state/subagents.v3.transaction.lock":"e3b0c44298fc1c14","README.md":"41f49bffb38de5ba"}}}
+{"harness_summary":{"model_requests":2,"non_streaming_requests":0,"workspace_after":{".codewhale/state/.gitignore":"cdbcae15105d6b78",".codewhale/state/subagents.v1.json/legacy.json":"9064f16acf90984c",".codewhale/state/subagents.v1.lock":"e3b0c44298fc1c14",".codewhale/state/subagents.v3.json":"e874da2951e53d2b",".codewhale/state/subagents.v3.transaction.lock":"e3b0c44298fc1c14","README.md":"41f49bffb38de5ba"}}}

crates/tui/tests/fixtures/conformance/events/cancel_midstream.golden.jsonl

@@ -15,4 +15,4 @@
 {"event":"tool_request_snapshot","snapshot":{"active_tool_catalog_sha256":"9017c53ee026ae4065bed3e4b2e7d773a828556b42d6ebe2c2ebfd8e65377280","capture_source":"prepared model-client request","delivery_status":"unknown (capture does not prove provider delivery)","omitted_tool_count":0,"payload_json_bytes":17394,"payload_measurement_status":"exact (within 1048576-byte measurement bound)","provider":{"model":"deepseek-v4-pro","provider":"Deepseek","status":"available"},"registry_facts_present":true,"registry_only_tools":{"status":"known","value":{"count":40,"omitted":8,"rendered":[{"truncated":false,"value":"Git"},{"truncated":false,"value":"Run"},{"truncated":false,"value":"Web"},{"truncated":false,"value":"apply_patch"},{"truncated":false,"value":"automation"},{"truncated":false,"value":"context_archive_get"},{"truncated":false,"value":"context_archive_search"},{"truncated":false,"value":"diagnostics"},{"truncated":false,"value":"finance"},{"truncated":false,"value":"github"},{"truncated":false,"value":"handle_read"},{"truncated":false,"value":"harness"},{"truncated":false,"value":"image_ocr"},{"truncated":false,"value":"list_dir"},{"truncated":false,"value":"lsp"},{"truncated":false,"value":"note"},{"truncated":false,"value":"notify"},{"truncated":false,"value":"pandoc_convert"},{"truncated":false,"value":"project_map"},{"truncated":false,"value":"registry_sync"},{"truncated":false,"value":"request_plugin_install"},{"truncated":false,"value":"request_user_input"},{"truncated":false,"value":"retrieve_tool_result"},{"truncated":false,"value":"revert_turn"},{"truncated":false,"value":"review"},{"truncated":false,"value":"send_later"},{"truncated":false,"value":"session_get"},{"truncated":false,"value":"session_search"},{"truncated":false,"value":"start_mcp_server"},{"truncated":false,"value":"start_registry_mcp_server"},{"truncated":false,"value":"tasks"},{"truncated":false,"value":"terminal/cancel"}]}},"registry_tool_count":{"status":"known","value":65},"rendered_tool_count":15,"schema_version":1,"step":0,"terminal":{"automatic_compaction_attempts":0,"effective_max_steps":null,"emergency_compaction_attempts":0,"empty_stop_retries":0,"final_report_requested":false,"last_prepared_output_limit_tokens":65536,"last_provider_finish_reason":null,"last_reported_input_tokens":120,"last_response_tool_calls":0,"last_response_tool_calls_suppressed":null,"model_requests_started":1,"model_step_index":0,"permission_denial_rounds_without_progress":0,"permission_strategy_switches":0,"reason":"interrupted","reasoning_only_reprompts":0,"route_context_window_tokens":1000000,"status":"interrupted","step_budget_source":"max_steps","stream_resumes":0,"transparent_stream_retries":0,"transport_retries":0},"tool_count":15,"tools":[{"allowed_callers":{"status":"known","value":{"count":1,"omitted":0,"rendered":[{"truncated":false,"value":"direct"}]}},"approval":{"status":"known","value":{"truncated":false,"value":"Required"}},"cache_control_type":{"reason":"request field absent","status":"unknown"},"capabilities":{"status":"known","value":{"count":3,"omitted":0,"rendered":[{"truncated":false,"value":"ExecutesCode"},{"truncated":false,"value":"Sandboxable"},{"truncated":false,"value":"RequiresApproval"}]}},"defer_loading":{"status":"known","value":false},"description":{"truncated":true,"value":"Execute a shell command in the workspace and return stdout and stderr (last 2000 lines or 50KB). Optional timeout is foreground seconds (default 30 seconds); a command still running then moves to the background by itself — never killed for time — returning output so far, last-output age, build phase, and a task_id. Call task_shell_wait with that task_id to read more or wait; do not sleep-poll with bash. Before a final answer, collect every still-relevant background task (wait only when genuinely blocked) an"},"input_examples":{"reason":"request field absent","status":"unknown"},"input_schema_json":{"truncated":false,"value":"{\"additionalProperties\":false,\"properties\":{\"command\":{\"description\":\"The command to execute. Actual execution shell: `/bin/bash`. Use Bash syntax: pipelines, redirections, $(command), and && / || are supported. Quote paths and variable expansions, such as \\\"$path\\\"; use single quotes for literal text. For literal multiline input, use a quoted heredoc delimiter (<<'EOF') with its closing delimiter on a separate line. Use only installed programs; do not assume GNU-specific flags on macOS/BSD. Example: printf '%s\\\\n' 'sample'.\",\"type\":\"string\"},\"description\":{\"description\":\"Optional few-word summary shown to the user beside the command.\",\"type\":\"string\"},\"justification\":{\"description\":\"Required with sandbox_permissions: one sentence explaining why this exact command needs wider access.\",\"type\":\"string\"},\"read_only\":{\"description\":\"Set true to run analysis code (including Python/SQLite) or a network fetch (curl) with mandatory native filesystem read-only isolation. Outbound networkLine truncated
 {"event":"turn_complete","parent_route_usage":{"input_tokens":120,"output_tokens":0},"routed_usage_dropped_records":0,"status":"interrupted","tool_catalog":"<pinned by the prompt family>","turn_id":"<uuid:1>","usage":{"input_tokens":120,"output_tokens":0}}
 {"event":"status","message":"Turn interrupted."}
-{"harness_summary":{"model_requests":1,"non_streaming_requests":0,"workspace_after":{".codewhale/state/.gitignore":"cdbcae15105d6b78",".codewhale/state/subagents.v1.json/legacy.json":"9064f16acf90984c",".codewhale/state/subagents.v3.json":"e874da2951e53d2b",".codewhale/state/subagents.v3.transaction.lock":"e3b0c44298fc1c14","NOTES.md":"f957b19529906961","README.md":"41f49bffb38de5ba"}}}
+{"harness_summary":{"model_requests":1,"non_streaming_requests":0,"workspace_after":{".codewhale/state/.gitignore":"cdbcae15105d6b78",".codewhale/state/subagents.v1.json/legacy.json":"9064f16acf90984c",".codewhale/state/subagents.v1.lock":"e3b0c44298fc1c14",".codewhale/state/subagents.v3.json":"e874da2951e53d2b",".codewhale/state/subagents.v3.transaction.lock":"e3b0c44298fc1c14","NOTES.md":"f957b19529906961","README.md":"41f49bffb38de5ba"}}}

crates/tui/tests/fixtures/conformance/events/one_tool_call.golden.jsonl

@@ -25,4 +25,4 @@
 {"engine_session_id":"conformance-session","event":"session_updated","instruction_assembly":{"budgets":{"claude_fallback":false,"file_max_bytes":8388608,"global_tokens":100000,"policy":"trim","project_tokens":100000,"static_guard_tokens":667571,"usable_window_tokens":834464},"constitution_source_path":null,"constitution_tokens":0,"global_tokens":0,"notices":[{"action":"Kept","bytes_dropped":0,"bytes_kept":514,"detail":null,"fallback_selected":null,"limit":"None","origin":"Instructions","path":null,"scope":"Project","tokens_dropped":0,"tokens_kept":172}],"project_tokens":172,"retained_block_tokens":193,"source_path":null,"warnings":[]},"messages":[{"content":[{"text":"What does README.md say?","type":"text"},{"text":"<turn_meta>\nCurrent local date: <today>\nCurrent workspace: <WORKSPACE>\nCurrent permission posture: Ask\nCurrent sandbox posture: workspace-write (writes inside the workspace; network blocked) (local OS sandbox applied)\nEstimated conversation input: 1972 / 1000000 context tokens (0.2%). Automatic compaction is enabled.\n## Repo Working Set\nKey files: README.md\nActive paths (prioritize these):\n- README.md (file)\nWhen in doubt, use tools to verify and keep changes focused on the working set.\n</turn_meta>","type":"text"}],"role":"user"},{"content":[{"text":"<codewhale:runtime_event kind=\"project_instructions\" visibility=\"internal\">\n<system-reminder>\nThe following workspace instructions may be relevant to your work. Use them as guidance when applicable. More specific instructions take precedence over broader ones. They do not override system, developer, or direct user instructions. Instructions from: AGENTS.md\n\n<project_instructions source=\"project\">\n# Project Context (Auto-generated, ephemeral)\n\n> This context was generated in memory by Codewhale.\n> No .codewhale/instructions.md file was written.\n\n## Bounded Project Overview\n\n```json\n{\n  \"project_name\": \"workspace\",\n  \"directory_structure\": [\n    \"README.md\",\n    \"NOTES.md\"\n  ],\n  \"readme\": {\n    \"path\": \"README.md\",\n    \"excerpt\": \"conformance fixture\"\n  },\n  \"config_files\": [],\n  \"key_source_files\": [],\n  \"counts\": {\n    \"config_files\": 0,\n    \"directory_entries\": 2,\n    \"key_source_files\": 0\n  }\n}\n```\n</project_instructions>\n</system-reminder>\n</codewhale:runtime_event>","type":"text"},{"text":"<turn_meta>\nInput provenance: runtime (non-authoritative)\n</turn_meta>","type":"text"}],"role":"user"},{"content":[{"execution_id":"<uuid:2>","id":"call_read_readme","input":{"action":"read","path":"README.md"},"name":"File","type":"tool_use"}],"role":"assistant"},{"content":[{"content":"content_hash=\"sha256:41f49bffb38de5ba88945f09121ffc590f9d7f6ff80fdb9ea8ad10224fecea61\"\nconformance fixture","execution_id":"<uuid:2>","tool_use_id":"call_read_readme","type":"tool_result"}],"role":"user"},{"content":[{"text":"It says: conformance fixture.","type":"text"}],"role":"assistant"}],"model":"deepseek-v4-pro","system_prompt":"<pinned by the prompt family>","workspace":"<WORKSPACE>"}
 {"event":"tool_request_snapshot","snapshot":{"active_tool_catalog_sha256":"9017c53ee026ae4065bed3e4b2e7d773a828556b42d6ebe2c2ebfd8e65377280","capture_source":"prepared model-client request","delivery_status":"unknown (capture does not prove provider delivery)","omitted_tool_count":0,"payload_json_bytes":17394,"payload_measurement_status":"exact (within 1048576-byte measurement bound)","provider":{"model":"deepseek-v4-pro","provider":"Deepseek","status":"available"},"registry_facts_present":true,"registry_only_tools":{"status":"known","value":{"count":40,"omitted":8,"rendered":[{"truncated":false,"value":"Git"},{"truncated":false,"value":"Run"},{"truncated":false,"value":"Web"},{"truncated":false,"value":"apply_patch"},{"truncated":false,"value":"automation"},{"truncated":false,"value":"context_archive_get"},{"truncated":false,"value":"context_archive_search"},{"truncated":false,"value":"diagnostics"},{"truncated":false,"value":"finance"},{"truncated":false,"value":"github"},{"truncated":false,"value":"handle_read"},{"truncated":false,"value":"harness"},{"truncated":false,"value":"image_ocr"},{"truncated":false,"value":"list_dir"},{"truncated":false,"value":"lsp"},{"truncated":false,"value":"note"},{"truncated":false,"value":"notify"},{"truncated":false,"value":"pandoc_convert"},{"truncated":false,"value":"project_map"},{"truncated":false,"value":"registry_sync"},{"truncated":false,"value":"request_plugin_install"},{"truncated":false,"value":"request_user_input"},{"truncated":false,"value":"retrieve_tool_result"},{"truncated":false,"value":"revert_turn"},{"truncated":false,"value":"review"},{"truncated":false,"value":"send_later"},{"truncated":false,"value":"session_get"},{"truncated":false,"value":"session_search"},{"truncated":false,"value":"start_mcp_server"},{"truncated":false,"value":"start_registry_mcp_server"},{"truncated":false,"value":"tasks"},{"truncated":false,"value":"terminal/cancel"}]}},"registry_tool_count":{"status":"known","value":65},"rendered_tool_count":15,"schema_version":1,"step":1,"terminal":{"automatic_compaction_attempts":0,"effective_max_steps":null,"emergency_compaction_attempts":0,"empty_stop_retries":0,"final_report_requested":false,"last_prepared_output_limit_tokens":65536,"last_provider_finish_reason":{"truncated":false,"value":"end_turn"},"last_reported_input_tokens":200,"last_response_tool_calls":0,"last_response_tool_calls_suppressed":0,"model_requests_started":2,"model_step_index":1,"permission_denial_rounds_without_progress":0,"permission_strategy_switches":0,"reason":"provider_no_tool_call","reasoning_only_reprompts":0,"route_context_window_tokens":1000000,"status":"completed","step_budget_source":"max_steps","stream_resumes":0,"transparent_stream_retries":0,"transport_retries":0},"tool_count":15,"tools":[{"allowed_callers":{"status":"known","value":{"count":1,"omitted":0,"rendered":[{"truncated":false,"value":"direct"}]}},"approval":{"status":"known","value":{"truncated":false,"value":"Required"}},"cache_control_type":{"reason":"request field absent","status":"unknown"},"capabilities":{"status":"known","value":{"count":3,"omitted":0,"rendered":[{"truncated":false,"value":"ExecutesCode"},{"truncated":false,"value":"Sandboxable"},{"truncated":false,"value":"RequiresApproval"}]}},"defer_loading":{"status":"known","value":false},"description":{"truncated":true,"value":"Execute a shell command in the workspace and return stdout and stderr (last 2000 lines or 50KB). Optional timeout is foreground seconds (default 30 seconds); a command still running then moves to the background by itself — never killed for time — returning output so far, last-output age, build phase, and a task_id. Call task_shell_wait with that task_id to read more or wait; do not sleep-poll with bash. Before a final answer, collect every still-relevant background task (wait only when genuinely blocked) an"},"input_examples":{"reason":"request field absent","status":"unknown"},"input_schema_json":{"truncated":false,"value":"{\"additionalProperties\":false,\"properties\":{\"command\":{\"description\":\"The command to execute. Actual execution shell: `/bin/bash`. Use Bash syntax: pipelines, redirections, $(command), and && / || are supported. Quote paths and variable expansions, such as \\\"$path\\\"; use single quotes for literal text. For literal multiline input, use a quoted heredoc delimiter (<<'EOF') with its closing delimiter on a separate line. Use only installed programs; do not assume GNU-specific flags on macOS/BSD. Example: printf '%s\\\\n' 'sample'.\",\"type\":\"string\"},\"description\":{\"description\":\"Optional few-word summary shown to the user beside the command.\",\"type\":\"string\"},\"justification\":{\"description\":\"Required with sandbox_permissions: one sentence explaining why this exact command needs wider access.\",\"type\":\"string\"},\"read_only\":{\"description\":\"Set true to run analysis code (including Python/SQLite) or a network fetch (curl) with mandatory native filesysteLine truncated
 {"event":"turn_complete","parent_route_usage":{"input_tokens":350,"output_tokens":26},"routed_usage_dropped_records":0,"status":"completed","tool_catalog":"<pinned by the prompt family>","turn_id":"<uuid:1>","usage":{"input_tokens":350,"output_tokens":26}}
-{"harness_summary":{"model_requests":2,"non_streaming_requests":0,"workspace_after":{".codewhale/state/.gitignore":"cdbcae15105d6b78",".codewhale/state/subagents.v1.json/legacy.json":"9064f16acf90984c",".codewhale/state/subagents.v3.json":"e874da2951e53d2b",".codewhale/state/subagents.v3.transaction.lock":"e3b0c44298fc1c14","NOTES.md":"f957b19529906961","README.md":"41f49bffb38de5ba"}}}
+{"harness_summary":{"model_requests":2,"non_streaming_requests":0,"workspace_after":{".codewhale/state/.gitignore":"cdbcae15105d6b78",".codewhale/state/subagents.v1.json/legacy.json":"9064f16acf90984c",".codewhale/state/subagents.v1.lock":"e3b0c44298fc1c14",".codewhale/state/subagents.v3.json":"e874da2951e53d2b",".codewhale/state/subagents.v3.transaction.lock":"e3b0c44298fc1c14","NOTES.md":"f957b19529906961","README.md":"41f49bffb38de5ba"}}}

crates/tui/tests/fixtures/conformance/events/parallel_tool_calls.golden.jsonl

@@ -31,4 +31,4 @@
 {"engine_session_id":"conformance-session","event":"session_updated","instruction_assembly":{"budgets":{"claude_fallback":false,"file_max_bytes":8388608,"global_tokens":100000,"policy":"trim","project_tokens":100000,"static_guard_tokens":667571,"usable_window_tokens":834464},"constitution_source_path":null,"constitution_tokens":0,"global_tokens":0,"notices":[{"action":"Kept","bytes_dropped":0,"bytes_kept":514,"detail":null,"fallback_selected":null,"limit":"None","origin":"Instructions","path":null,"scope":"Project","tokens_dropped":0,"tokens_kept":172}],"project_tokens":172,"retained_block_tokens":193,"source_path":null,"warnings":[]},"messages":[{"content":[{"text":"Read README.md and NOTES.md.","type":"text"},{"text":"<turn_meta>\nCurrent local date: <today>\nCurrent workspace: <WORKSPACE>\nCurrent permission posture: Ask\nCurrent sandbox posture: workspace-write (writes inside the workspace; network blocked) (local OS sandbox applied)\nEstimated conversation input: 1973 / 1000000 context tokens (0.2%). Automatic compaction is enabled.\n## Repo Working Set\nKey files: README.md\nActive paths (prioritize these):\n- NOTES.md (file)\n- README.md (file)\nWhen in doubt, use tools to verify and keep changes focused on the working set.\n</turn_meta>","type":"text"}],"role":"user"},{"content":[{"text":"<codewhale:runtime_event kind=\"project_instructions\" visibility=\"internal\">\n<system-reminder>\nThe following workspace instructions may be relevant to your work. Use them as guidance when applicable. More specific instructions take precedence over broader ones. They do not override system, developer, or direct user instructions. Instructions from: AGENTS.md\n\n<project_instructions source=\"project\">\n# Project Context (Auto-generated, ephemeral)\n\n> This context was generated in memory by Codewhale.\n> No .codewhale/instructions.md file was written.\n\n## Bounded Project Overview\n\n```json\n{\n  \"project_name\": \"workspace\",\n  \"directory_structure\": [\n    \"README.md\",\n    \"NOTES.md\"\n  ],\n  \"readme\": {\n    \"path\": \"README.md\",\n    \"excerpt\": \"conformance fixture\"\n  },\n  \"config_files\": [],\n  \"key_source_files\": [],\n  \"counts\": {\n    \"config_files\": 0,\n    \"directory_entries\": 2,\n    \"key_source_files\": 0\n  }\n}\n```\n</project_instructions>\n</system-reminder>\n</codewhale:runtime_event>","type":"text"},{"text":"<turn_meta>\nInput provenance: runtime (non-authoritative)\n</turn_meta>","type":"text"}],"role":"user"},{"content":[{"execution_id":"<uuid:2>","id":"call_read_a","input":{"action":"read","path":"README.md"},"name":"File","type":"tool_use"},{"execution_id":"<uuid:3>","id":"call_read_b","input":{"action":"read","path":"NOTES.md"},"name":"File","type":"tool_use"}],"role":"assistant"},{"content":[{"content":"content_hash=\"sha256:41f49bffb38de5ba88945f09121ffc590f9d7f6ff80fdb9ea8ad10224fecea61\"\nconformance fixture","execution_id":"<uuid:2>","tool_use_id":"call_read_a","type":"tool_result"}],"role":"user"},{"content":[{"content":"content_hash=\"sha256:f957b19529906961933c5c30f8713c500a9bb5d9d0695c40d48c97a26a3594ec\"\nsecond file","execution_id":"<uuid:3>","tool_use_id":"call_read_b","type":"tool_result"}],"role":"user"},{"content":[{"text":"Both files read.","type":"text"}],"role":"assistant"}],"model":"deepseek-v4-pro","system_prompt":"<pinned by the prompt family>","workspace":"<WORKSPACE>"}
 {"event":"tool_request_snapshot","snapshot":{"active_tool_catalog_sha256":"9017c53ee026ae4065bed3e4b2e7d773a828556b42d6ebe2c2ebfd8e65377280","capture_source":"prepared model-client request","delivery_status":"unknown (capture does not prove provider delivery)","omitted_tool_count":0,"payload_json_bytes":17394,"payload_measurement_status":"exact (within 1048576-byte measurement bound)","provider":{"model":"deepseek-v4-pro","provider":"Deepseek","status":"available"},"registry_facts_present":true,"registry_only_tools":{"status":"known","value":{"count":40,"omitted":8,"rendered":[{"truncated":false,"value":"Git"},{"truncated":false,"value":"Run"},{"truncated":false,"value":"Web"},{"truncated":false,"value":"apply_patch"},{"truncated":false,"value":"automation"},{"truncated":false,"value":"context_archive_get"},{"truncated":false,"value":"context_archive_search"},{"truncated":false,"value":"diagnostics"},{"truncated":false,"value":"finance"},{"truncated":false,"value":"github"},{"truncated":false,"value":"handle_read"},{"truncated":false,"value":"harness"},{"truncated":false,"value":"image_ocr"},{"truncated":false,"value":"list_dir"},{"truncated":false,"value":"lsp"},{"truncated":false,"value":"note"},{"truncated":false,"value":"notify"},{"truncated":false,"value":"pandoc_convert"},{"truncated":false,"value":"project_map"},{"truncated":false,"value":"registry_sync"},{"truncated":false,"value":"request_plugin_install"},{"truncated":false,"value":"request_user_input"},{"truncated":false,"value":"retrieve_tool_result"},{"truncated":false,"value":"revert_turn"},{"truncated":false,"value":"review"},{"truncated":false,"value":"send_later"},{"truncated":false,"value":"session_get"},{"truncated":false,"value":"session_search"},{"truncated":false,"value":"start_mcp_server"},{"truncated":false,"value":"start_registry_mcp_server"},{"truncated":false,"value":"tasks"},{"truncated":false,"value":"terminal/cancel"}]}},"registry_tool_count":{"status":"known","value":65},"rendered_tool_count":15,"schema_version":1,"step":1,"terminal":{"automatic_compaction_attempts":0,"effective_max_steps":null,"emergency_compaction_attempts":0,"empty_stop_retries":0,"final_report_requested":false,"last_prepared_output_limit_tokens":65536,"last_provider_finish_reason":{"truncated":false,"value":"end_turn"},"last_reported_input_tokens":200,"last_response_tool_calls":0,"last_response_tool_calls_suppressed":0,"model_requests_started":2,"model_step_index":1,"permission_denial_rounds_without_progress":0,"permission_strategy_switches":0,"reason":"provider_no_tool_call","reasoning_only_reprompts":0,"route_context_window_tokens":1000000,"status":"completed","step_budget_source":"max_steps","stream_resumes":0,"transparent_stream_retries":0,"transport_retries":0},"tool_count":15,"tools":[{"allowed_callers":{"status":"known","value":{"count":1,"omitted":0,"rendered":[{"truncated":false,"value":"direct"}]}},"approval":{"status":"known","value":{"truncated":false,"value":"Required"}},"cache_control_type":{"reason":"request field absent","status":"unknown"},"capabilities":{"status":"known","value":{"count":3,"omitted":0,"rendered":[{"truncated":false,"value":"ExecutesCode"},{"truncated":false,"value":"Sandboxable"},{"truncated":false,"value":"RequiresApproval"}]}},"defer_loading":{"status":"known","value":false},"description":{"truncated":true,"value":"Execute a shell command in the workspace and return stdout and stderr (last 2000 lines or 50KB). Optional timeout is foreground seconds (default 30 seconds); a command still running then moves to the background by itself — never killed for time — returning output so far, last-output age, build phase, and a task_id. Call task_shell_wait with that task_id to read more or wait; do not sleep-poll with bash. Before a final answer, collect every still-relevant background task (wait only when genuinely blocked) an"},"input_examples":{"reason":"request field absent","status":"unknown"},"input_schema_json":{"truncated":false,"value":"{\"additionalProperties\":false,\"properties\":{\"command\":{\"description\":\"The command to execute. Actual execution shell: `/bin/bash`. Use Bash syntax: pipelines, redirections, $(command), and && / || are supported. Quote paths and variable expansions, such as \\\"$path\\\"; use single quotes for literal text. For literal multiline input, use a quoted heredoc delimiter (<<'EOF') with its closing delimiter on a separate line. Use only installed programs; do not assume GNU-specific flags on macOS/BSD. Example: printf '%s\\\\n' 'sample'.\",\"type\":\"string\"},\"description\":{\"description\":\"Optional few-word summary shown to the user beside the command.\",\"type\":\"string\"},\"justification\":{\"description\":\"Required with sandbox_permissions: one sentence explaining why this exact command needs wider access.\",\"type\":\"string\"},\"read_only\":{\"description\":\"Set true to run analysis code (including Python/SQLite) or a network fetch (curl) with mandatory native filesysteLine truncated
 {"event":"turn_complete","parent_route_usage":{"input_tokens":360,"output_tokens":46},"routed_usage_dropped_records":0,"status":"completed","tool_catalog":"<pinned by the prompt family>","turn_id":"<uuid:1>","usage":{"input_tokens":360,"output_tokens":46}}
-{"harness_summary":{"model_requests":2,"non_streaming_requests":0,"workspace_after":{".codewhale/state/.gitignore":"cdbcae15105d6b78",".codewhale/state/subagents.v1.json/legacy.json":"9064f16acf90984c",".codewhale/state/subagents.v3.json":"e874da2951e53d2b",".codewhale/state/subagents.v3.transaction.lock":"e3b0c44298fc1c14","NOTES.md":"f957b19529906961","README.md":"41f49bffb38de5ba"}}}
+{"harness_summary":{"model_requests":2,"non_streaming_requests":0,"workspace_after":{".codewhale/state/.gitignore":"cdbcae15105d6b78",".codewhale/state/subagents.v1.json/legacy.json":"9064f16acf90984c",".codewhale/state/subagents.v1.lock":"e3b0c44298fc1c14",".codewhale/state/subagents.v3.json":"e874da2951e53d2b",".codewhale/state/subagents.v3.transaction.lock":"e3b0c44298fc1c14","NOTES.md":"f957b19529906961","README.md":"41f49bffb38de5ba"}}}

crates/tui/tests/fixtures/conformance/events/plain_answer.golden.jsonl

@@ -15,4 +15,4 @@
 {"engine_session_id":"conformance-session","event":"session_updated","instruction_assembly":{"budgets":{"claude_fallback":false,"file_max_bytes":8388608,"global_tokens":100000,"policy":"trim","project_tokens":100000,"static_guard_tokens":667571,"usable_window_tokens":834464},"constitution_source_path":null,"constitution_tokens":0,"global_tokens":0,"notices":[{"action":"Kept","bytes_dropped":0,"bytes_kept":514,"detail":null,"fallback_selected":null,"limit":"None","origin":"Instructions","path":null,"scope":"Project","tokens_dropped":0,"tokens_kept":172}],"project_tokens":172,"retained_block_tokens":193,"source_path":null,"warnings":[]},"messages":[{"content":[{"text":"Say hello.","type":"text"},{"text":"<turn_meta>\nCurrent local date: <today>\nCurrent workspace: <WORKSPACE>\nCurrent permission posture: Ask\nCurrent sandbox posture: workspace-write (writes inside the workspace; network blocked) (local OS sandbox applied)\nEstimated conversation input: 1968 / 1000000 context tokens (0.2%). Automatic compaction is enabled.\n## Repo Working Set\nKey files: README.md\nWhen in doubt, use tools to verify and keep changes focused on the working set.\n</turn_meta>","type":"text"}],"role":"user"},{"content":[{"text":"<codewhale:runtime_event kind=\"project_instructions\" visibility=\"internal\">\n<system-reminder>\nThe following workspace instructions may be relevant to your work. Use them as guidance when applicable. More specific instructions take precedence over broader ones. They do not override system, developer, or direct user instructions. Instructions from: AGENTS.md\n\n<project_instructions source=\"project\">\n# Project Context (Auto-generated, ephemeral)\n\n> This context was generated in memory by Codewhale.\n> No .codewhale/instructions.md file was written.\n\n## Bounded Project Overview\n\n```json\n{\n  \"project_name\": \"workspace\",\n  \"directory_structure\": [\n    \"README.md\",\n    \"NOTES.md\"\n  ],\n  \"readme\": {\n    \"path\": \"README.md\",\n    \"excerpt\": \"conformance fixture\"\n  },\n  \"config_files\": [],\n  \"key_source_files\": [],\n  \"counts\": {\n    \"config_files\": 0,\n    \"directory_entries\": 2,\n    \"key_source_files\": 0\n  }\n}\n```\n</project_instructions>\n</system-reminder>\n</codewhale:runtime_event>","type":"text"},{"text":"<turn_meta>\nInput provenance: runtime (non-authoritative)\n</turn_meta>","type":"text"}],"role":"user"},{"content":[{"text":"Hello there.","type":"text"}],"role":"assistant"}],"model":"deepseek-v4-pro","system_prompt":"<pinned by the prompt family>","workspace":"<WORKSPACE>"}
 {"event":"tool_request_snapshot","snapshot":{"active_tool_catalog_sha256":"9017c53ee026ae4065bed3e4b2e7d773a828556b42d6ebe2c2ebfd8e65377280","capture_source":"prepared model-client request","delivery_status":"unknown (capture does not prove provider delivery)","omitted_tool_count":0,"payload_json_bytes":17394,"payload_measurement_status":"exact (within 1048576-byte measurement bound)","provider":{"model":"deepseek-v4-pro","provider":"Deepseek","status":"available"},"registry_facts_present":true,"registry_only_tools":{"status":"known","value":{"count":40,"omitted":8,"rendered":[{"truncated":false,"value":"Git"},{"truncated":false,"value":"Run"},{"truncated":false,"value":"Web"},{"truncated":false,"value":"apply_patch"},{"truncated":false,"value":"automation"},{"truncated":false,"value":"context_archive_get"},{"truncated":false,"value":"context_archive_search"},{"truncated":false,"value":"diagnostics"},{"truncated":false,"value":"finance"},{"truncated":false,"value":"github"},{"truncated":false,"value":"handle_read"},{"truncated":false,"value":"harness"},{"truncated":false,"value":"image_ocr"},{"truncated":false,"value":"list_dir"},{"truncated":false,"value":"lsp"},{"truncated":false,"value":"note"},{"truncated":false,"value":"notify"},{"truncated":false,"value":"pandoc_convert"},{"truncated":false,"value":"project_map"},{"truncated":false,"value":"registry_sync"},{"truncated":false,"value":"request_plugin_install"},{"truncated":false,"value":"request_user_input"},{"truncated":false,"value":"retrieve_tool_result"},{"truncated":false,"value":"revert_turn"},{"truncated":false,"value":"review"},{"truncated":false,"value":"send_later"},{"truncated":false,"value":"session_get"},{"truncated":false,"value":"session_search"},{"truncated":false,"value":"start_mcp_server"},{"truncated":false,"value":"start_registry_mcp_server"},{"truncated":false,"value":"tasks"},{"truncated":false,"value":"terminal/cancel"}]}},"registry_tool_count":{"status":"known","value":65},"rendered_tool_count":15,"schema_version":1,"step":0,"terminal":{"automatic_compaction_attempts":0,"effective_max_steps":null,"emergency_compaction_attempts":0,"empty_stop_retries":0,"final_report_requested":false,"last_prepared_output_limit_tokens":65536,"last_provider_finish_reason":{"truncated":false,"value":"end_turn"},"last_reported_input_tokens":120,"last_response_tool_calls":0,"last_response_tool_calls_suppressed":0,"model_requests_started":1,"model_step_index":0,"permission_denial_rounds_without_progress":0,"permission_strategy_switches":0,"reason":"provider_no_tool_call","reasoning_only_reprompts":0,"route_context_window_tokens":1000000,"status":"completed","step_budget_source":"max_steps","stream_resumes":0,"transparent_stream_retries":0,"transport_retries":0},"tool_count":15,"tools":[{"allowed_callers":{"status":"known","value":{"count":1,"omitted":0,"rendered":[{"truncated":false,"value":"direct"}]}},"approval":{"status":"known","value":{"truncated":false,"value":"Required"}},"cache_control_type":{"reason":"request field absent","status":"unknown"},"capabilities":{"status":"known","value":{"count":3,"omitted":0,"rendered":[{"truncated":false,"value":"ExecutesCode"},{"truncated":false,"value":"Sandboxable"},{"truncated":false,"value":"RequiresApproval"}]}},"defer_loading":{"status":"known","value":false},"description":{"truncated":true,"value":"Execute a shell command in the workspace and return stdout and stderr (last 2000 lines or 50KB). Optional timeout is foreground seconds (default 30 seconds); a command still running then moves to the background by itself — never killed for time — returning output so far, last-output age, build phase, and a task_id. Call task_shell_wait with that task_id to read more or wait; do not sleep-poll with bash. Before a final answer, collect every still-relevant background task (wait only when genuinely blocked) an"},"input_examples":{"reason":"request field absent","status":"unknown"},"input_schema_json":{"truncated":false,"value":"{\"additionalProperties\":false,\"properties\":{\"command\":{\"description\":\"The command to execute. Actual execution shell: `/bin/bash`. Use Bash syntax: pipelines, redirections, $(command), and && / || are supported. Quote paths and variable expansions, such as \\\"$path\\\"; use single quotes for literal text. For literal multiline input, use a quoted heredoc delimiter (<<'EOF') with its closing delimiter on a separate line. Use only installed programs; do not assume GNU-specific flags on macOS/BSD. Example: printf '%s\\\\n' 'sample'.\",\"type\":\"string\"},\"description\":{\"description\":\"Optional few-word summary shown to the user beside the command.\",\"type\":\"string\"},\"justification\":{\"description\":\"Required with sandbox_permissions: one sentence explaining why this exact command needs wider access.\",\"type\":\"string\"},\"read_only\":{\"description\":\"Set true to run analysis code (including Python/SQLite) or a network fetch (curl) with mandatory native filesysteLine truncated
 {"event":"turn_complete","parent_route_usage":{"input_tokens":120,"output_tokens":4},"routed_usage_dropped_records":0,"status":"completed","tool_catalog":"<pinned by the prompt family>","turn_id":"<uuid:1>","usage":{"input_tokens":120,"output_tokens":4}}
-{"harness_summary":{"model_requests":1,"non_streaming_requests":0,"workspace_after":{".codewhale/state/.gitignore":"cdbcae15105d6b78",".codewhale/state/subagents.v1.json/legacy.json":"9064f16acf90984c",".codewhale/state/subagents.v3.json":"e874da2951e53d2b",".codewhale/state/subagents.v3.transaction.lock":"e3b0c44298fc1c14","NOTES.md":"f957b19529906961","README.md":"41f49bffb38de5ba"}}}
+{"harness_summary":{"model_requests":1,"non_streaming_requests":0,"workspace_after":{".codewhale/state/.gitignore":"cdbcae15105d6b78",".codewhale/state/subagents.v1.json/legacy.json":"9064f16acf90984c",".codewhale/state/subagents.v1.lock":"e3b0c44298fc1c14",".codewhale/state/subagents.v3.json":"e874da2951e53d2b",".codewhale/state/subagents.v3.transaction.lock":"e3b0c44298fc1c14","NOTES.md":"f957b19529906961","README.md":"41f49bffb38de5ba"}}}

crates/tui/tests/fixtures/conformance/events/provider_error_after_tool_call.golden.jsonl

@@ -13,4 +13,4 @@
 {"event":"tool_call_complete","result":{"content":"Not executed: the provider stream failed before the model response completed (Model not exist.).","metadata":{"error_category":"model_stream_failed","model_output_incomplete":true,"side_effect_status":"not_started"},"outcome":"ok","success":false},"tool_call_id":"<uuid:2>","tool_name":"File"}
 {"event":"tool_request_snapshot","snapshot":{"active_tool_catalog_sha256":"9017c53ee026ae4065bed3e4b2e7d773a828556b42d6ebe2c2ebfd8e65377280","capture_source":"prepared model-client request","delivery_status":"unknown (capture does not prove provider delivery)","omitted_tool_count":0,"payload_json_bytes":17394,"payload_measurement_status":"exact (within 1048576-byte measurement bound)","provider":{"model":"deepseek-v4-pro","provider":"Deepseek","status":"available"},"registry_facts_present":true,"registry_only_tools":{"status":"known","value":{"count":40,"omitted":8,"rendered":[{"truncated":false,"value":"Git"},{"truncated":false,"value":"Run"},{"truncated":false,"value":"Web"},{"truncated":false,"value":"apply_patch"},{"truncated":false,"value":"automation"},{"truncated":false,"value":"context_archive_get"},{"truncated":false,"value":"context_archive_search"},{"truncated":false,"value":"diagnostics"},{"truncated":false,"value":"finance"},{"truncated":false,"value":"github"},{"truncated":false,"value":"handle_read"},{"truncated":false,"value":"harness"},{"truncated":false,"value":"image_ocr"},{"truncated":false,"value":"list_dir"},{"truncated":false,"value":"lsp"},{"truncated":false,"value":"note"},{"truncated":false,"value":"notify"},{"truncated":false,"value":"pandoc_convert"},{"truncated":false,"value":"project_map"},{"truncated":false,"value":"registry_sync"},{"truncated":false,"value":"request_plugin_install"},{"truncated":false,"value":"request_user_input"},{"truncated":false,"value":"retrieve_tool_result"},{"truncated":false,"value":"revert_turn"},{"truncated":false,"value":"review"},{"truncated":false,"value":"send_later"},{"truncated":false,"value":"session_get"},{"truncated":false,"value":"session_search"},{"truncated":false,"value":"start_mcp_server"},{"truncated":false,"value":"start_registry_mcp_server"},{"truncated":false,"value":"tasks"},{"truncated":false,"value":"terminal/cancel"}]}},"registry_tool_count":{"status":"known","value":65},"rendered_tool_count":15,"schema_version":1,"step":0,"terminal":{"automatic_compaction_attempts":0,"effective_max_steps":null,"emergency_compaction_attempts":0,"empty_stop_retries":0,"final_report_requested":false,"last_prepared_output_limit_tokens":65536,"last_provider_finish_reason":null,"last_reported_input_tokens":120,"last_response_tool_calls":1,"last_response_tool_calls_suppressed":1,"model_requests_started":1,"model_step_index":0,"permission_denial_rounds_without_progress":0,"permission_strategy_switches":0,"reason":"failed","reasoning_only_reprompts":0,"route_context_window_tokens":1000000,"status":"failed","step_budget_source":"max_steps","stream_resumes":0,"transparent_stream_retries":0,"transport_retries":0},"tool_count":15,"tools":[{"allowed_callers":{"status":"known","value":{"count":1,"omitted":0,"rendered":[{"truncated":false,"value":"direct"}]}},"approval":{"status":"known","value":{"truncated":false,"value":"Required"}},"cache_control_type":{"reason":"request field absent","status":"unknown"},"capabilities":{"status":"known","value":{"count":3,"omitted":0,"rendered":[{"truncated":false,"value":"ExecutesCode"},{"truncated":false,"value":"Sandboxable"},{"truncated":false,"value":"RequiresApproval"}]}},"defer_loading":{"status":"known","value":false},"description":{"truncated":true,"value":"Execute a shell command in the workspace and return stdout and stderr (last 2000 lines or 50KB). Optional timeout is foreground seconds (default 30 seconds); a command still running then moves to the background by itself — never killed for time — returning output so far, last-output age, build phase, and a task_id. Call task_shell_wait with that task_id to read more or wait; do not sleep-poll with bash. Before a final answer, collect every still-relevant background task (wait only when genuinely blocked) an"},"input_examples":{"reason":"request field absent","status":"unknown"},"input_schema_json":{"truncated":false,"value":"{\"additionalProperties\":false,\"properties\":{\"command\":{\"description\":\"The command to execute. Actual execution shell: `/bin/bash`. Use Bash syntax: pipelines, redirections, $(command), and && / || are supported. Quote paths and variable expansions, such as \\\"$path\\\"; use single quotes for literal text. For literal multiline input, use a quoted heredoc delimiter (<<'EOF') with its closing delimiter on a separate line. Use only installed programs; do not assume GNU-specific flags on macOS/BSD. Example: printf '%s\\\\n' 'sample'.\",\"type\":\"string\"},\"description\":{\"description\":\"Optional few-word summary shown to the user beside the command.\",\"type\":\"string\"},\"justification\":{\"description\":\"Required with sandbox_permissions: one sentence explaining why this exact command needs wider access.\",\"type\":\"string\"},\"read_only\":{\"description\":\"Set true to run analysis code (including Python/SQLite) or a network fetch (curl) with mandatory native filesystem read-only isolation. Outbound network is on only iLine truncated
 {"error":"Model not exist.","event":"turn_complete","parent_route_usage":{"input_tokens":120,"output_tokens":0},"routed_usage_dropped_records":0,"status":"failed","tool_catalog":"<pinned by the prompt family>","turn_id":"<uuid:1>","usage":{"input_tokens":120,"output_tokens":0}}
-{"harness_summary":{"model_requests":1,"non_streaming_requests":0,"workspace_after":{".codewhale/state/.gitignore":"cdbcae15105d6b78",".codewhale/state/subagents.v1.json/legacy.json":"9064f16acf90984c",".codewhale/state/subagents.v3.json":"e874da2951e53d2b",".codewhale/state/subagents.v3.transaction.lock":"e3b0c44298fc1c14","NOTES.md":"f957b19529906961","README.md":"41f49bffb38de5ba"}}}
+{"harness_summary":{"model_requests":1,"non_streaming_requests":0,"workspace_after":{".codewhale/state/.gitignore":"cdbcae15105d6b78",".codewhale/state/subagents.v1.json/legacy.json":"9064f16acf90984c",".codewhale/state/subagents.v1.lock":"e3b0c44298fc1c14",".codewhale/state/subagents.v3.json":"e874da2951e53d2b",".codewhale/state/subagents.v3.transaction.lock":"e3b0c44298fc1c14","NOTES.md":"f957b19529906961","README.md":"41f49bffb38de5ba"}}}

docs/CONFIGURATION.md

@@ -2230,10 +2230,7 @@ Common settings keys:
   `@`-mention completions are populated. `fuzzy` searches the workspace and
   applies mention frecency. `browser` lists only the immediate children of the
   currently typed directory segment in deterministic alphabetical order.
-- `show_thinking` (on/off, default on): shows model thinking in the transcript,
-  live while it streams and collapsed under a `Thought for …` header once the
-  answer starts. Off hides it; answers and tools stay visible. A value already
-  saved in `settings.toml` is kept.
+- `show_thinking` (on/off)
 - `thinking_default_expanded` (on/off, default off): renders thinking blocks
   expanded initially when `show_thinking` is enabled. Space still toggles the
   selected block, and it decides only the blocks you have not touched: a block

docs/CONTRIBUTORS.md

@@ -44,6 +44,7 @@ changelog or release notes, and this contributor record.
 
 **Contributions**
 
+- **[MannXo](https://github.com/MannXo)** — added Qwen Code to the external agent registry so an installed Qwen Code CLI runs as a session over ACP ([#7293](https://github.com/codewhale-hq/Codewhale/pull/7293)).
 - **[SamhandsomeLee](https://github.com/SamhandsomeLee)** — drove Responses request, URL and stream behavior from one checked dialect profile table ([#7261](https://github.com/codewhale-hq/Codewhale/pull/7261)).
 - **[gaord](https://github.com/gaord)** — ran the read-only `/system`, `/tokens` and `/cost` builtins over the Runtime API through one gated terminal-host catalog ([#7277](https://github.com/codewhale-hq/Codewhale/pull/7277)).
 

docs/CONVERSATION_LIFECYCLE_CONTRACT_PROPOSAL.md

@@ -68,8 +68,8 @@ ids to customers.
 ## Subagent state
 
 Subagent state lives in the workspace, under `.codewhale/state/`, in
-`subagents.v3.json` (the manager's `SUBAGENT_STATE_FILE`), loaded by
-`load_state` and written by
+`subagents.v1.json` (the manager's `SUBAGENT_STATE_FILE`; the lock is
+`subagents.v1.lock`), loaded by `load_state` and written by
 `persist_state*` in `crates/engine/src/tools/subagent/mod.rs`. A legacy
 `.deepseek/state/` copy is still read for migration.
 

docs/ENVIRONMENTS.md

@@ -161,6 +161,17 @@ separately from the #6953 reporter-observed exports; the x64 profile now allows
 468 pairs. The candidate still has to pass the import check against its final
 executable bytes, and neither addition is a 14393 or installer run.
 
+Release candidate run 43186389246 on source `f5a3dda975` then stopped the same
+check on three more pairs, `GetSidIdentifierAuthority`, `GetSidSubAuthority` and
+`GetSidSubAuthorityCount` from advapi32.dll, on both 64-bit targets. Probe run
+43197611105 (Windows 11 ARM64 build 26200.9725) resolved all 447 imports of
+that candidate's ARM64 executables, and the 3 pairs were added. For x64 the 3
+pairs are documented Microsoft API contracts (Windows XP minimum on each
+function's reference page), labelled apart from the #6953 reporter-observed
+exports; the x64 profile now allows 471 pairs and the ARM64 profile 473. The
+same limit applies: the candidate still has to pass the import check against
+its final executable bytes, and neither addition is a 14393 or installer run.
+
 At this snapshot, startup/installation of the final release bytes on an actual
 1607 machine and the candidate's generated uninstaller were still unqualified.
 The final frozen source, release import dumps, archive/native/npm installation,

docs/RELEASE_RUNBOOK.md

@@ -244,7 +244,30 @@ The manual `ci.yml` path verifies that the dispatch resolved to
 `expected_sha`, disables light-change shortcuts, and forces the heavy Rust,
 workflow, mobile, Actions, Linux, macOS, Windows, npm-wrapper, and documentation
 gates. A mismatch fails before those gates start; it never silently tests a
-different head.
+different head. The Linux, macOS and Windows suites run as `Sharded tests (<os>)`
+(`ci-test-suite.yml`): one test build per OS, parallel nextest shards of that
+archive, and a `Suite check` job that fails unless the shards together ran
+exactly the tests the archive lists; read that job, not only the shard jobs.
+
+`ci.yml` also takes optional scoped-run inputs for a quick answer from an agent
+or a maintainer. With `crate` (one workspace package, for example
+`codewhale-tui`) and/or `filter` (a nextest filterset such as
+`test(auto_review)`) set, only the `Scoped test` job runs, on the operating
+systems named by `os` (default `ubuntu-latest`; a comma-separated list or
+`all`), and the full suite and every other gate are skipped. `expected_sha`
+stays required and is still verified:
+
+```bash
+gh workflow run ci.yml --ref ci/my-task \
+  -f expected_sha="$(git rev-parse HEAD)" \
+  -f crate=codewhale-tui -f filter='test(auto_review)'
+```
+
+A scoped run proves a slice, never a release; release evidence is a full
+dispatch with the scoped inputs empty. A newer dispatch on the same non-main
+branch, and a newer push to a pull request, cancels the older run of the same
+workflow; runs on `main` are keyed by SHA and never cancel each other, and a
+scoped run never cancels a full one.
 
 Verify beforehand that clean local `HEAD`, `main` and `origin/main` agree.
 Record both run URLs, resolved SHAs and required job results; inspect executed

docs/RUNTIME_API.md

@@ -3270,7 +3270,7 @@ refuses OpenAI Codex. An Engine without this route answers `404`.
 
 Reports which external agent CLIs this computer already has, without starting a
 session. It covers Claude, Codex, DeepSeek Harness, Antigravity (`agy acp`),
-OpenCode, Kimi Code, Mistral Vibe, Qoder, CodeBuddy and Grok. `state` is `installed` or `not_installed`; a binary is
+OpenCode, Kimi Code, Mistral Vibe, Qoder, CodeBuddy, Grok and Qwen Code. `state` is `installed` or `not_installed`; a binary is
 looked up on `PATH` (absolute entries only; on Windows the spawnable PATHEXT suffixes `.com/.exe/.bat/.cmd` and `%APPDATA%\npm`) and in the usual per-user bin directories. For an installed
 agent the route runs `--version` and, where the CLI speaks Agent Client Protocol
 itself, one `initialize` request on stdin, concurrently. `acp` is `native` when

docs/architecture/delegated-coordination.md

@@ -29,7 +29,7 @@ resolved values are persisted in their launch manifest.
 
 By default, delegated control-plane state remains workspace-scoped at
 `<workspace>/.codewhale/state`: the worker ledger, complete transcript
-artifacts, and store locks share that root. An embedding host may set
+artifacts, and coordination lock share that root. An embedding host may set
 `EngineConfig::subagent_state_root` to keep those files under a session-owned
 root without changing child cwd, tool path authority, or the execution
 workspace recorded in receipts.
@@ -124,32 +124,19 @@ cannot replace a newer one, and a terminal receipt cannot be resurrected as
 Running by a checkpoint. An explicit terminal retry accepts only the same
 winning outcome; authoritative newer accounting metadata is retained.
 
-There is no workspace-wide session lock. Interactive, exec, Fleet, background
-and Runtime managers in one folder are ordinary neighbours: each holds only its
-own boot lease and takes the transaction lock per operation. The Runtime
-constructor keeps one manager per state path inside its process. No manager
-holds the transaction lock while a child executes or while it waits for a
-provider or channel capacity. Conflicting lock contention returns a bounded
-retry error. Unsupported locking or failed store admission refuses durable
-writes rather than silently creating an empty writable ledger, and admission is
-retried on the next delegated write instead of staying refused for the life of
-the process.
-
-A Runtime that admits a selected workspace other than its own state root keeps
-that workspace's claims in its own store. Before it admits a shared-workspace
-writer there, and each time it leases such a writer, it refuses while any
-session holds a live boot lease in the selected workspace's own store, and it
-refuses when that store's directory cannot be listed. Only sessions that hold
-a boot lease are seen: a Codewhale 0.10.2 or older session holds none. A
-session that opens in that workspace later does not see the Runtime's claims,
-and its lease refuses the Runtime's next shared-workspace writer there; give
-writers isolated worktrees when both are expected.
+Normal interactive/exec/Fleet/background sessions continue to share the
+existing admission flock. The explicit exclusive Runtime constructor retains
+its host admission policy; it is not the mechanism that conserves shared
+snapshots. No manager holds the transaction lock while a child executes or
+while it waits for a provider or channel capacity. Conflicting lock contention
+returns a bounded retry error. Unsupported locking or failed store admission
+refuses durable writes rather than silently creating an empty writable ledger.
 
 Each writer holds its own boot lease. Other managers preserve its live records
 and use its active claims during admission. An existing unlocked lease can
 establish that its writer ended; a missing, invalid or unreadable lease is
-uncertain and fails closed. Local task handles do not establish that another
-process's work stopped. Parked and headless
+uncertain and fails closed. Local task handles and a failed admission flock do
+not establish that another process's work stopped. Parked and headless
 WaitingForUser claim rules retain their existing lifecycle meaning; they do not
 transfer a still-live writer's record ownership to a different boot.
 
@@ -229,15 +216,10 @@ terminal receipt or claim that a remote effect was undone.
 
 ### Legacy cutover and recovery
 
-The first current-format admission migrates under the short transaction lock.
-It does not wait for older sessions that still hold the retired
-`subagents.v1.lock`. It retains the original
+The first current-format admission takes the legacy admission lock exclusively
+only for migration, after previous holders have left. It retains the original
 legacy bytes and a validated recovery copy, then leaves
 `subagents.v1.json` as a private directory containing that recovery material.
-An older session still open in that folder can no longer save its delegated
-state once the directory replaces its file; restart it. If it publishes its
-file between the two cutover renames, the cutover stops at a retained-file
-refusal until a person inspects both files; nothing is deleted.
 The current format3 file is published under the same short transaction lock.
 Interrupted cutover states are validated and resumed without replacing a later
 current-format revision with the old backup. Future or malformed envelopes

docs/features.toml

@@ -805,7 +805,7 @@ owner = "crates/tui/src/commands/groups/session/export.rs"
 [[feature]]
 id = "live-thinking"
 name = "Live thinking"
-summary = "Model thinking shows live in the transcript, then collapses under a Thought header; /verbose shows it in full."
+summary = "/verbose toggles live model thinking in the transcript."
 status = "stable"
 surfaces = { tui = "stable" }
 owner = "crates/tui/src/commands/groups/config/mod.rs"

docs/skills/cw-land/SKILL.md

@@ -97,8 +97,13 @@ Stage 5 of the loop: [cw-orient](../cw-orient/SKILL.md) →
 
    Co-authored-by: <maintainer or agent identity>
    ```
-   GitHub's contributor list counts authors only, and
-   `check-contributor-credit.py` flags a harvested handle who is not credited.
+   GitHub's contributor count takes authors and `Co-authored-by` trailers in
+   the final trailer block (no blank line between trailers) whose address is
+   linked to the account, and `check-contributor-credit.py` flags a harvested
+   handle who is not credited, a trailer GitHub cannot parse, and a noreply id
+   that belongs to someone else. It accepts credit only from a noreply address
+   whose id matches `.github/AUTHOR_MAP`, or from a raw email listed there, so
+   add the contributor's line to `AUTHOR_MAP` when it is missing.
    Resolve the address from `.github/AUTHOR_MAP`, else
    `gh api users/<handle> --jq '"\(.id)+\(.login)@users.noreply.github.com"'`.
    A clean cherry-pick keeps the author by itself. This step is the single

docs/zh_hans/CONFIGURATION.md

@@ -1152,7 +1152,7 @@ codewhale 还把用户偏好存储在：
 - `mention_menu_limit`(整数，默认 `128`)：输入区渲染可见窗口之前保留的 `@`-提及弹出候选最大数。可见行仍取决于终端高度。
 - `mention_walk_depth`(整数，默认 `10`)：`@`-提及补全遍历的最大工作区深度。在深度嵌套的工作区设为 `0` 表示无限深度；在非常大的仓库保持默认，除非需要。
 - `mention_menu_behavior`(`fuzzy`、`browser`；默认 `fuzzy`)：控制 `@`-提及补全如何填充。`fuzzy` 搜索工作区并应用提及 frecency。`browser` 只按确定性字母顺序列出当前输入目录段的直接子项。
-- `show_thinking`（on/off，默认 on）：在对话记录中显示模型思考，流式输出时实时显示，回答开始后折叠到“思考用时 …”标题下。设为 off 则隐藏思考；回答和工具仍然可见。`settings.toml` 中已保存的值保持不变。
+- `show_thinking`(on/off)
 - `thinking_default_expanded`（on/off，默认 off）：启用 `show_thinking` 时，让思考块初始以展开状态渲染。Space 仍可切换所选块，并且它只决定你尚未动过的块：你自己展开或折叠过的块，即使之后更改这个设置，也会保持该状态。这在 Space 绑定可能被拦截的 SSH/tmux 环境中很有用。
 - `thinking_preview_lines`(整数，默认 `2`):**折叠**的已完成思考仍显示多少正文行。`0` 仅头部；`10` 是较旧的完整显示。实时流式预览不变。用 Space 展开块，或设置 `thinking_default_expanded` 打开每个块。
 - `help_expand_groups`(on/off，默认 off)：让帮助/快捷键从每个组展开开始。默认折叠长尾(Grok 风格)；输入过滤仍展开匹配。

docs/zh_hans/ENVIRONMENTS.md

@@ -127,6 +127,11 @@ DLL 的身份一致，[该 API 的文档前提是 Vista](https://learn.microsoft
 x64 没有导出探测，因此这 12 对作为低于 14393 下限的微软文档 API 契约单独记录，与 #6953 报告者实测导出分开标注；x64 允许列表现为 468 对。
 候选仍须以最终可执行文件通过导入检查，两项新增都不是 14393 或安装器运行。
 
+源 `f5a3dda975` 的发布候选运行 43186389246 随后在两个 64 位目标上又停止于三对导入：advapi32.dll 的 `GetSidIdentifierAuthority`、`GetSidSubAuthority` 和 `GetSidSubAuthorityCount`。
+探测运行 43197611105（Windows 11 ARM64 内部版本 26200.9725）解析了该候选 ARM64 程序的全部 447 个导入，新增的 3 对已加入基线。
+x64 的这 3 对作为微软文档 API 契约记录（各函数参考页的最低客户端为 Windows XP），与 #6953 报告者实测导出分开标注；x64 允许列表现为 471 对，ARM64 为 473 对。
+限制不变：候选仍须以最终可执行文件通过导入检查，两项新增都不是 14393 或安装器运行。
+
 此快照中，最终发布文件在真实 1607 上启动/安装及候选程序生成的卸载器仍未验收。
 最终冻结源码、发布导入记录、各安装途径、打包/原生应用崩溃恢复和最低 OS 回执应
 进入发布验收记录，见[英文发布检查单](../RELEASE_CHECKLIST.md)。

scripts/check-blocking-calls-budget.json

@@ -257,7 +257,7 @@
   },
   "crates/engine/src/tools/subagent/mod.rs": {
     "path_canonicalize": 4,
-    "std_fs": 2,
+    "std_fs": 3,
     "thread_sleep": 1
   },
   "crates/engine/src/tools/subagent/worktree.rs": {

scripts/check-contributor-credit.py

@@ -17,7 +17,7 @@ live in the commit bodies, so no network call is needed.
 Usage: scripts/check-contributor-credit.py [<since-rev>]
 """
 from __future__ import annotations
-import re, subprocess, sys
+import functools, re, subprocess, sys
 from pathlib import Path
 
 ROOT = Path(__file__).resolve().parent.parent
@@ -35,6 +35,7 @@ def sh(*args: str) -> str:
     return subprocess.run(args, cwd=ROOT, capture_output=True, text=True, check=True).stdout
 
 
+@functools.cache
 def author_map() -> dict[str, str]:
     """alias -> github handle, from .github/AUTHOR_MAP."""
     out: dict[str, str] = {}
@@ -52,19 +53,43 @@ def author_map() -> dict[str, str]:
     return out
 
 
+NOREPLY = re.compile(r"\d+\+([^@<>\s]+)@users\.noreply\.github\.com")
+
+
+def not_a_person(name: str, email: str) -> bool:
+    low, addr = name.strip().lower(), email.lower()
+    return any(low.startswith(skip) for skip in SKIP) or "[bot]" in addr \
+        or addr.endswith(("[email protected]", "[email protected]"))
+
+
+def maintainer(name: str, email: str) -> bool:
+    handle = author_map().get(email.lower()) or author_map().get(name.strip().lower()) or name.strip()
+    return any(h.lower().startswith(skip) for skip in SKIP for h in (name.strip(), handle))
+
+
+def restoring(subject: str, name: str, email: str) -> bool:
+    return subject.startswith("chore(credit):") and maintainer(name, email)
+
+
+def noreply_ids() -> dict[str, str]:
+    out: dict[str, str] = {}
+    path = ROOT / ".github" / "AUTHOR_MAP"
+    if path.exists():
+        for login_id, login in re.findall(r"(\d+)\+([^@<>\s]+)@users\.noreply\.github\.com", path.read_text()):
+            out[login.lower()] = login_id
+    return out
+
+
 def expected(since: str) -> dict[str, str]:
     """handle -> why we think they are owed credit."""
     amap = author_map()
     found: dict[str, str] = {}
 
     def add(name: str, email: str, why: str) -> None:
-        low = name.strip().lower()
         # Model co-authorship ("Claude Opus 5 (1M context)") and bot addresses
         # are not people owed credit; match on prefix since the model name and
         # its context window change release to release.
-        if any(low.startswith(skip) for skip in SKIP):
-            return
-        if "[bot]" in email.lower() or email.lower().endswith(("[email protected]", "[email protected]")):
+        if not_a_person(name, email):
             return
         handle = amap.get(email.lower()) or amap.get(name.strip().lower())
         if not handle:
@@ -74,16 +99,17 @@ def expected(since: str) -> dict[str, str]:
             return
         found.setdefault(handle, why)
 
-    for line in sh("git", "log", f"{since}..HEAD", "--format=%an\t%ae").splitlines():
-        name, _, email = line.partition("\t")
-        add(name, email, "commit author")
-
-    body = sh("git", "log", f"{since}..HEAD", "--format=%b")
-    for name, email in re.findall(r"(?im)^co-authored-by:\s*(.+?)\s*<([^>]+)>", body):
-        add(name, email, "co-author trailer")
-    for pr, handle in re.findall(r"(?i)harvested from PR #(\d+) by @([A-Za-z0-9-]+)", body):
-        if handle.lower() not in SKIP:
-            found.setdefault(handle, f"harvested PR #{pr}")
+    log = sh("git", "log", f"{since}..HEAD", "--format=%an%x1f%ae%x1f%s%x1f%b%x00")
+    work = [parts for parts in (entry.lstrip("\n").split("\x1f", 3) for entry in log.split("\0"))
+            if len(parts) == 4 and not restoring(parts[2], parts[0], parts[1])]
+    for author, email, _, _ in work:
+        add(author, email, "commit author")
+    for _, _, _, body in work:
+        for name, addr in re.findall(r"(?im)^co-authored-by:\s*(.+?)\s*<([^>]+)>", body):
+            add(name, addr, "co-author trailer")
+        for pr, handle in re.findall(r"(?i)harvested from PR #(\d+) by @([A-Za-z0-9-]+)", body):
+            if handle.lower() not in SKIP:
+                found.setdefault(handle, f"harvested PR #{pr}")
     return found
 
 
@@ -108,17 +134,50 @@ def main() -> int:
 
     print(f"Contributor credit since {since}: {len(owed)} contributor(s) found.")
     amap = author_map()
+    ids = noreply_ids()
     lost = []
-    for entry in sh("git", "log", f"{since}..HEAD", "--format=%h\t%an\t%ae\t%B%x00").split("\0"):
-        parts = entry.strip().split("\t", 3)
-        if len(parts) < 4:
+    broken: list[tuple[str, str]] = []
+    credited: set[str] = set()
+
+    def login(name: str, addr: str) -> str:
+        nm = NOREPLY.search(addr)
+        return (amap.get(addr.lower()) or amap.get(name.strip().lower()) or (nm.group(1) if nm else name.strip())).lower()
+
+    def who(name: str, addr: str) -> str:
+        nm = re.search(r"(\d+)\+([^@<>\s]+)@users\.noreply\.github\.com", addr)
+        if not nm:
+            return amap.get(addr.lower(), "").lower() if "@" in addr else ""
+        known = ids.get(nm.group(2).lower())
+        if known is not None and known != nm.group(1):
+            broken.append((nm.group(2).lower(), f"uses {addr}, but @{nm.group(2)} is id {known} "
+                                                 f"in AUTHOR_MAP; GitHub credits whoever owns id {nm.group(1)}"))
+        return nm.group(2).lower() if known == nm.group(1) else ""
+
+    log = sh("git", "log", f"{since}..HEAD",
+             "--format=%h%x1f%an%x1f%ae%x1f%(trailers:key=Co-authored-by,unfold)%x1f%B%x00")
+    for entry in log.split("\0"):
+        parts = entry.strip().split("\x1f", 4)
+        if len(parts) < 5:
             continue
-        sha, name, email, body = parts
+        sha, name, email, parsed, body = parts
         m = re.search(r"\d+\+([^@<>\s]+)@users\.noreply\.github\.com", email)
         author = (amap.get(email.lower()) or amap.get(name.lower()) or (m.group(1) if m else name)).lower()
+        trailers = re.findall(r"(?im)^co-authored-by:\s*(.+?)\s*<([^>]+)>", parsed)
+        mark = len(broken)
+        credited.add(who(name, email))
+        credited.update(who(n, e) for n, e in trailers)
+        broken[mark:] = [(login, f"{sha} {why}") for login, why in broken[mark:]]
+        for co_name, co_email in re.findall(r"(?im)^\s*co-authored-by:\s*(.+?)\s*<([^>]+)>", body):
+            if (co_name, co_email) not in trailers and not not_a_person(co_name, co_email):
+                broken.append((login(co_name, co_email),
+                               f"{sha} has 'Co-authored-by: {co_name}' outside the final trailer block; GitHub ignores it"))
+        if restoring(body.split("\n", 1)[0], name, email):
+            continue
         for handle in re.findall(r"(?i)harvested from PR #\d+ by @([A-Za-z0-9-]+)", body):
             if handle.lower() != author:
                 lost.append(f"  {sha} harvested @{handle} but is authored by {name}")
+            broken.append((amap.get(handle.lower(), handle).lower(),
+                           f"{sha} harvested @{handle} with neither authorship nor a Co-authored-by trailer for them"))
         if any(author.startswith(skip) for skip in SKIP):
             for co_name, co_email in re.findall(r"(?im)^co-authored-by:\s*(.+?)\s*<([^>]+)>", body):
                 cm = re.search(r"\d+\+([^@<>\s]+)@users\.noreply\.github\.com", co_email)
@@ -128,8 +187,18 @@ def main() -> int:
                         and "[bot]" not in co_email.lower() and not co_email.lower().endswith(("[email protected]", "[email protected]")):
                     lost.append(f"  {sha} credits @{human} only as a co-author; authored by {name}")
     if lost:
-        print("\nCommits whose human contributor is not the commit author (GitHub's contributor list will miss them):\n")
+        print("\nCommits whose human contributor is not the commit author (GitHub counts them only through a\n"
+              "well-formed Co-authored-by trailer whose address is linked to their account):\n")
         print("\n".join(lost))
+    broken = [(login, why) for login, why in broken if login.lower() not in credited]
+    if broken:
+        print("\nCredit GitHub cannot see:\n")
+        print("\n".join(f"  {why}" for _, why in broken))
+        print("\nPut each contributor in the final trailer block as\n"
+              "'Co-authored-by: Name <[email protected]>', with no blank line between trailers,\n"
+              "using the id from `gh api users/<login>`. A later commit can carry the trailer when history cannot change.\n"
+              "Credit is taken only from such an address, or from a raw email listed in .github/AUTHOR_MAP; add the\n"
+              "email a contributor commits with there once it is linked to their account.")
     if missing:
         print("\nUncredited work landed in this release:\n")
         print("\n".join(missing))
@@ -139,6 +208,8 @@ def main() -> int:
             "is still their work -- see AGENTS.md 'Landing other people's work'."
         )
         return 1
+    if broken:
+        return 1
     print("Every contributor in the window is credited on all three surfaces.")
     return 0
 

scripts/check-provider-registry.py

@@ -538,6 +538,9 @@ def report_huggingface_coverage(
 def report_env_lookup_order(
     label: str, source: str, expected_order: list[str], context: str
 ) -> list[str]:
+    non_blank_chain = "env_non_blank(&[" + ", ".join(f'"{name}"' for name in expected_order) + "])"
+    if non_blank_chain in source:
+        return []
     lookup_needles = [f'std::env::var("{name}")' for name in expected_order]
     return report_string_order(label, source, lookup_needles, context)
 

scripts/ci/compiled-host-test-input.sh

@@ -0,0 +1,18 @@
+#!/usr/bin/env bash
+set -euo pipefail
+bun_binary="$(bun -p 'process.execPath')"
+host_binary="$(node -p "require('path').join(process.env.RUNNER_TEMP, 'codewhale-extension-host' + (process.platform === 'win32' ? '.exe' : ''))")"
+compiled_info="$(node crates/engine/extension-host/compile-host.mjs --bun "$bun_binary" --output "$host_binary")"
+printf 'CODEWHALE_COMPILED_HOST_TEST_BINARY=%s\n' "$host_binary" >> "$GITHUB_ENV"
+node - "$host_binary" "$bun_binary" "$compiled_info" >> "$GITHUB_ENV" <<'JS'
+const fs = require('node:fs'), crypto = require('node:crypto');
+const digest = (file) => crypto.createHash('sha256').update(fs.readFileSync(file)).digest('hex');
+const info = JSON.parse(process.argv[4]);
+if (!info.platform || !info.arch || info.platform !== info.compiler?.platform || info.arch !== info.compiler?.arch) throw new Error('image/compiler identity mismatch');
+console.log('CODEWHALE_COMPILED_HOST_TEST_PLATFORM=' + info.platform);
+console.log('CODEWHALE_COMPILED_HOST_TEST_ARCH=' + info.arch);
+console.log('CODEWHALE_COMPILED_HOST_TEST_SHA256=' + digest(process.argv[2]));
+console.log('CODEWHALE_COMPILED_HOST_TEST_BUN=' + process.argv[3]);
+console.log('CODEWHALE_COMPILED_HOST_TEST_RUNTIME_SHA256=' + digest(process.argv[3]));
+console.log('CODEWHALE_COMPILED_HOST_TEST_BUNDLE_SHA256=' + digest('crates/engine/extension-host/dist/codewhale-extension-host.mjs'));
+JS

scripts/ci/nextest-shard-count.py

@@ -0,0 +1,83 @@
+#!/usr/bin/env python3
+"""Count nextest tests so a sharded run can prove it ran every test once.
+
+  list LISTING_JSON         tests the archive selects (nextest list --message-format json)
+  junit JUNIT_XML           tests one nextest invocation ran
+  check DIRECTORY           per OS, the shard files must cover the expected total exactly
+
+The sharded CI suite splits one archive across jobs. A shard that is skipped,
+runs a different filter, or double-counts a partition changes the sum, so the
+suite gate fails instead of reporting a smaller run as a pass.
+"""
+from __future__ import annotations
+
+import json
+import sys
+import xml.etree.ElementTree as ET
+from pathlib import Path
+
+
+def listed(path: str) -> int:
+    suites = json.loads(Path(path).read_text(encoding="utf-8"))["rust-suites"]
+    return sum(
+        1
+        for suite in suites.values()
+        for case in suite["testcases"].values()
+        if case["filter-match"]["status"] == "matches"
+    )
+
+
+def ran(path: str) -> int:
+    return int(ET.parse(path).getroot().get("tests", "0"))
+
+
+def check(directory: str) -> int:
+    files = sorted(Path(directory).rglob("*.json"))
+    expected = {}
+    shards: dict[str, dict[int, int]] = {}
+    for file in files:
+        record = json.loads(file.read_text(encoding="utf-8"))
+        if file.name.startswith("expected-"):
+            expected[record["os"]] = record
+        elif file.name.startswith("shard-"):
+            slot = shards.setdefault(record["os"], {})
+            if record["shard"] in slot:
+                print(f"{record['os']}: shard {record['shard']} reported twice")
+                return 1
+            slot[record["shard"]] = record["tests"]
+    if not expected:
+        print("no expected-<os>.json found")
+        return 1
+    failed = False
+    for os_name, record in sorted(expected.items()):
+        got = shards.get(os_name, {})
+        want = set(range(1, record["shards"] + 1))
+        total = sum(got.values())
+        print(f"{os_name}: archive lists {record['tests']}, shards ran {total} {dict(sorted(got.items()))}")
+        if set(got) != want:
+            print(f"{os_name}: expected shards {sorted(want)}, found {sorted(got)}")
+            failed = True
+        if total != record["tests"]:
+            print(f"{os_name}: shard total {total} differs from the archive's {record['tests']}")
+            failed = True
+        if any(count <= 0 for count in got.values()):
+            print(f"{os_name}: a shard ran no tests")
+            failed = True
+    return 1 if failed else 0
+
+
+def main(argv: list[str]) -> int:
+    if len(argv) == 3 and argv[1] == "list":
+        print(listed(argv[2]))
+        return 0
+    if len(argv) == 3 and argv[1] == "junit":
+        print(ran(argv[2]))
+        return 0
+    if len(argv) == 3 and argv[1] == "check":
+        return check(argv[2])
+    print(__doc__, file=sys.stderr)
+    return 2
+
+
+if __name__ == "__main__":
+    raise SystemExit(main(sys.argv))

scripts/contributor-wall.py

@@ -1,10 +1,10 @@
 #!/usr/bin/env python3
 """Everyone whose work landed in or is credited by Codewhale, on one wall.
 
-GitHub's contributor list counts only commit authors whose email is linked to
-an account. Harvested pull requests, co-author trailers, unlinked commit
-emails and credited bug reports never reach it, so it shows about half of the
-people who built Codewhale.
+GitHub's contributor count takes only commit authors and co-author trailers
+whose email is linked to an account. Harvested pull requests without a trailer,
+unlinked commit emails and credited bug reports never reach it, so it shows
+about half of the people who built Codewhale.
 
   refresh  rebuild docs/contributors.json from git history, the CHANGELOG,
            docs/CONTRIBUTORS.md and GitHub (needs `gh`)
@@ -202,8 +202,8 @@ def render() -> None:
         "<!-- contributor-wall:start -->\n## Everyone\n\n"
         f"[![{total} people built Codewhale](assets/contributors.png)](#everyone)\n\n"
         f"{total} people have code, co-authored commits, carried-forward pull requests or credited reports in Codewhale. "
-        "GitHub's contributor list shows only commit authors whose email is linked to their account, so it counts "
-        f"{data['github_contributor_list']} of them. If your commits are missing there, add your commit email to your "
+        "GitHub's contributor count includes only commit authors and co-authors whose email is linked to their "
+        "account, so it misses many of them. If your commits are missing there, add your commit email to your "
         "GitHub account and they will link.\n\n"
         f"{links}\n\nAlso with us, from commits whose email is not linked to an account: {' · '.join(extra)}.\n"
         "<!-- contributor-wall:end -->\n"

scripts/release/windows-import-floor.json

@@ -6,7 +6,7 @@
       "osArchitecture": "x64",
       "qualification": {
         "result": "PASS",
-        "kind": "reporter exact-OS export resolution plus 12 documented Microsoft named API minima",
+        "kind": "reporter exact-OS export resolution plus 15 documented Microsoft named API minima",
         "evidence": {
           "issue": "https://github.com/codewhale-hq/Codewhale/issues/6953",
           "binary": "https://github.com/codewhale-hq/Codewhale/releases/download/v0.10.1/codewhale-windows-x64.exe",
@@ -16,7 +16,7 @@
           "previous_sha256": "f7b35a1e928f67884637f36b01fc72854f50690130f191bac3bcf62ece3c1d84"
         },
         "runtimeAcceptance": "Final 0.10.2 release bytes on 14393 not run",
-        "qualifiedPairs": 468,
+        "qualifiedPairs": 471,
         "namedApiContracts": [
           {
             "dll": "advapi32.dll",
@@ -89,6 +89,24 @@
             "symbol": "expf",
             "minimumClient": "Exact API-set pair introduced in 10.0.10240",
             "source": "https://learn.microsoft.com/en-us/uwp/win32-and-com/win32-extension-apis"
+          },
+          {
+            "dll": "advapi32.dll",
+            "symbol": "GetSidIdentifierAuthority",
+            "minimumClient": "Windows XP",
+            "source": "https://learn.microsoft.com/en-us/windows/win32/api/securitybaseapi/nf-securitybaseapi-getsididentifierauthority"
+          },
+          {
+            "dll": "advapi32.dll",
+            "symbol": "GetSidSubAuthority",
+            "minimumClient": "Windows XP",
+            "source": "https://learn.microsoft.com/en-us/windows/win32/api/securitybaseapi/nf-securitybaseapi-getsidsubauthority"
+          },
+          {
+            "dll": "advapi32.dll",
+            "symbol": "GetSidSubAuthorityCount",
+            "minimumClient": "Windows XP",
+            "source": "https://learn.microsoft.com/en-us/windows/win32/api/securitybaseapi/nf-securitybaseapi-getsidsubauthoritycount"
           }
         ],
         "namedApiContractEvidence": {
@@ -97,7 +115,21 @@
           "binarySha256": "cd5a1f9d4b53eb0f2ac1e29fbbedc71f8845d63ab85115192f60b78cf197e2fa",
           "importReceiptSha256": "a872374cb767467732c5f36e4b315d58fd424107a9a725800673b9181a714f02",
           "ucrtPlatformSource": "https://learn.microsoft.com/en-us/cpp/porting/upgrade-your-code-to-the-universal-crt?view=msvc-170",
-          "scope": "Documentation-backed static API contracts below the unchanged 14393 floor; not reporter-observed exports, a native x64 export probe, executable startup, installer or ordinary-user acceptance"
+          "scope": "Documentation-backed static API contracts below the unchanged 14393 floor; not reporter-observed exports, a native x64 export probe, executable startup, installer or ordinary-user acceptance",
+          "additionalCandidates": [
+            {
+              "candidateRun": "https://github.com/codewhale-hq/Codewhale/actions/runs/43186389246",
+              "source": "f5a3dda9755170aeafecd0841347a5578a540290",
+              "binarySha256": "6fa43dc8c8680cf3e30a38ea950154b28e30aa97fe373fe82fae6086aa58e1c1",
+              "importReceiptSha256": "9cc807d09bee35101a926773934937427a53214e6e1a54fb85e622b3f18a6ae3",
+              "pairs": [
+                "advapi32.dll!GetSidIdentifierAuthority",
+                "advapi32.dll!GetSidSubAuthority",
+                "advapi32.dll!GetSidSubAuthorityCount"
+              ],
+              "scope": "Three pairs from crates/config private_directory.rs; documentation-backed static contracts, same limits as above"
+            }
+          ]
         }
       },
       "imports": {
@@ -121,6 +153,9 @@
           "GetSecurityDescriptorControl",
           "GetSecurityDescriptorDacl",
           "GetSecurityInfo",
+          "GetSidIdentifierAuthority",
+          "GetSidSubAuthority",
+          "GetSidSubAuthorityCount",
           "GetTokenInformation",
           "ImpersonateAnonymousToken",
           "InitializeAcl",
@@ -631,25 +666,25 @@
       "qualification": {
         "result": "PASS",
         "kind": "exact native system export resolution",
-        "time": "2026-11-07T16:01:13.7658879Z",
+        "time": "2026-11-12T10:08:11.6545236Z",
         "osVersion": "10.0.26200.0",
         "osBuild": "26200.9725",
         "processArchitecture": "ARM64",
         "runner": "GitHub-hosted windows-11-arm; its registry ProductName reads \"Windows 10 Enterprise\", as Windows 11 reports it",
-        "source": "549ff01ebb78856408164058fcb52db1e639ee0d",
-        "sourceUrl": "https://github.com/codewhale-hq/Codewhale/commit/549ff01ebb78856408164058fcb52db1e639ee0d",
-        "binarySha256": "38558ab2b4dc54590503543a0a584c5a82da6f078dd616b5b68e301bf09621b0",
+        "source": "f5a3dda9755170aeafecd0841347a5578a540290",
+        "sourceUrl": "https://github.com/codewhale-hq/Codewhale/commit/f5a3dda9755170aeafecd0841347a5578a540290",
+        "binarySha256": "c8c6f97d0ccf20e392b2457cb970187e79a29caa28d091dbfefe3a1393b99040",
         "binaries": [
           "codewhale-windows-arm64.exe",
           "codew-windows-arm64.exe",
           "codewhale-tui-windows-arm64.exe"
         ],
-        "resolved": 444,
+        "resolved": 447,
         "unresolved": 0,
-        "buildRun": "https://github.com/codewhale-hq/Codewhale/actions/runs/42381818665",
-        "probeRun": "https://github.com/codewhale-hq/Codewhale/actions/runs/42389141087",
+        "buildRun": "https://github.com/codewhale-hq/Codewhale/actions/runs/43186389246",
+        "probeRun": "https://github.com/codewhale-hq/Codewhale/actions/runs/43197611105",
         "probeWorkflow": ".github/workflows/windows-arm64-export-probe.yml",
-        "profile": "dist, built by release-artifacts.yml (windows-arm64 job) in release candidate run 42381818665",
+        "profile": "dist, built by release-artifacts.yml (windows-arm64 job) in release candidate run 43186389246",
         "runtimeAcceptance": "Static export resolution only; this probe did not launch the release executables. The only ARM64 runtime acceptance is the cf7199b52e debug application run of 2026-10-13; the packaged release and older OS builds are unproven",
         "exportProbeMethod": "ARM64 process: LoadLibraryExW per imported DLL with LOAD_LIBRARY_SEARCH_SYSTEM32 (0x00000800), then GetProcAddress for every imported symbol; module path and SHA256 recorded",
         "scope": "System-context export resolution only; no Authenticode check, oldest supported OS, or ordinary-user install inference",
@@ -683,66 +718,119 @@
         "supersedes": {
           "result": "PASS",
           "kind": "exact native system export resolution",
-          "time": "2026-10-16T10:57:53.0754367Z",
+          "time": "2026-11-07T16:01:13.7658879Z",
           "osVersion": "10.0.26200.0",
-          "osBuild": "26200.9457",
+          "osBuild": "26200.9725",
           "processArchitecture": "ARM64",
           "runner": "GitHub-hosted windows-11-arm; its registry ProductName reads \"Windows 10 Enterprise\", as Windows 11 reports it",
-          "source": "db34bc8949a5b5e7e6fd5a88cb67e998011ad9df",
-          "sourceUrl": "https://github.com/codewhale-hq/Codewhale/commit/db34bc8949a5b5e7e6fd5a88cb67e998011ad9df",
-          "binarySha256": "a9a2a521d8d0c971b9d49c2c463cdb49d0c7a4eb700b73e6fed2c743e688eebb",
+          "source": "549ff01ebb78856408164058fcb52db1e639ee0d",
+          "sourceUrl": "https://github.com/codewhale-hq/Codewhale/commit/549ff01ebb78856408164058fcb52db1e639ee0d",
+          "binarySha256": "38558ab2b4dc54590503543a0a584c5a82da6f078dd616b5b68e301bf09621b0",
           "binaries": [
             "codewhale-windows-arm64.exe",
             "codew-windows-arm64.exe",
             "codewhale-tui-windows-arm64.exe"
           ],
-          "resolved": 445,
+          "resolved": 444,
           "unresolved": 0,
-          "buildRun": "https://github.com/codewhale-hq/Codewhale/actions/runs/38867608774",
-          "probeRun": "https://github.com/codewhale-hq/Codewhale/actions/runs/38890634267",
+          "buildRun": "https://github.com/codewhale-hq/Codewhale/actions/runs/42381818665",
+          "probeRun": "https://github.com/codewhale-hq/Codewhale/actions/runs/42389141087",
           "probeWorkflow": ".github/workflows/windows-arm64-export-probe.yml",
-          "profile": "release, built by release-artifacts.yml (windows-arm64 job) in release candidate run 38867608774",
+          "profile": "dist, built by release-artifacts.yml (windows-arm64 job) in release candidate run 42381818665",
           "runtimeAcceptance": "Static export resolution only; this probe did not launch the release executables. The only ARM64 runtime acceptance is the cf7199b52e debug application run of 2026-10-13; the packaged release and older OS builds are unproven",
           "exportProbeMethod": "ARM64 process: LoadLibraryExW per imported DLL with LOAD_LIBRARY_SEARCH_SYSTEM32 (0x00000800), then GetProcAddress for every imported symbol; module path and SHA256 recorded",
           "scope": "System-context export resolution only; no Authenticode check, oldest supported OS, or ordinary-user install inference",
           "systemModuleHashes": {
-            "advapi32.dll": "021452c20f89c50d64c304391ec18169f95a9fa14762f76cfd2dded627dbf0bb",
-            "api-ms-win-core-synch-l1-2-0.dll": "26be64c26bac33368a650372cb5fb1e3235f1bd7883471bd3291dbf6372ae274",
-            "api-ms-win-crt-convert-l1-1-0.dll": "529e795875178b906ea8758abc8de5c28339f336b3e7a121d7778ae040908de9",
-            "api-ms-win-crt-heap-l1-1-0.dll": "529e795875178b906ea8758abc8de5c28339f336b3e7a121d7778ae040908de9",
-            "api-ms-win-crt-locale-l1-1-0.dll": "529e795875178b906ea8758abc8de5c28339f336b3e7a121d7778ae040908de9",
-            "api-ms-win-crt-math-l1-1-0.dll": "529e795875178b906ea8758abc8de5c28339f336b3e7a121d7778ae040908de9",
-            "api-ms-win-crt-runtime-l1-1-0.dll": "529e795875178b906ea8758abc8de5c28339f336b3e7a121d7778ae040908de9",
-            "api-ms-win-crt-stdio-l1-1-0.dll": "529e795875178b906ea8758abc8de5c28339f336b3e7a121d7778ae040908de9",
-            "api-ms-win-crt-string-l1-1-0.dll": "529e795875178b906ea8758abc8de5c28339f336b3e7a121d7778ae040908de9",
-            "api-ms-win-crt-time-l1-1-0.dll": "529e795875178b906ea8758abc8de5c28339f336b3e7a121d7778ae040908de9",
-            "api-ms-win-crt-utility-l1-1-0.dll": "529e795875178b906ea8758abc8de5c28339f336b3e7a121d7778ae040908de9",
-            "bcrypt.dll": "12e3562488e5ca3a26884408d99d81898237c8ca3c5c3c9e40ca6b8f5ed361cd",
-            "bcryptprimitives.dll": "d3da1097e9e1fc16c910350e40c4405ef52f36f134f9bf165fbe02194c1eaa82",
-            "combase.dll": "850b43094ae61a5d3b8f40f0cddf89b4a90190bf5fc57c5242e0bf5b06644b2f",
-            "crypt32.dll": "e3e800f1905658674ce4af1d3efb31d722b9255aaae4fa46c54302d329ba4ab6",
-            "kernel32.dll": "6a7bd4daff9c0c4cc8b9478ecd9828c3f19be257a4a18db9559cae929197fffa",
-            "ntdll.dll": "60ef561645b50dce71ad0919dbbef3ad334a0489432cde718e75b73b558c7997",
-            "oleaut32.dll": "5ebb761b3f96b227006e79d6a92cf5d98ac5a5fd3c136d35dd05391bed867753",
-            "shell32.dll": "ef0fbd71a46fe4b35bff893ea65241cd08ea5231de704c106e39920c03698fe9",
-            "shlwapi.dll": "67f713491bc18614d6ea29b920e3c609cabf9eecf054990e030476a2e9f76a2c",
-            "user32.dll": "f2fc06e7e5016a15ad7f7f4ee250154b7d41d0a0706bf14927a3bbd4c42727dd",
-            "userenv.dll": "18bd1660ab7cacd90789f1bebe35bc6a5730b1e75e5582f53575473fae3db530",
+            "advapi32.dll": "31a87f532c0e99521de26fb25d1ec33cfbb92359952580c88350891602b6cf35",
+            "api-ms-win-core-synch-l1-2-0.dll": "06bf19bcd0b4af9e78c46fc81fe8fb1bf950fb5f295d44de889c5028129b5b56",
+            "api-ms-win-crt-convert-l1-1-0.dll": "e59858bf4245cb0889d8da773612b75d58ac716be5d6bb801e1b39594dccfc6d",
+            "api-ms-win-crt-heap-l1-1-0.dll": "e59858bf4245cb0889d8da773612b75d58ac716be5d6bb801e1b39594dccfc6d",
+            "api-ms-win-crt-locale-l1-1-0.dll": "e59858bf4245cb0889d8da773612b75d58ac716be5d6bb801e1b39594dccfc6d",
+            "api-ms-win-crt-math-l1-1-0.dll": "e59858bf4245cb0889d8da773612b75d58ac716be5d6bb801e1b39594dccfc6d",
+            "api-ms-win-crt-runtime-l1-1-0.dll": "e59858bf4245cb0889d8da773612b75d58ac716be5d6bb801e1b39594dccfc6d",
+            "api-ms-win-crt-stdio-l1-1-0.dll": "e59858bf4245cb0889d8da773612b75d58ac716be5d6bb801e1b39594dccfc6d",
+            "api-ms-win-crt-string-l1-1-0.dll": "e59858bf4245cb0889d8da773612b75d58ac716be5d6bb801e1b39594dccfc6d",
+            "api-ms-win-crt-time-l1-1-0.dll": "e59858bf4245cb0889d8da773612b75d58ac716be5d6bb801e1b39594dccfc6d",
+            "api-ms-win-crt-utility-l1-1-0.dll": "e59858bf4245cb0889d8da773612b75d58ac716be5d6bb801e1b39594dccfc6d",
+            "bcrypt.dll": "2a016e6f29683a7e6be564424d286d0885ff4846fff87cefab4d4e5df47a8d4a",
+            "bcryptprimitives.dll": "6626bf08ddb38e3c37804d7a000c9684977fb0f5f7ab8eb1c0a91a654ec975fc",
+            "combase.dll": "6082ef654c1a0978e127a1e045cb3078b1e80f8f7108b5c6cb4bc601c69b7f51",
+            "crypt32.dll": "3f53b3351e75cb8847b87ce40349d48279cdddab4b19c4c6c9ad38eef6aa6baf",
+            "kernel32.dll": "905c5e00f64ddf9e3439a617b25d4dd1bd0fe54143ade689422b89ccb324ba3a",
+            "ntdll.dll": "0a7201420df35a67e76aa2f2f4adb793b0c0b4fbc927d7da112268a7beb74930",
+            "oleaut32.dll": "61cd70f21624866fb46102644d5af24563a30c85921ed8deed89139cab340f0b",
+            "shell32.dll": "b9a118ff822ab1e950c7a1009de223ecfea5d5f0a23b8a969d58ff19ce28d9cf",
+            "shlwapi.dll": "b4df8c76ea7999912ddf12ba026c5f4a55b683fe9234878f506080575481779a",
+            "user32.dll": "b95f8cce02f51a7cbf77f17582ad3eee8abc577c324963d0026e8dbb977fa097",
+            "userenv.dll": "e9b94fa6123048d535265214322a6a09e450dd85941292bf5258bf39d5b86f9e",
             "vcruntime140.dll": "3c56f4167e2b3d8e6338497731e6aae8cd7ec46bd6789f9423a8d9cf9a630310",
-            "winmm.dll": "2ebb728e5c510492df8fde4ec81b8a8554f2f33d1686b837d22c962b7d2fa9b7",
-            "ws2_32.dll": "e1266872f796f68b4be9c20640ca940a7a96e1ba98c741d472ae305fccf87ca9"
+            "winmm.dll": "4fd628e0a495b1fa7e7653b562a4f5484592c9cfe733184f9808dfce18e26680",
+            "ws2_32.dll": "d10c37755ef07b55829406cfaa22c623ae6f86cce5a5fed42d77d75c29ae824a"
           },
           "supersedes": {
-            "time": "2026-10-13T10:43:07.5742747Z",
-            "osVersion": "10.0.26300",
-            "osBuild": "26300",
-            "source": "cf7199b52ecb9147e4670ae7874924ffc30a6a67",
-            "binarySha256": "ca51f91562a0569ee43855715856e1cfee74e8a834bd6206183a3175373fdb2d",
-            "resolved": 453,
+            "result": "PASS",
+            "kind": "exact native system export resolution",
+            "time": "2026-10-16T10:57:53.0754367Z",
+            "osVersion": "10.0.26200.0",
+            "osBuild": "26200.9457",
+            "processArchitecture": "ARM64",
+            "runner": "GitHub-hosted windows-11-arm; its registry ProductName reads \"Windows 10 Enterprise\", as Windows 11 reports it",
+            "source": "db34bc8949a5b5e7e6fd5a88cb67e998011ad9df",
+            "sourceUrl": "https://github.com/codewhale-hq/Codewhale/commit/db34bc8949a5b5e7e6fd5a88cb67e998011ad9df",
+            "binarySha256": "a9a2a521d8d0c971b9d49c2c463cdb49d0c7a4eb700b73e6fed2c743e688eebb",
+            "binaries": [
+              "codewhale-windows-arm64.exe",
+              "codew-windows-arm64.exe",
+              "codewhale-tui-windows-arm64.exe"
+            ],
+            "resolved": 445,
             "unresolved": 0,
-            "buildCommand": "cargo build --locked --target aarch64-pc-windows-msvc -p codewhale-cli --bin codewhale",
-            "profile": "debug, default features",
-            "runtimeAcceptance": "Only cf7199b52e ARM64 debug application acceptance; final packaged release and older OS unproven"
+            "buildRun": "https://github.com/codewhale-hq/Codewhale/actions/runs/38867608774",
+            "probeRun": "https://github.com/codewhale-hq/Codewhale/actions/runs/38890634267",
+            "probeWorkflow": ".github/workflows/windows-arm64-export-probe.yml",
+            "profile": "release, built by release-artifacts.yml (windows-arm64 job) in release candidate run 38867608774",
+            "runtimeAcceptance": "Static export resolution only; this probe did not launch the release executables. The only ARM64 runtime acceptance is the cf7199b52e debug application run of 2026-10-13; the packaged release and older OS builds are unproven",
+            "exportProbeMethod": "ARM64 process: LoadLibraryExW per imported DLL with LOAD_LIBRARY_SEARCH_SYSTEM32 (0x00000800), then GetProcAddress for every imported symbol; module path and SHA256 recorded",
+            "scope": "System-context export resolution only; no Authenticode check, oldest supported OS, or ordinary-user install inference",
+            "systemModuleHashes": {
+              "advapi32.dll": "021452c20f89c50d64c304391ec18169f95a9fa14762f76cfd2dded627dbf0bb",
+              "api-ms-win-core-synch-l1-2-0.dll": "26be64c26bac33368a650372cb5fb1e3235f1bd7883471bd3291dbf6372ae274",
+              "api-ms-win-crt-convert-l1-1-0.dll": "529e795875178b906ea8758abc8de5c28339f336b3e7a121d7778ae040908de9",
+              "api-ms-win-crt-heap-l1-1-0.dll": "529e795875178b906ea8758abc8de5c28339f336b3e7a121d7778ae040908de9",
+              "api-ms-win-crt-locale-l1-1-0.dll": "529e795875178b906ea8758abc8de5c28339f336b3e7a121d7778ae040908de9",
+              "api-ms-win-crt-math-l1-1-0.dll": "529e795875178b906ea8758abc8de5c28339f336b3e7a121d7778ae040908de9",
+              "api-ms-win-crt-runtime-l1-1-0.dll": "529e795875178b906ea8758abc8de5c28339f336b3e7a121d7778ae040908de9",
+              "api-ms-win-crt-stdio-l1-1-0.dll": "529e795875178b906ea8758abc8de5c28339f336b3e7a121d7778ae040908de9",
+              "api-ms-win-crt-string-l1-1-0.dll": "529e795875178b906ea8758abc8de5c28339f336b3e7a121d7778ae040908de9",
+              "api-ms-win-crt-time-l1-1-0.dll": "529e795875178b906ea8758abc8de5c28339f336b3e7a121d7778ae040908de9",
+              "api-ms-win-crt-utility-l1-1-0.dll": "529e795875178b906ea8758abc8de5c28339f336b3e7a121d7778ae040908de9",
+              "bcrypt.dll": "12e3562488e5ca3a26884408d99d81898237c8ca3c5c3c9e40ca6b8f5ed361cd",
+              "bcryptprimitives.dll": "d3da1097e9e1fc16c910350e40c4405ef52f36f134f9bf165fbe02194c1eaa82",
+              "combase.dll": "850b43094ae61a5d3b8f40f0cddf89b4a90190bf5fc57c5242e0bf5b06644b2f",
+              "crypt32.dll": "e3e800f1905658674ce4af1d3efb31d722b9255aaae4fa46c54302d329ba4ab6",
+              "kernel32.dll": "6a7bd4daff9c0c4cc8b9478ecd9828c3f19be257a4a18db9559cae929197fffa",
+              "ntdll.dll": "60ef561645b50dce71ad0919dbbef3ad334a0489432cde718e75b73b558c7997",
+              "oleaut32.dll": "5ebb761b3f96b227006e79d6a92cf5d98ac5a5fd3c136d35dd05391bed867753",
+              "shell32.dll": "ef0fbd71a46fe4b35bff893ea65241cd08ea5231de704c106e39920c03698fe9",
+              "shlwapi.dll": "67f713491bc18614d6ea29b920e3c609cabf9eecf054990e030476a2e9f76a2c",
+              "user32.dll": "f2fc06e7e5016a15ad7f7f4ee250154b7d41d0a0706bf14927a3bbd4c42727dd",
+              "userenv.dll": "18bd1660ab7cacd90789f1bebe35bc6a5730b1e75e5582f53575473fae3db530",
+              "vcruntime140.dll": "3c56f4167e2b3d8e6338497731e6aae8cd7ec46bd6789f9423a8d9cf9a630310",
+              "winmm.dll": "2ebb728e5c510492df8fde4ec81b8a8554f2f33d1686b837d22c962b7d2fa9b7",
+              "ws2_32.dll": "e1266872f796f68b4be9c20640ca940a7a96e1ba98c741d472ae305fccf87ca9"
+            },
+            "supersedes": {
+              "time": "2026-10-13T10:43:07.5742747Z",
+              "osVersion": "10.0.26300",
+              "osBuild": "26300",
+              "source": "cf7199b52ecb9147e4670ae7874924ffc30a6a67",
+              "binarySha256": "ca51f91562a0569ee43855715856e1cfee74e8a834bd6206183a3175373fdb2d",
+              "resolved": 453,
+              "unresolved": 0,
+              "buildCommand": "cargo build --locked --target aarch64-pc-windows-msvc -p codewhale-cli --bin codewhale",
+              "profile": "debug, default features",
+              "runtimeAcceptance": "Only cf7199b52e ARM64 debug application acceptance; final packaged release and older OS unproven"
+            }
           }
         }
       },
@@ -767,6 +855,9 @@
           "GetSecurityDescriptorControl",
           "GetSecurityDescriptorDacl",
           "GetSecurityInfo",
+          "GetSidIdentifierAuthority",
+          "GetSidSubAuthority",
+          "GetSidSubAuthorityCount",
           "GetTokenInformation",
           "ImpersonateAnonymousToken",
           "InitializeAcl",

scripts/test_ci_migration_wiring.py

@@ -273,7 +273,7 @@ class CiWiringTests(unittest.TestCase):
         block = ci[start:next_step]
         self.assertIn(
             "sh scripts/with-hermetic-test-home.sh cargo nextest run "
-            "-p codewhale-tui -p codewhale-runtime --lib --locked --no-tests=fail "
+            "-p codewhale-engine -p codewhale-tui -p codewhale-runtime --lib --locked --no-tests=fail "
             "-E 'test(auto_review) | test(authority) | test(sandbox)'", block
         )
         self.assertIn(
@@ -290,27 +290,27 @@ class CiWiringTests(unittest.TestCase):
         self.assertIn("check-command-migration-manifest.py", ci)
         self.assertIn("needs.changes.outputs.heavy == 'true'", ci)
 
-    def test_shared_process_mode_is_manual_hosted_and_preserves_default_runner(self) -> None:
+    def test_shared_process_mode_is_manual_hosted_ubuntu_only_and_replaces_the_sharded_suite(self) -> None:
         ci = load_ci()
         self.assertIn("default: nextest", ci)
-        matrix = ci.split("  test:\n", 1)[1].split("    steps:\n", 1)[0]
-        self.assertIn(
-            SHARED_PROCESS_MODE + " && '[\"ubuntu-latest\"]' || "
-            "'[\"ubuntu-latest\",\"macos-latest\",\"windows-latest\"]'", matrix,
-        )
-        self.assertIn("needs.changes.outputs.trusted == 'true'", matrix)
+        job = ci.split("  test:\n", 1)[1].split("    steps:\n", 1)[0]
+        self.assertIn("runs-on: ubuntu-latest", job)
+        self.assertIn(SHARED_PROCESS_MODE, job)
+        self.assertNotIn("pull_request", job)
+        for name in ("test-ubuntu", "test-macos", "test-windows"):
+            caller = ci.split(f"  {name}:\n", 1)[1].split("    uses:", 1)[0]
+            self.assertIn(f"!({SHARED_PROCESS_MODE})", caller)
         qualification = named_step(ci, "Shared-process workspace qualification")
-        self.assertIn("if: " + SHARED_PROCESS_MODE + " && matrix.os == 'ubuntu-latest'", qualification)
-        for name, command in (
-            ("Run tests", "cargo nextest run --workspace --all-features --locked --profile ci"),
-            ("Run doctests", "cargo test --workspace --all-features --locked --doc"),
-        ):
-            step = named_step(ci, name)
-            self.assertIn(command, step)
-            self.assertIn(f"!({SHARED_PROCESS_MODE})", step)
+        self.assertIn("if: " + SHARED_PROCESS_MODE, qualification)
         upload = named_step(ci, "Retain shared-process qualification evidence")
         self.assertIn("if: always() && " + SHARED_PROCESS_MODE, upload)
         self.assertIn("if-no-files-found: error", upload)
+        suite = (ROOT / ".github" / "workflows" / "ci-test-suite.yml").read_text(encoding="utf-8")
+        for name, command in (
+            ("Run tests", "cargo nextest run"),
+            ("Run doctests", "cargo test --workspace --all-features --locked --doc"),
+        ):
+            self.assertIn(command, named_step(suite, name))
 
     def test_shared_process_runs_exact_command_twice_with_fresh_hermetic_homes(self) -> None:
         proof = run_shared_process_fixture()
@@ -332,31 +332,24 @@ class CiWiringTests(unittest.TestCase):
             for identity in ISSUE_6698_TESTS:
                 self.assertIn(f"test {identity} ... ok", proof["evidence"][f"run-{run}.log"])
 
-    def test_shared_process_keeps_macos_budget_when_self_hosted_test_leg_is_absent(self) -> None:
+    def test_macos_budget_runs_on_every_heavy_dispatch_and_never_on_pull_requests(self) -> None:
         job = load_ci().split("  macos-budget:\n", 1)[1].split("    steps:\n", 1)[0]
         condition = next(line.strip()[4:] for line in job.splitlines() if line.strip().startswith("if: "))
         self.assertIn("runs-on: macos-latest", job)
         for event in ("pull_request", "push", "schedule", "workflow_dispatch"):
-            for mode in ("nextest", "shared-process-twice"):
-                for heavy, trusted, self_hosted in (
-                    (True, True, True), (True, True, False),
-                    (True, False, True), (False, True, True),
-                ):
-                    with self.subTest(event=event, mode=mode, heavy=heavy, trusted=trusted, self_hosted=self_hosted):
+            for heavy in (True, False):
+                for scoped in (True, False):
+                    with self.subTest(event=event, heavy=heavy, scoped=scoped):
                         expression = condition
                         for name, value in {
                             "github.event_name": event,
-                            "inputs.workspace_test_mode": mode,
                             "needs.changes.outputs.heavy": str(heavy).lower(),
-                            "needs.changes.outputs.trusted": str(trusted).lower(),
-                            "vars.CW_SELF_HOSTED_MAC": str(self_hosted).lower(),
+                            "needs.changes.outputs.scoped": str(scoped).lower(),
                         }.items():
                             expression = expression.replace(name, repr(value))
                         expression = re.sub(r"!(?!=)", "not ", expression.replace("&&", "and").replace("||", "or"))
                         actual = eval(expression, {"__builtins__": {}}, {})
-                        previous = heavy and event != "pull_request" and not (trusted and self_hosted)
-                        qualification = heavy and event == "workflow_dispatch" and mode == "shared-process-twice"
-                        self.assertEqual(actual, previous or qualification)
+                        self.assertEqual(actual, heavy and not scoped and event != "pull_request")
 
     def test_shared_process_retains_failed_first_run_and_still_runs_second_once(self) -> None:
         proof = run_shared_process_fixture("first_failure")

web/data/agent-docs.generated.json

@@ -58,8 +58,8 @@
     "sources": [
       {
         "path": "docs/CONFIGURATION.md",
-        "sha256": "00f8d3b2b40dc4145a78b5bf93bd6d4e96c83478fdb5b4e0b7316108926ba1b7",
-        "content": "# Configuration\n\ncodewhale reads configuration from a TOML file plus environment variables.\nAt process startup it may also load literal built-in-provider credentials from\na workspace-local `.env` file. Use the tracked `.env.example` as the template;\ncopy it to `.env`, then add only credential values.\n\nA workspace is not configuration authority. Codewhale therefore ignores\nconfig/profile/home paths, provider/model/base-URL routing, MCP/plugin state,\napproval/sandbox/shell posture, executable paths, runtime settings, and every\nother non-credential `.env` entry. Variable expansion is rejected so a\nrepository cannot substitute an ambient secret into a credential value. Use\n`config.toml`, CLI flags, or values exported by the launching shell for those\nexplicit control-plane settings. `.env` is read from a stable regular-file\nhandle, is capped at 1 MiB, and symbolic links, reparse points, and multiply\nlinked files are rejected.\n\n## Reading and checking configuration from the CLI\n\n`codewhale config get <key>` reads scalar keys, whole tables such as `tools`,\nand nested paths such as `tools.user_input_timeout_seconds`. Displayed tables\nand nested values apply the same recursive credential redaction as `config dump`.\n\n`config set` supports its named scalar keys and the provider, route, and\nnotification commands. It refuses a key that nothing reads and suggests the\nnearest real key (`config set calm_mod on` names `calm_mode`). A settings.toml\nkey such as `calm_mode` or `tool_collapse` is validated and written to the\nuser-global settings.toml, not config.toml; `config get` reads it back from\nthere even when an old config.toml copy (which nothing reads) is still present,\nand names that copy so you can `config unset` it. Settings keys have no project\nscope, so `--project` refuses them. A config.toml key is checked against the\ntype its reader expects and stored as a TOML boolean or number where the\nreader needs one (`yolo = true`, `max_subagents = 4`); `reasoning_effort`\naccepts the same aliases as `/effort` (#6563). Other dotted writes fail before\nmodifying the file and name the TOML table to edit. For example, set a tools\ntimeout in the file as:\n\n```toml\n[tools]\nuser_input_timeout_seconds = 0\n```\n\n`codewhale config doctor` checks credential presence and endpoint shape, and\nwarns about each config.toml root key that nothing reads: a settings.toml key\nleft in config.toml is named as misplaced, and any other unread key gets a\ndid-you-mean. Keys read by any runtime reader are not reported. The warnings do\nnot fail the check, and a clean result does not validate every runtime setting's\nvalue (#6083, #6563).\n\n## Usage and execution statistics\n\n`codewhale stats` is an alias of `codewhale metrics`; both use the same local,\nread-only backend. Use `--json` for the structured `stats` report. For example:\n\n```sh\ncodewhale stats --days 7 --models 10 --tools 20 --roles 10\ncodewhale metrics --project /path/to/workspace --json\n```\n\n`--days N` accepts a positive number of days and conflicts with `--since`.\n`--since 24h` and the existing duration syntax remain available. `--models N`,\n`--tools N` and `--roles N` select the top positive N rows by known tokens,\nidentified tool executions and recorded Fleet tasks respectively. These limits\nonly shorten the display; they do not remove stored receipts.\n\nThe report separates saved-document and Runtime views, frozen native-source\nmoney and token subtotals by actual model, retained tool-result history\nestimates and outcomes, completed compaction observations, and recorded Fleet\nroles. USD and CNY stay separate. A cumulative correction replaces its proven\npredecessor; copied Session/Runtime receipts and inherited Runtime fork rows\nare not added as new provider work. Conflicting source versions are shown\nseparately and excluded from the subtotal. Saved legacy display totals remain\nvisible as document projections without being charged again globally.\n\nCoverage is explicit. A missing quote, usage receipt, tool outcome, output\nshape or recovery link is unknown, not zero. Empty scalar usage from older\ntransport records cannot prove a reported zero. Token counts describe known\nreported data; tool history tokens are estimates of retained text, not bytes\nproven to have been replayed on a provider request. Latest prepared-request\ncomponents and source-linked compaction/role economics are unavailable where\nthey were not captured. The report does not infer provider invoices or prove\ncomplete historical coverage.\n\n`--project WORKSPACE` filters current saved/thread/Fleet views. Original source\nworkspace is not yet captured, so native money and model subtotals are\nunavailable for that filter; mutable current metadata is not used to assign a\ncharge. Older receipt projections without reliable workspace scope are also\nunavailable under this filter. Without it, those legacy receipt views remain\nseparate fromLine truncated
+        "sha256": "fb934929d28d87dd161a26c427ffa21595ac358d994a46806c53f84fd49cb0f8",
+        "content": "# Configuration\n\ncodewhale reads configuration from a TOML file plus environment variables.\nAt process startup it may also load literal built-in-provider credentials from\na workspace-local `.env` file. Use the tracked `.env.example` as the template;\ncopy it to `.env`, then add only credential values.\n\nA workspace is not configuration authority. Codewhale therefore ignores\nconfig/profile/home paths, provider/model/base-URL routing, MCP/plugin state,\napproval/sandbox/shell posture, executable paths, runtime settings, and every\nother non-credential `.env` entry. Variable expansion is rejected so a\nrepository cannot substitute an ambient secret into a credential value. Use\n`config.toml`, CLI flags, or values exported by the launching shell for those\nexplicit control-plane settings. `.env` is read from a stable regular-file\nhandle, is capped at 1 MiB, and symbolic links, reparse points, and multiply\nlinked files are rejected.\n\n## Reading and checking configuration from the CLI\n\n`codewhale config get <key>` reads scalar keys, whole tables such as `tools`,\nand nested paths such as `tools.user_input_timeout_seconds`. Displayed tables\nand nested values apply the same recursive credential redaction as `config dump`.\n\n`config set` supports its named scalar keys and the provider, route, and\nnotification commands. It refuses a key that nothing reads and suggests the\nnearest real key (`config set calm_mod on` names `calm_mode`). A settings.toml\nkey such as `calm_mode` or `tool_collapse` is validated and written to the\nuser-global settings.toml, not config.toml; `config get` reads it back from\nthere even when an old config.toml copy (which nothing reads) is still present,\nand names that copy so you can `config unset` it. Settings keys have no project\nscope, so `--project` refuses them. A config.toml key is checked against the\ntype its reader expects and stored as a TOML boolean or number where the\nreader needs one (`yolo = true`, `max_subagents = 4`); `reasoning_effort`\naccepts the same aliases as `/effort` (#6563). Other dotted writes fail before\nmodifying the file and name the TOML table to edit. For example, set a tools\ntimeout in the file as:\n\n```toml\n[tools]\nuser_input_timeout_seconds = 0\n```\n\n`codewhale config doctor` checks credential presence and endpoint shape, and\nwarns about each config.toml root key that nothing reads: a settings.toml key\nleft in config.toml is named as misplaced, and any other unread key gets a\ndid-you-mean. Keys read by any runtime reader are not reported. The warnings do\nnot fail the check, and a clean result does not validate every runtime setting's\nvalue (#6083, #6563).\n\n## Usage and execution statistics\n\n`codewhale stats` is an alias of `codewhale metrics`; both use the same local,\nread-only backend. Use `--json` for the structured `stats` report. For example:\n\n```sh\ncodewhale stats --days 7 --models 10 --tools 20 --roles 10\ncodewhale metrics --project /path/to/workspace --json\n```\n\n`--days N` accepts a positive number of days and conflicts with `--since`.\n`--since 24h` and the existing duration syntax remain available. `--models N`,\n`--tools N` and `--roles N` select the top positive N rows by known tokens,\nidentified tool executions and recorded Fleet tasks respectively. These limits\nonly shorten the display; they do not remove stored receipts.\n\nThe report separates saved-document and Runtime views, frozen native-source\nmoney and token subtotals by actual model, retained tool-result history\nestimates and outcomes, completed compaction observations, and recorded Fleet\nroles. USD and CNY stay separate. A cumulative correction replaces its proven\npredecessor; copied Session/Runtime receipts and inherited Runtime fork rows\nare not added as new provider work. Conflicting source versions are shown\nseparately and excluded from the subtotal. Saved legacy display totals remain\nvisible as document projections without being charged again globally.\n\nCoverage is explicit. A missing quote, usage receipt, tool outcome, output\nshape or recovery link is unknown, not zero. Empty scalar usage from older\ntransport records cannot prove a reported zero. Token counts describe known\nreported data; tool history tokens are estimates of retained text, not bytes\nproven to have been replayed on a provider request. Latest prepared-request\ncomponents and source-linked compaction/role economics are unavailable where\nthey were not captured. The report does not infer provider invoices or prove\ncomplete historical coverage.\n\n`--project WORKSPACE` filters current saved/thread/Fleet views. Original source\nworkspace is not yet captured, so native money and model subtotals are\nunavailable for that filter; mutable current metadata is not used to assign a\ncharge. Older receipt projections without reliable workspace scope are also\nunavailable under this filter. Without it, those legacy receipt views remain\nseparate fromLine truncated
       },
       {
         "path": "docs/PROVIDERS.md",
@@ -90,8 +90,8 @@
     "sources": [
       {
         "path": "docs/CONFIGURATION.md",
-        "sha256": "00f8d3b2b40dc4145a78b5bf93bd6d4e96c83478fdb5b4e0b7316108926ba1b7",
-        "content": "# Configuration\n\ncodewhale reads configuration from a TOML file plus environment variables.\nAt process startup it may also load literal built-in-provider credentials from\na workspace-local `.env` file. Use the tracked `.env.example` as the template;\ncopy it to `.env`, then add only credential values.\n\nA workspace is not configuration authority. Codewhale therefore ignores\nconfig/profile/home paths, provider/model/base-URL routing, MCP/plugin state,\napproval/sandbox/shell posture, executable paths, runtime settings, and every\nother non-credential `.env` entry. Variable expansion is rejected so a\nrepository cannot substitute an ambient secret into a credential value. Use\n`config.toml`, CLI flags, or values exported by the launching shell for those\nexplicit control-plane settings. `.env` is read from a stable regular-file\nhandle, is capped at 1 MiB, and symbolic links, reparse points, and multiply\nlinked files are rejected.\n\n## Reading and checking configuration from the CLI\n\n`codewhale config get <key>` reads scalar keys, whole tables such as `tools`,\nand nested paths such as `tools.user_input_timeout_seconds`. Displayed tables\nand nested values apply the same recursive credential redaction as `config dump`.\n\n`config set` supports its named scalar keys and the provider, route, and\nnotification commands. It refuses a key that nothing reads and suggests the\nnearest real key (`config set calm_mod on` names `calm_mode`). A settings.toml\nkey such as `calm_mode` or `tool_collapse` is validated and written to the\nuser-global settings.toml, not config.toml; `config get` reads it back from\nthere even when an old config.toml copy (which nothing reads) is still present,\nand names that copy so you can `config unset` it. Settings keys have no project\nscope, so `--project` refuses them. A config.toml key is checked against the\ntype its reader expects and stored as a TOML boolean or number where the\nreader needs one (`yolo = true`, `max_subagents = 4`); `reasoning_effort`\naccepts the same aliases as `/effort` (#6563). Other dotted writes fail before\nmodifying the file and name the TOML table to edit. For example, set a tools\ntimeout in the file as:\n\n```toml\n[tools]\nuser_input_timeout_seconds = 0\n```\n\n`codewhale config doctor` checks credential presence and endpoint shape, and\nwarns about each config.toml root key that nothing reads: a settings.toml key\nleft in config.toml is named as misplaced, and any other unread key gets a\ndid-you-mean. Keys read by any runtime reader are not reported. The warnings do\nnot fail the check, and a clean result does not validate every runtime setting's\nvalue (#6083, #6563).\n\n## Usage and execution statistics\n\n`codewhale stats` is an alias of `codewhale metrics`; both use the same local,\nread-only backend. Use `--json` for the structured `stats` report. For example:\n\n```sh\ncodewhale stats --days 7 --models 10 --tools 20 --roles 10\ncodewhale metrics --project /path/to/workspace --json\n```\n\n`--days N` accepts a positive number of days and conflicts with `--since`.\n`--since 24h` and the existing duration syntax remain available. `--models N`,\n`--tools N` and `--roles N` select the top positive N rows by known tokens,\nidentified tool executions and recorded Fleet tasks respectively. These limits\nonly shorten the display; they do not remove stored receipts.\n\nThe report separates saved-document and Runtime views, frozen native-source\nmoney and token subtotals by actual model, retained tool-result history\nestimates and outcomes, completed compaction observations, and recorded Fleet\nroles. USD and CNY stay separate. A cumulative correction replaces its proven\npredecessor; copied Session/Runtime receipts and inherited Runtime fork rows\nare not added as new provider work. Conflicting source versions are shown\nseparately and excluded from the subtotal. Saved legacy display totals remain\nvisible as document projections without being charged again globally.\n\nCoverage is explicit. A missing quote, usage receipt, tool outcome, output\nshape or recovery link is unknown, not zero. Empty scalar usage from older\ntransport records cannot prove a reported zero. Token counts describe known\nreported data; tool history tokens are estimates of retained text, not bytes\nproven to have been replayed on a provider request. Latest prepared-request\ncomponents and source-linked compaction/role economics are unavailable where\nthey were not captured. The report does not infer provider invoices or prove\ncomplete historical coverage.\n\n`--project WORKSPACE` filters current saved/thread/Fleet views. Original source\nworkspace is not yet captured, so native money and model subtotals are\nunavailable for that filter; mutable current metadata is not used to assign a\ncharge. Older receipt projections without reliable workspace scope are also\nunavailable under this filter. Without it, those legacy receipt views remain\nseparate fromLine truncated
+        "sha256": "fb934929d28d87dd161a26c427ffa21595ac358d994a46806c53f84fd49cb0f8",
+        "content": "# Configuration\n\ncodewhale reads configuration from a TOML file plus environment variables.\nAt process startup it may also load literal built-in-provider credentials from\na workspace-local `.env` file. Use the tracked `.env.example` as the template;\ncopy it to `.env`, then add only credential values.\n\nA workspace is not configuration authority. Codewhale therefore ignores\nconfig/profile/home paths, provider/model/base-URL routing, MCP/plugin state,\napproval/sandbox/shell posture, executable paths, runtime settings, and every\nother non-credential `.env` entry. Variable expansion is rejected so a\nrepository cannot substitute an ambient secret into a credential value. Use\n`config.toml`, CLI flags, or values exported by the launching shell for those\nexplicit control-plane settings. `.env` is read from a stable regular-file\nhandle, is capped at 1 MiB, and symbolic links, reparse points, and multiply\nlinked files are rejected.\n\n## Reading and checking configuration from the CLI\n\n`codewhale config get <key>` reads scalar keys, whole tables such as `tools`,\nand nested paths such as `tools.user_input_timeout_seconds`. Displayed tables\nand nested values apply the same recursive credential redaction as `config dump`.\n\n`config set` supports its named scalar keys and the provider, route, and\nnotification commands. It refuses a key that nothing reads and suggests the\nnearest real key (`config set calm_mod on` names `calm_mode`). A settings.toml\nkey such as `calm_mode` or `tool_collapse` is validated and written to the\nuser-global settings.toml, not config.toml; `config get` reads it back from\nthere even when an old config.toml copy (which nothing reads) is still present,\nand names that copy so you can `config unset` it. Settings keys have no project\nscope, so `--project` refuses them. A config.toml key is checked against the\ntype its reader expects and stored as a TOML boolean or number where the\nreader needs one (`yolo = true`, `max_subagents = 4`); `reasoning_effort`\naccepts the same aliases as `/effort` (#6563). Other dotted writes fail before\nmodifying the file and name the TOML table to edit. For example, set a tools\ntimeout in the file as:\n\n```toml\n[tools]\nuser_input_timeout_seconds = 0\n```\n\n`codewhale config doctor` checks credential presence and endpoint shape, and\nwarns about each config.toml root key that nothing reads: a settings.toml key\nleft in config.toml is named as misplaced, and any other unread key gets a\ndid-you-mean. Keys read by any runtime reader are not reported. The warnings do\nnot fail the check, and a clean result does not validate every runtime setting's\nvalue (#6083, #6563).\n\n## Usage and execution statistics\n\n`codewhale stats` is an alias of `codewhale metrics`; both use the same local,\nread-only backend. Use `--json` for the structured `stats` report. For example:\n\n```sh\ncodewhale stats --days 7 --models 10 --tools 20 --roles 10\ncodewhale metrics --project /path/to/workspace --json\n```\n\n`--days N` accepts a positive number of days and conflicts with `--since`.\n`--since 24h` and the existing duration syntax remain available. `--models N`,\n`--tools N` and `--roles N` select the top positive N rows by known tokens,\nidentified tool executions and recorded Fleet tasks respectively. These limits\nonly shorten the display; they do not remove stored receipts.\n\nThe report separates saved-document and Runtime views, frozen native-source\nmoney and token subtotals by actual model, retained tool-result history\nestimates and outcomes, completed compaction observations, and recorded Fleet\nroles. USD and CNY stay separate. A cumulative correction replaces its proven\npredecessor; copied Session/Runtime receipts and inherited Runtime fork rows\nare not added as new provider work. Conflicting source versions are shown\nseparately and excluded from the subtotal. Saved legacy display totals remain\nvisible as document projections without being charged again globally.\n\nCoverage is explicit. A missing quote, usage receipt, tool outcome, output\nshape or recovery link is unknown, not zero. Empty scalar usage from older\ntransport records cannot prove a reported zero. Token counts describe known\nreported data; tool history tokens are estimates of retained text, not bytes\nproven to have been replayed on a provider request. Latest prepared-request\ncomponents and source-linked compaction/role economics are unavailable where\nthey were not captured. The report does not infer provider invoices or prove\ncomplete historical coverage.\n\n`--project WORKSPACE` filters current saved/thread/Fleet views. Original source\nworkspace is not yet captured, so native money and model subtotals are\nunavailable for that filter; mutable current metadata is not used to assign a\ncharge. Older receipt projections without reliable workspace scope are also\nunavailable under this filter. Without it, those legacy receipt views remain\nseparate fromLine truncated
       },
       {
         "path": "docs/LEGACY_PATHS.md",
@@ -122,8 +122,8 @@
       },
       {
         "path": "docs/CONFIGURATION.md",
-        "sha256": "00f8d3b2b40dc4145a78b5bf93bd6d4e96c83478fdb5b4e0b7316108926ba1b7",
-        "content": "# Configuration\n\ncodewhale reads configuration from a TOML file plus environment variables.\nAt process startup it may also load literal built-in-provider credentials from\na workspace-local `.env` file. Use the tracked `.env.example` as the template;\ncopy it to `.env`, then add only credential values.\n\nA workspace is not configuration authority. Codewhale therefore ignores\nconfig/profile/home paths, provider/model/base-URL routing, MCP/plugin state,\napproval/sandbox/shell posture, executable paths, runtime settings, and every\nother non-credential `.env` entry. Variable expansion is rejected so a\nrepository cannot substitute an ambient secret into a credential value. Use\n`config.toml`, CLI flags, or values exported by the launching shell for those\nexplicit control-plane settings. `.env` is read from a stable regular-file\nhandle, is capped at 1 MiB, and symbolic links, reparse points, and multiply\nlinked files are rejected.\n\n## Reading and checking configuration from the CLI\n\n`codewhale config get <key>` reads scalar keys, whole tables such as `tools`,\nand nested paths such as `tools.user_input_timeout_seconds`. Displayed tables\nand nested values apply the same recursive credential redaction as `config dump`.\n\n`config set` supports its named scalar keys and the provider, route, and\nnotification commands. It refuses a key that nothing reads and suggests the\nnearest real key (`config set calm_mod on` names `calm_mode`). A settings.toml\nkey such as `calm_mode` or `tool_collapse` is validated and written to the\nuser-global settings.toml, not config.toml; `config get` reads it back from\nthere even when an old config.toml copy (which nothing reads) is still present,\nand names that copy so you can `config unset` it. Settings keys have no project\nscope, so `--project` refuses them. A config.toml key is checked against the\ntype its reader expects and stored as a TOML boolean or number where the\nreader needs one (`yolo = true`, `max_subagents = 4`); `reasoning_effort`\naccepts the same aliases as `/effort` (#6563). Other dotted writes fail before\nmodifying the file and name the TOML table to edit. For example, set a tools\ntimeout in the file as:\n\n```toml\n[tools]\nuser_input_timeout_seconds = 0\n```\n\n`codewhale config doctor` checks credential presence and endpoint shape, and\nwarns about each config.toml root key that nothing reads: a settings.toml key\nleft in config.toml is named as misplaced, and any other unread key gets a\ndid-you-mean. Keys read by any runtime reader are not reported. The warnings do\nnot fail the check, and a clean result does not validate every runtime setting's\nvalue (#6083, #6563).\n\n## Usage and execution statistics\n\n`codewhale stats` is an alias of `codewhale metrics`; both use the same local,\nread-only backend. Use `--json` for the structured `stats` report. For example:\n\n```sh\ncodewhale stats --days 7 --models 10 --tools 20 --roles 10\ncodewhale metrics --project /path/to/workspace --json\n```\n\n`--days N` accepts a positive number of days and conflicts with `--since`.\n`--since 24h` and the existing duration syntax remain available. `--models N`,\n`--tools N` and `--roles N` select the top positive N rows by known tokens,\nidentified tool executions and recorded Fleet tasks respectively. These limits\nonly shorten the display; they do not remove stored receipts.\n\nThe report separates saved-document and Runtime views, frozen native-source\nmoney and token subtotals by actual model, retained tool-result history\nestimates and outcomes, completed compaction observations, and recorded Fleet\nroles. USD and CNY stay separate. A cumulative correction replaces its proven\npredecessor; copied Session/Runtime receipts and inherited Runtime fork rows\nare not added as new provider work. Conflicting source versions are shown\nseparately and excluded from the subtotal. Saved legacy display totals remain\nvisible as document projections without being charged again globally.\n\nCoverage is explicit. A missing quote, usage receipt, tool outcome, output\nshape or recovery link is unknown, not zero. Empty scalar usage from older\ntransport records cannot prove a reported zero. Token counts describe known\nreported data; tool history tokens are estimates of retained text, not bytes\nproven to have been replayed on a provider request. Latest prepared-request\ncomponents and source-linked compaction/role economics are unavailable where\nthey were not captured. The report does not infer provider invoices or prove\ncomplete historical coverage.\n\n`--project WORKSPACE` filters current saved/thread/Fleet views. Original source\nworkspace is not yet captured, so native money and model subtotals are\nunavailable for that filter; mutable current metadata is not used to assign a\ncharge. Older receipt projections without reliable workspace scope are also\nunavailable under this filter. Without it, those legacy receipt views remain\nseparate fromLine truncated
+        "sha256": "fb934929d28d87dd161a26c427ffa21595ac358d994a46806c53f84fd49cb0f8",
+        "content": "# Configuration\n\ncodewhale reads configuration from a TOML file plus environment variables.\nAt process startup it may also load literal built-in-provider credentials from\na workspace-local `.env` file. Use the tracked `.env.example` as the template;\ncopy it to `.env`, then add only credential values.\n\nA workspace is not configuration authority. Codewhale therefore ignores\nconfig/profile/home paths, provider/model/base-URL routing, MCP/plugin state,\napproval/sandbox/shell posture, executable paths, runtime settings, and every\nother non-credential `.env` entry. Variable expansion is rejected so a\nrepository cannot substitute an ambient secret into a credential value. Use\n`config.toml`, CLI flags, or values exported by the launching shell for those\nexplicit control-plane settings. `.env` is read from a stable regular-file\nhandle, is capped at 1 MiB, and symbolic links, reparse points, and multiply\nlinked files are rejected.\n\n## Reading and checking configuration from the CLI\n\n`codewhale config get <key>` reads scalar keys, whole tables such as `tools`,\nand nested paths such as `tools.user_input_timeout_seconds`. Displayed tables\nand nested values apply the same recursive credential redaction as `config dump`.\n\n`config set` supports its named scalar keys and the provider, route, and\nnotification commands. It refuses a key that nothing reads and suggests the\nnearest real key (`config set calm_mod on` names `calm_mode`). A settings.toml\nkey such as `calm_mode` or `tool_collapse` is validated and written to the\nuser-global settings.toml, not config.toml; `config get` reads it back from\nthere even when an old config.toml copy (which nothing reads) is still present,\nand names that copy so you can `config unset` it. Settings keys have no project\nscope, so `--project` refuses them. A config.toml key is checked against the\ntype its reader expects and stored as a TOML boolean or number where the\nreader needs one (`yolo = true`, `max_subagents = 4`); `reasoning_effort`\naccepts the same aliases as `/effort` (#6563). Other dotted writes fail before\nmodifying the file and name the TOML table to edit. For example, set a tools\ntimeout in the file as:\n\n```toml\n[tools]\nuser_input_timeout_seconds = 0\n```\n\n`codewhale config doctor` checks credential presence and endpoint shape, and\nwarns about each config.toml root key that nothing reads: a settings.toml key\nleft in config.toml is named as misplaced, and any other unread key gets a\ndid-you-mean. Keys read by any runtime reader are not reported. The warnings do\nnot fail the check, and a clean result does not validate every runtime setting's\nvalue (#6083, #6563).\n\n## Usage and execution statistics\n\n`codewhale stats` is an alias of `codewhale metrics`; both use the same local,\nread-only backend. Use `--json` for the structured `stats` report. For example:\n\n```sh\ncodewhale stats --days 7 --models 10 --tools 20 --roles 10\ncodewhale metrics --project /path/to/workspace --json\n```\n\n`--days N` accepts a positive number of days and conflicts with `--since`.\n`--since 24h` and the existing duration syntax remain available. `--models N`,\n`--tools N` and `--roles N` select the top positive N rows by known tokens,\nidentified tool executions and recorded Fleet tasks respectively. These limits\nonly shorten the display; they do not remove stored receipts.\n\nThe report separates saved-document and Runtime views, frozen native-source\nmoney and token subtotals by actual model, retained tool-result history\nestimates and outcomes, completed compaction observations, and recorded Fleet\nroles. USD and CNY stay separate. A cumulative correction replaces its proven\npredecessor; copied Session/Runtime receipts and inherited Runtime fork rows\nare not added as new provider work. Conflicting source versions are shown\nseparately and excluded from the subtotal. Saved legacy display totals remain\nvisible as document projections without being charged again globally.\n\nCoverage is explicit. A missing quote, usage receipt, tool outcome, output\nshape or recovery link is unknown, not zero. Empty scalar usage from older\ntransport records cannot prove a reported zero. Token counts describe known\nreported data; tool history tokens are estimates of retained text, not bytes\nproven to have been replayed on a provider request. Latest prepared-request\ncomponents and source-linked compaction/role economics are unavailable where\nthey were not captured. The report does not infer provider invoices or prove\ncomplete historical coverage.\n\n`--project WORKSPACE` filters current saved/thread/Fleet views. Original source\nworkspace is not yet captured, so native money and model subtotals are\nunavailable for that filter; mutable current metadata is not used to assign a\ncharge. Older receipt projections without reliable workspace scope are also\nunavailable under this filter. Without it, those legacy receipt views remain\nseparate fromLine truncated
       }
     ]
   },
@@ -265,8 +265,8 @@
       },
       {
         "path": "docs/CONFIGURATION.md",
-        "sha256": "00f8d3b2b40dc4145a78b5bf93bd6d4e96c83478fdb5b4e0b7316108926ba1b7",
-        "content": "# Configuration\n\ncodewhale reads configuration from a TOML file plus environment variables.\nAt process startup it may also load literal built-in-provider credentials from\na workspace-local `.env` file. Use the tracked `.env.example` as the template;\ncopy it to `.env`, then add only credential values.\n\nA workspace is not configuration authority. Codewhale therefore ignores\nconfig/profile/home paths, provider/model/base-URL routing, MCP/plugin state,\napproval/sandbox/shell posture, executable paths, runtime settings, and every\nother non-credential `.env` entry. Variable expansion is rejected so a\nrepository cannot substitute an ambient secret into a credential value. Use\n`config.toml`, CLI flags, or values exported by the launching shell for those\nexplicit control-plane settings. `.env` is read from a stable regular-file\nhandle, is capped at 1 MiB, and symbolic links, reparse points, and multiply\nlinked files are rejected.\n\n## Reading and checking configuration from the CLI\n\n`codewhale config get <key>` reads scalar keys, whole tables such as `tools`,\nand nested paths such as `tools.user_input_timeout_seconds`. Displayed tables\nand nested values apply the same recursive credential redaction as `config dump`.\n\n`config set` supports its named scalar keys and the provider, route, and\nnotification commands. It refuses a key that nothing reads and suggests the\nnearest real key (`config set calm_mod on` names `calm_mode`). A settings.toml\nkey such as `calm_mode` or `tool_collapse` is validated and written to the\nuser-global settings.toml, not config.toml; `config get` reads it back from\nthere even when an old config.toml copy (which nothing reads) is still present,\nand names that copy so you can `config unset` it. Settings keys have no project\nscope, so `--project` refuses them. A config.toml key is checked against the\ntype its reader expects and stored as a TOML boolean or number where the\nreader needs one (`yolo = true`, `max_subagents = 4`); `reasoning_effort`\naccepts the same aliases as `/effort` (#6563). Other dotted writes fail before\nmodifying the file and name the TOML table to edit. For example, set a tools\ntimeout in the file as:\n\n```toml\n[tools]\nuser_input_timeout_seconds = 0\n```\n\n`codewhale config doctor` checks credential presence and endpoint shape, and\nwarns about each config.toml root key that nothing reads: a settings.toml key\nleft in config.toml is named as misplaced, and any other unread key gets a\ndid-you-mean. Keys read by any runtime reader are not reported. The warnings do\nnot fail the check, and a clean result does not validate every runtime setting's\nvalue (#6083, #6563).\n\n## Usage and execution statistics\n\n`codewhale stats` is an alias of `codewhale metrics`; both use the same local,\nread-only backend. Use `--json` for the structured `stats` report. For example:\n\n```sh\ncodewhale stats --days 7 --models 10 --tools 20 --roles 10\ncodewhale metrics --project /path/to/workspace --json\n```\n\n`--days N` accepts a positive number of days and conflicts with `--since`.\n`--since 24h` and the existing duration syntax remain available. `--models N`,\n`--tools N` and `--roles N` select the top positive N rows by known tokens,\nidentified tool executions and recorded Fleet tasks respectively. These limits\nonly shorten the display; they do not remove stored receipts.\n\nThe report separates saved-document and Runtime views, frozen native-source\nmoney and token subtotals by actual model, retained tool-result history\nestimates and outcomes, completed compaction observations, and recorded Fleet\nroles. USD and CNY stay separate. A cumulative correction replaces its proven\npredecessor; copied Session/Runtime receipts and inherited Runtime fork rows\nare not added as new provider work. Conflicting source versions are shown\nseparately and excluded from the subtotal. Saved legacy display totals remain\nvisible as document projections without being charged again globally.\n\nCoverage is explicit. A missing quote, usage receipt, tool outcome, output\nshape or recovery link is unknown, not zero. Empty scalar usage from older\ntransport records cannot prove a reported zero. Token counts describe known\nreported data; tool history tokens are estimates of retained text, not bytes\nproven to have been replayed on a provider request. Latest prepared-request\ncomponents and source-linked compaction/role economics are unavailable where\nthey were not captured. The report does not infer provider invoices or prove\ncomplete historical coverage.\n\n`--project WORKSPACE` filters current saved/thread/Fleet views. Original source\nworkspace is not yet captured, so native money and model subtotals are\nunavailable for that filter; mutable current metadata is not used to assign a\ncharge. Older receipt projections without reliable workspace scope are also\nunavailable under this filter. Without it, those legacy receipt views remain\nseparate fromLine truncated
+        "sha256": "fb934929d28d87dd161a26c427ffa21595ac358d994a46806c53f84fd49cb0f8",
+        "content": "# Configuration\n\ncodewhale reads configuration from a TOML file plus environment variables.\nAt process startup it may also load literal built-in-provider credentials from\na workspace-local `.env` file. Use the tracked `.env.example` as the template;\ncopy it to `.env`, then add only credential values.\n\nA workspace is not configuration authority. Codewhale therefore ignores\nconfig/profile/home paths, provider/model/base-URL routing, MCP/plugin state,\napproval/sandbox/shell posture, executable paths, runtime settings, and every\nother non-credential `.env` entry. Variable expansion is rejected so a\nrepository cannot substitute an ambient secret into a credential value. Use\n`config.toml`, CLI flags, or values exported by the launching shell for those\nexplicit control-plane settings. `.env` is read from a stable regular-file\nhandle, is capped at 1 MiB, and symbolic links, reparse points, and multiply\nlinked files are rejected.\n\n## Reading and checking configuration from the CLI\n\n`codewhale config get <key>` reads scalar keys, whole tables such as `tools`,\nand nested paths such as `tools.user_input_timeout_seconds`. Displayed tables\nand nested values apply the same recursive credential redaction as `config dump`.\n\n`config set` supports its named scalar keys and the provider, route, and\nnotification commands. It refuses a key that nothing reads and suggests the\nnearest real key (`config set calm_mod on` names `calm_mode`). A settings.toml\nkey such as `calm_mode` or `tool_collapse` is validated and written to the\nuser-global settings.toml, not config.toml; `config get` reads it back from\nthere even when an old config.toml copy (which nothing reads) is still present,\nand names that copy so you can `config unset` it. Settings keys have no project\nscope, so `--project` refuses them. A config.toml key is checked against the\ntype its reader expects and stored as a TOML boolean or number where the\nreader needs one (`yolo = true`, `max_subagents = 4`); `reasoning_effort`\naccepts the same aliases as `/effort` (#6563). Other dotted writes fail before\nmodifying the file and name the TOML table to edit. For example, set a tools\ntimeout in the file as:\n\n```toml\n[tools]\nuser_input_timeout_seconds = 0\n```\n\n`codewhale config doctor` checks credential presence and endpoint shape, and\nwarns about each config.toml root key that nothing reads: a settings.toml key\nleft in config.toml is named as misplaced, and any other unread key gets a\ndid-you-mean. Keys read by any runtime reader are not reported. The warnings do\nnot fail the check, and a clean result does not validate every runtime setting's\nvalue (#6083, #6563).\n\n## Usage and execution statistics\n\n`codewhale stats` is an alias of `codewhale metrics`; both use the same local,\nread-only backend. Use `--json` for the structured `stats` report. For example:\n\n```sh\ncodewhale stats --days 7 --models 10 --tools 20 --roles 10\ncodewhale metrics --project /path/to/workspace --json\n```\n\n`--days N` accepts a positive number of days and conflicts with `--since`.\n`--since 24h` and the existing duration syntax remain available. `--models N`,\n`--tools N` and `--roles N` select the top positive N rows by known tokens,\nidentified tool executions and recorded Fleet tasks respectively. These limits\nonly shorten the display; they do not remove stored receipts.\n\nThe report separates saved-document and Runtime views, frozen native-source\nmoney and token subtotals by actual model, retained tool-result history\nestimates and outcomes, completed compaction observations, and recorded Fleet\nroles. USD and CNY stay separate. A cumulative correction replaces its proven\npredecessor; copied Session/Runtime receipts and inherited Runtime fork rows\nare not added as new provider work. Conflicting source versions are shown\nseparately and excluded from the subtotal. Saved legacy display totals remain\nvisible as document projections without being charged again globally.\n\nCoverage is explicit. A missing quote, usage receipt, tool outcome, output\nshape or recovery link is unknown, not zero. Empty scalar usage from older\ntransport records cannot prove a reported zero. Token counts describe known\nreported data; tool history tokens are estimates of retained text, not bytes\nproven to have been replayed on a provider request. Latest prepared-request\ncomponents and source-linked compaction/role economics are unavailable where\nthey were not captured. The report does not infer provider invoices or prove\ncomplete historical coverage.\n\n`--project WORKSPACE` filters current saved/thread/Fleet views. Original source\nworkspace is not yet captured, so native money and model subtotals are\nunavailable for that filter; mutable current metadata is not used to assign a\ncharge. Older receipt projections without reliable workspace scope are also\nunavailable under this filter. Without it, those legacy receipt views remain\nseparate fromLine truncated
       }
     ]
   },
@@ -297,8 +297,8 @@
       },
       {
         "path": "docs/CONFIGURATION.md",
-        "sha256": "00f8d3b2b40dc4145a78b5bf93bd6d4e96c83478fdb5b4e0b7316108926ba1b7",
-        "content": "# Configuration\n\ncodewhale reads configuration from a TOML file plus environment variables.\nAt process startup it may also load literal built-in-provider credentials from\na workspace-local `.env` file. Use the tracked `.env.example` as the template;\ncopy it to `.env`, then add only credential values.\n\nA workspace is not configuration authority. Codewhale therefore ignores\nconfig/profile/home paths, provider/model/base-URL routing, MCP/plugin state,\napproval/sandbox/shell posture, executable paths, runtime settings, and every\nother non-credential `.env` entry. Variable expansion is rejected so a\nrepository cannot substitute an ambient secret into a credential value. Use\n`config.toml`, CLI flags, or values exported by the launching shell for those\nexplicit control-plane settings. `.env` is read from a stable regular-file\nhandle, is capped at 1 MiB, and symbolic links, reparse points, and multiply\nlinked files are rejected.\n\n## Reading and checking configuration from the CLI\n\n`codewhale config get <key>` reads scalar keys, whole tables such as `tools`,\nand nested paths such as `tools.user_input_timeout_seconds`. Displayed tables\nand nested values apply the same recursive credential redaction as `config dump`.\n\n`config set` supports its named scalar keys and the provider, route, and\nnotification commands. It refuses a key that nothing reads and suggests the\nnearest real key (`config set calm_mod on` names `calm_mode`). A settings.toml\nkey such as `calm_mode` or `tool_collapse` is validated and written to the\nuser-global settings.toml, not config.toml; `config get` reads it back from\nthere even when an old config.toml copy (which nothing reads) is still present,\nand names that copy so you can `config unset` it. Settings keys have no project\nscope, so `--project` refuses them. A config.toml key is checked against the\ntype its reader expects and stored as a TOML boolean or number where the\nreader needs one (`yolo = true`, `max_subagents = 4`); `reasoning_effort`\naccepts the same aliases as `/effort` (#6563). Other dotted writes fail before\nmodifying the file and name the TOML table to edit. For example, set a tools\ntimeout in the file as:\n\n```toml\n[tools]\nuser_input_timeout_seconds = 0\n```\n\n`codewhale config doctor` checks credential presence and endpoint shape, and\nwarns about each config.toml root key that nothing reads: a settings.toml key\nleft in config.toml is named as misplaced, and any other unread key gets a\ndid-you-mean. Keys read by any runtime reader are not reported. The warnings do\nnot fail the check, and a clean result does not validate every runtime setting's\nvalue (#6083, #6563).\n\n## Usage and execution statistics\n\n`codewhale stats` is an alias of `codewhale metrics`; both use the same local,\nread-only backend. Use `--json` for the structured `stats` report. For example:\n\n```sh\ncodewhale stats --days 7 --models 10 --tools 20 --roles 10\ncodewhale metrics --project /path/to/workspace --json\n```\n\n`--days N` accepts a positive number of days and conflicts with `--since`.\n`--since 24h` and the existing duration syntax remain available. `--models N`,\n`--tools N` and `--roles N` select the top positive N rows by known tokens,\nidentified tool executions and recorded Fleet tasks respectively. These limits\nonly shorten the display; they do not remove stored receipts.\n\nThe report separates saved-document and Runtime views, frozen native-source\nmoney and token subtotals by actual model, retained tool-result history\nestimates and outcomes, completed compaction observations, and recorded Fleet\nroles. USD and CNY stay separate. A cumulative correction replaces its proven\npredecessor; copied Session/Runtime receipts and inherited Runtime fork rows\nare not added as new provider work. Conflicting source versions are shown\nseparately and excluded from the subtotal. Saved legacy display totals remain\nvisible as document projections without being charged again globally.\n\nCoverage is explicit. A missing quote, usage receipt, tool outcome, output\nshape or recovery link is unknown, not zero. Empty scalar usage from older\ntransport records cannot prove a reported zero. Token counts describe known\nreported data; tool history tokens are estimates of retained text, not bytes\nproven to have been replayed on a provider request. Latest prepared-request\ncomponents and source-linked compaction/role economics are unavailable where\nthey were not captured. The report does not infer provider invoices or prove\ncomplete historical coverage.\n\n`--project WORKSPACE` filters current saved/thread/Fleet views. Original source\nworkspace is not yet captured, so native money and model subtotals are\nunavailable for that filter; mutable current metadata is not used to assign a\ncharge. Older receipt projections without reliable workspace scope are also\nunavailable under this filter. Without it, those legacy receipt views remain\nseparate fromLine truncated
+        "sha256": "fb934929d28d87dd161a26c427ffa21595ac358d994a46806c53f84fd49cb0f8",
+        "content": "# Configuration\n\ncodewhale reads configuration from a TOML file plus environment variables.\nAt process startup it may also load literal built-in-provider credentials from\na workspace-local `.env` file. Use the tracked `.env.example` as the template;\ncopy it to `.env`, then add only credential values.\n\nA workspace is not configuration authority. Codewhale therefore ignores\nconfig/profile/home paths, provider/model/base-URL routing, MCP/plugin state,\napproval/sandbox/shell posture, executable paths, runtime settings, and every\nother non-credential `.env` entry. Variable expansion is rejected so a\nrepository cannot substitute an ambient secret into a credential value. Use\n`config.toml`, CLI flags, or values exported by the launching shell for those\nexplicit control-plane settings. `.env` is read from a stable regular-file\nhandle, is capped at 1 MiB, and symbolic links, reparse points, and multiply\nlinked files are rejected.\n\n## Reading and checking configuration from the CLI\n\n`codewhale config get <key>` reads scalar keys, whole tables such as `tools`,\nand nested paths such as `tools.user_input_timeout_seconds`. Displayed tables\nand nested values apply the same recursive credential redaction as `config dump`.\n\n`config set` supports its named scalar keys and the provider, route, and\nnotification commands. It refuses a key that nothing reads and suggests the\nnearest real key (`config set calm_mod on` names `calm_mode`). A settings.toml\nkey such as `calm_mode` or `tool_collapse` is validated and written to the\nuser-global settings.toml, not config.toml; `config get` reads it back from\nthere even when an old config.toml copy (which nothing reads) is still present,\nand names that copy so you can `config unset` it. Settings keys have no project\nscope, so `--project` refuses them. A config.toml key is checked against the\ntype its reader expects and stored as a TOML boolean or number where the\nreader needs one (`yolo = true`, `max_subagents = 4`); `reasoning_effort`\naccepts the same aliases as `/effort` (#6563). Other dotted writes fail before\nmodifying the file and name the TOML table to edit. For example, set a tools\ntimeout in the file as:\n\n```toml\n[tools]\nuser_input_timeout_seconds = 0\n```\n\n`codewhale config doctor` checks credential presence and endpoint shape, and\nwarns about each config.toml root key that nothing reads: a settings.toml key\nleft in config.toml is named as misplaced, and any other unread key gets a\ndid-you-mean. Keys read by any runtime reader are not reported. The warnings do\nnot fail the check, and a clean result does not validate every runtime setting's\nvalue (#6083, #6563).\n\n## Usage and execution statistics\n\n`codewhale stats` is an alias of `codewhale metrics`; both use the same local,\nread-only backend. Use `--json` for the structured `stats` report. For example:\n\n```sh\ncodewhale stats --days 7 --models 10 --tools 20 --roles 10\ncodewhale metrics --project /path/to/workspace --json\n```\n\n`--days N` accepts a positive number of days and conflicts with `--since`.\n`--since 24h` and the existing duration syntax remain available. `--models N`,\n`--tools N` and `--roles N` select the top positive N rows by known tokens,\nidentified tool executions and recorded Fleet tasks respectively. These limits\nonly shorten the display; they do not remove stored receipts.\n\nThe report separates saved-document and Runtime views, frozen native-source\nmoney and token subtotals by actual model, retained tool-result history\nestimates and outcomes, completed compaction observations, and recorded Fleet\nroles. USD and CNY stay separate. A cumulative correction replaces its proven\npredecessor; copied Session/Runtime receipts and inherited Runtime fork rows\nare not added as new provider work. Conflicting source versions are shown\nseparately and excluded from the subtotal. Saved legacy display totals remain\nvisible as document projections without being charged again globally.\n\nCoverage is explicit. A missing quote, usage receipt, tool outcome, output\nshape or recovery link is unknown, not zero. Empty scalar usage from older\ntransport records cannot prove a reported zero. Token counts describe known\nreported data; tool history tokens are estimates of retained text, not bytes\nproven to have been replayed on a provider request. Latest prepared-request\ncomponents and source-linked compaction/role economics are unavailable where\nthey were not captured. The report does not infer provider invoices or prove\ncomplete historical coverage.\n\n`--project WORKSPACE` filters current saved/thread/Fleet views. Original source\nworkspace is not yet captured, so native money and model subtotals are\nunavailable for that filter; mutable current metadata is not used to assign a\ncharge. Older receipt projections without reliable workspace scope are also\nunavailable under this filter. Without it, those legacy receipt views remain\nseparate fromLine truncated
       }
     ]
   },
@@ -334,8 +334,8 @@
     "sources": [
       {
         "path": "docs/RUNTIME_API.md",
-        "sha256": "c4f348c27735a7df47a79faf11a6d5e3284a1e67a76c9372e76a8809236af45a",
-        "content": "# Codewhale Engine Runtime API & Integration Contract\n\n> 阅读简体中文版：[zh_hans/RUNTIME_API.md](zh_hans/RUNTIME_API.md)。\n\n`codewhale app-server --http` exposes the [Codewhale Engine](ARCHITECTURE.md) through\nits canonical local Runtime API and control plane.\nLocal SDKs, mobile/remote-control clients, and editor integrations talk to it\ninstead of screen-scraping terminal output. It serves the full HTTP/SSE runtime\nAPI (`/v1/*`), a JSON-RPC control transport over stdio, and the phone-friendly\nmobile page. `codewhale doctor --json` provides machine-readable health, and\n`codewhale serve --acp` speaks the Agent Client Protocol over stdio for editors\nsuch as Zed.\n\n`codewhale serve --http` / `serve --mobile` remain as **compatibility aliases**\nfor `codewhale app-server --http` / `--mobile`; both launch the identical\nserver. Desktop, web, SDK and custom integrations should target\n`app-server --http`. `--stdio` and `--socket` expose compatibility controls.\nBare `app-server` selects initialized JSONL control over stdio. Local stdio,\nsocket and ACP owners do not bind a hidden TCP listener. Explicit `--http`\nretains HTTP/SSE and compatibility routes on the same Runtime owner.\n\n`codewhale exec` is the separate one-shot headless worker path (stream-json,\nfleet worker subprocess, CI primitive). It is not part of this API, but it\nshares the same runtime, provider/model resolution, permission profiles, and\nevent vocabulary.\n\nThis document is the stable integration contract for native workbench\napplications (and other local supervisors) that embed the Codewhale Engine.\n\n## Architecture\n\n```\nlocal supervisor / SDK / automation harness\n        │\n        ├─ codewhale app-server --http     → HTTP/SSE runtime API (/v1/*)        [canonical]\n        ├─ codewhale app-server --mobile   → runtime API + mobile control page\n        ├─ codewhale app-server --stdio    → JSON-RPC control transport over stdio\n        ├─ codewhale app-server --socket   → same JSON-RPC over a unix domain socket (desktop daemon)\n        ├─ codewhale doctor --json         → machine-readable health & capability\n        ├─ codewhale serve --acp           → ACP stdio agent for editors such as Zed\n        ├─ codewhale serve --mcp           → MCP stdio server\n        ├─ codewhale serve --http/--mobile → legacy aliases for `app-server --http/--mobile`\n        └─ codewhale exec [args]           → one-shot headless worker (stream-json)\n```\n\nThe engine runs as a local-only process. All APIs bind to `localhost` by\ndefault. No hosted relay, no provider-token custody, no secret leakage.\n\nFor the read-only record of what a thread or turn did, see\n[`docs/RECEIPTS.md`](RECEIPTS.md): `codewhale receipts` on the CLI and the\n`/receipt` routes under **Threads** below.\n\n## Runtime API entrypoints\n\n| Entry | Transport | Use |\n|---|---|---|\n| `codewhale web [--port 7878]` | HTTP/SSE on `127.0.0.1:7878` + embedded client | First-class loopback-only browser client; opens the default browser |\n| `codewhale app-server --http` | HTTP/SSE on `127.0.0.1:7878` | Full `/v1/*` runtime API (canonical) |\n| `codewhale app-server --mobile` | HTTP/SSE on loopback + `/mobile` | Runtime API + local mobile control page |\n| `codewhale app-server --stdio` | Initialized JSONL over stdio | Compatibility controls and explicit tool-free turns through the held manager; no hidden TCP listener or full Codex interaction channel. |\n| `codewhale app-server --socket [--socket-path P]` | RPC text messages over WebSocket on a `0600` Unix domain socket (macOS/Linux) or an owner-only named pipe (Windows) | Desktop daemon: multi-client, kernel peer identity checked, `daemon/attach` claim handshake. Windows native-app qualification remains open. |\n| `codewhale app-server` | Initialized JSONL over stdio | Same selection as `--stdio`; HTTP requires explicit `--http`. |\n| `codewhale serve --http` / `--mobile` | same server as `app-server --http`/`--mobile` | Compatibility aliases |\n\n`app-server --http` and `--mobile` launch the same mature runtime API server\nhistorically reached through `serve --http` — no routes or behavior changed, so\nevery endpoint documented below is identical across both entrypoints. The\nruntime API token is read from `--auth-token`, then `CODEWHALE_RUNTIME_TOKEN`,\nthen `DEEPSEEK_RUNTIME_TOKEN`; use `--insecure-no-auth` only with a loopback\nbind. The `serve` compatibility aliases keep their `--insecure` flag.\nBare `app-server` does not open an HTTP port. Selected attached HTTP listeners\nretain their explicit bearer and non-loopback security checks.\n\nInitialized stdio/private-socket controls support `thread/loaded/list` with\n`{cursor?: string | null, limit?: number | null}`. The result is\n`{data: string[], nextCursor: string | null}` from the held manager's loaded\nengine snapshot, including idle and archived engines still in memory. It dLine truncated
+        "sha256": "6f3b8885978ef2e1b33d0fdf66ad6ccf5c25464c1951b59658aec00745390db2",
+        "content": "# Codewhale Engine Runtime API & Integration Contract\n\n> 阅读简体中文版：[zh_hans/RUNTIME_API.md](zh_hans/RUNTIME_API.md)。\n\n`codewhale app-server --http` exposes the [Codewhale Engine](ARCHITECTURE.md) through\nits canonical local Runtime API and control plane.\nLocal SDKs, mobile/remote-control clients, and editor integrations talk to it\ninstead of screen-scraping terminal output. It serves the full HTTP/SSE runtime\nAPI (`/v1/*`), a JSON-RPC control transport over stdio, and the phone-friendly\nmobile page. `codewhale doctor --json` provides machine-readable health, and\n`codewhale serve --acp` speaks the Agent Client Protocol over stdio for editors\nsuch as Zed.\n\n`codewhale serve --http` / `serve --mobile` remain as **compatibility aliases**\nfor `codewhale app-server --http` / `--mobile`; both launch the identical\nserver. Desktop, web, SDK and custom integrations should target\n`app-server --http`. `--stdio` and `--socket` expose compatibility controls.\nBare `app-server` selects initialized JSONL control over stdio. Local stdio,\nsocket and ACP owners do not bind a hidden TCP listener. Explicit `--http`\nretains HTTP/SSE and compatibility routes on the same Runtime owner.\n\n`codewhale exec` is the separate one-shot headless worker path (stream-json,\nfleet worker subprocess, CI primitive). It is not part of this API, but it\nshares the same runtime, provider/model resolution, permission profiles, and\nevent vocabulary.\n\nThis document is the stable integration contract for native workbench\napplications (and other local supervisors) that embed the Codewhale Engine.\n\n## Architecture\n\n```\nlocal supervisor / SDK / automation harness\n        │\n        ├─ codewhale app-server --http     → HTTP/SSE runtime API (/v1/*)        [canonical]\n        ├─ codewhale app-server --mobile   → runtime API + mobile control page\n        ├─ codewhale app-server --stdio    → JSON-RPC control transport over stdio\n        ├─ codewhale app-server --socket   → same JSON-RPC over a unix domain socket (desktop daemon)\n        ├─ codewhale doctor --json         → machine-readable health & capability\n        ├─ codewhale serve --acp           → ACP stdio agent for editors such as Zed\n        ├─ codewhale serve --mcp           → MCP stdio server\n        ├─ codewhale serve --http/--mobile → legacy aliases for `app-server --http/--mobile`\n        └─ codewhale exec [args]           → one-shot headless worker (stream-json)\n```\n\nThe engine runs as a local-only process. All APIs bind to `localhost` by\ndefault. No hosted relay, no provider-token custody, no secret leakage.\n\nFor the read-only record of what a thread or turn did, see\n[`docs/RECEIPTS.md`](RECEIPTS.md): `codewhale receipts` on the CLI and the\n`/receipt` routes under **Threads** below.\n\n## Runtime API entrypoints\n\n| Entry | Transport | Use |\n|---|---|---|\n| `codewhale web [--port 7878]` | HTTP/SSE on `127.0.0.1:7878` + embedded client | First-class loopback-only browser client; opens the default browser |\n| `codewhale app-server --http` | HTTP/SSE on `127.0.0.1:7878` | Full `/v1/*` runtime API (canonical) |\n| `codewhale app-server --mobile` | HTTP/SSE on loopback + `/mobile` | Runtime API + local mobile control page |\n| `codewhale app-server --stdio` | Initialized JSONL over stdio | Compatibility controls and explicit tool-free turns through the held manager; no hidden TCP listener or full Codex interaction channel. |\n| `codewhale app-server --socket [--socket-path P]` | RPC text messages over WebSocket on a `0600` Unix domain socket (macOS/Linux) or an owner-only named pipe (Windows) | Desktop daemon: multi-client, kernel peer identity checked, `daemon/attach` claim handshake. Windows native-app qualification remains open. |\n| `codewhale app-server` | Initialized JSONL over stdio | Same selection as `--stdio`; HTTP requires explicit `--http`. |\n| `codewhale serve --http` / `--mobile` | same server as `app-server --http`/`--mobile` | Compatibility aliases |\n\n`app-server --http` and `--mobile` launch the same mature runtime API server\nhistorically reached through `serve --http` — no routes or behavior changed, so\nevery endpoint documented below is identical across both entrypoints. The\nruntime API token is read from `--auth-token`, then `CODEWHALE_RUNTIME_TOKEN`,\nthen `DEEPSEEK_RUNTIME_TOKEN`; use `--insecure-no-auth` only with a loopback\nbind. The `serve` compatibility aliases keep their `--insecure` flag.\nBare `app-server` does not open an HTTP port. Selected attached HTTP listeners\nretain their explicit bearer and non-loopback security checks.\n\nInitialized stdio/private-socket controls support `thread/loaded/list` with\n`{cursor?: string | null, limit?: number | null}`. The result is\n`{data: string[], nextCursor: string | null}` from the held manager's loaded\nengine snapshot, including idle and archived engines still in memory. It dLine truncated
       }
     ]
   },
@@ -377,8 +377,8 @@
     "sources": [
       {
         "path": "CONTRIBUTING.md",
-        "sha256": "7b3e523585eb5535555b1d8aaa66153ce0451b1cedc165dfba212e372dd71d41",
-        "content": "# Contributing to Codewhale\n\nCodewhale is what it is because of the people who used it, reported what broke,\ntranslated it, added their providers and sent fixes. Thank you for being here.\nEveryone is welcome, in English or 中文, with or without code.\n\nAll feedback is a gift to this project, including \"this looked wrong\" or \"I got\nlost here\" with no fix attached. The [roadmap](docs/ROADMAP.md) says where\nCodewhale is going and where help counts most.\n\n## Where things go\n\n| You have | Best place |\n| --- | --- |\n| A bug, a crash, or behavior that surprised you | [Bug report](https://github.com/codewhale-hq/CodeWhale/issues/new?template=bug_report.yml) |\n| An idea or a feature you want | [Feature request](https://github.com/codewhale-hq/CodeWhale/issues/new?template=feature_request.yml), or [Ideas](https://github.com/codewhale-hq/CodeWhale/discussions/categories/ideas) to talk it through first |\n| Feedback on how something looks, reads or feels | [Feedback](https://github.com/codewhale-hq/CodeWhale/issues/new?template=feedback.yml) |\n| A question | [Q&A](https://github.com/codewhale-hq/CodeWhale/discussions/categories/q-a) or [Discord](https://discord.gg/37gfS3ksug) |\n| A security problem | A [private advisory](https://github.com/codewhale-hq/CodeWhale/security/advisories/new), never a public issue ([policy](.github/SECURITY.md)) |\n| A fix for an open issue | A pull request that links the issue |\n| A translation, a provider or model, a docs fix | A pull request, no issue needed ([standing projects](#standing-projects)) |\n| A new feature, already written or not | An issue first ([why](#new-features-start-as-an-issue)) |\n\nIf you are not sure, open an issue. Nothing is filed in the wrong place for\nlong: a maintainer moves it.\n\n## New features start as an issue\n\nIssues are the best place for feature requests, ideas and bugs. A pull request\nis the best place for work we have already agreed on.\n\nA feature that first appears as a pull request takes longer to land than one\nthat starts as an issue, even when we want it. The questions that decide a\nfeature are about the product, not the diff: does it belong in the engine or in\na plugin, what does it do in each permission mode, how does it show up in\nreceipts, what is it called in 19 languages. Answering those in a review means\nrewriting code you already finished. Answering them in an issue costs you a few\nsentences.\n\nSo for a new feature:\n\n1. Open a [feature request](https://github.com/codewhale-hq/CodeWhale/issues/new?template=feature_request.yml).\n   Say what you are trying to do and what gets in the way. A proposed design is\n   welcome and optional.\n2. Tick \"I would like to open a pull request\" if you want to build it.\n3. A maintainer answers with the shape that fits, or with what is still\n   undecided. Then the pull request is small and merges quickly.\n\nA feature pull request that arrives without an issue is still read, and it is\nnever closed for that reason. We will most likely ask the same questions there\nand it will take longer.\n\n## Standing projects\n\nThese are always open, and a pull request needs no issue first:\n\n- **Translations.** You can improve the terminal language packs, the translated\n  READMEs in `docs/i18n/`, the website, and the translated docs in\n  `docs/zh_hans/` and `docs/id/`. The [localization matrix](docs/LOCALIZATION.md)\n  shows what each language has and how to add one, and the\n  [Chinese localization group](https://github.com/codewhale-hq/CodeWhale/issues/6804)\n  is recruiting.\n- **Providers and models.** You can add a missing provider or a new model, or\n  correct a price, a context size or a provider quirk.\n  [Providers](docs/PROVIDERS.md) lists what ships, and\n  `python3 scripts/check-provider-registry.py` checks your change.\n- **Documentation.** You can fix a typo, an unclear step, a missing example or\n  an instruction that no longer matches the product.\n- **Bug fixes.** If there is no issue yet, describe the bug in the pull request\n  by saying what you ran, what happened and what you expected.\n- **Planned work.** You can take anything in a\n  [milestone](docs/ROADMAP.md#releases) or anything labelled\n  [help wanted](https://github.com/codewhale-hq/CodeWhale/issues?q=is%3Aissue+is%3Aopen+label%3A%22help+wanted%22).\n  Comment on the issue to claim it so two people do not build the same thing.\n\nLarge changes to planned work are easier in layers; see\n[Layered and EPIC-Sized Work](#layered-and-epic-sized-work).\n\n### What a planned-work issue gives you\n\nEvery issue a maintainer opens for planned work, and every release tracking issue, has the same parts, so you can pick one up without asking first:\n\n- **Status:** what is on `main` already, with commit links, and what is still missing.\n- **Ways to help:** each piece open to contributors, with the files to read, the command to test it locally (usually `scripts/dev-test.sh <crate> <filter>Line truncated
+        "sha256": "096a609f93d82c8bcea9f967072d5306807cd4375390218f9d3784023df0dcb0",
+        "content": "# Contributing to Codewhale\n\nCodewhale is what it is because of the people who used it, reported what broke,\ntranslated it, added their providers and sent fixes. Thank you for being here.\nEveryone is welcome, in English or 中文, with or without code.\n\nAll feedback is a gift to this project, including \"this looked wrong\" or \"I got\nlost here\" with no fix attached. The [roadmap](docs/ROADMAP.md) says where\nCodewhale is going and where help counts most.\n\n## Where things go\n\n| You have | Best place |\n| --- | --- |\n| A bug, a crash, or behavior that surprised you | [Bug report](https://github.com/codewhale-hq/CodeWhale/issues/new?template=bug_report.yml) |\n| An idea or a feature you want | [Feature request](https://github.com/codewhale-hq/CodeWhale/issues/new?template=feature_request.yml), or [Ideas](https://github.com/codewhale-hq/CodeWhale/discussions/categories/ideas) to talk it through first |\n| Feedback on how something looks, reads or feels | [Feedback](https://github.com/codewhale-hq/CodeWhale/issues/new?template=feedback.yml) |\n| A question | [Q&A](https://github.com/codewhale-hq/CodeWhale/discussions/categories/q-a) or [Discord](https://discord.gg/37gfS3ksug) |\n| A security problem | A [private advisory](https://github.com/codewhale-hq/CodeWhale/security/advisories/new), never a public issue ([policy](.github/SECURITY.md)) |\n| A fix for an open issue | A pull request that links the issue |\n| A translation, a provider or model, a docs fix | A pull request, no issue needed ([standing projects](#standing-projects)) |\n| A new feature, already written or not | An issue first ([why](#new-features-start-as-an-issue)) |\n\nIf you are not sure, open an issue. Nothing is filed in the wrong place for\nlong: a maintainer moves it.\n\n## New features start as an issue\n\nIssues are the best place for feature requests, ideas and bugs. A pull request\nis the best place for work we have already agreed on.\n\nA feature that first appears as a pull request takes longer to land than one\nthat starts as an issue, even when we want it. The questions that decide a\nfeature are about the product, not the diff: does it belong in the engine or in\na plugin, what does it do in each permission mode, how does it show up in\nreceipts, what is it called in 19 languages. Answering those in a review means\nrewriting code you already finished. Answering them in an issue costs you a few\nsentences.\n\nSo for a new feature:\n\n1. Open a [feature request](https://github.com/codewhale-hq/CodeWhale/issues/new?template=feature_request.yml).\n   Say what you are trying to do and what gets in the way. A proposed design is\n   welcome and optional.\n2. Tick \"I would like to open a pull request\" if you want to build it.\n3. A maintainer answers with the shape that fits, or with what is still\n   undecided. Then the pull request is small and merges quickly.\n\nA feature pull request that arrives without an issue is still read, and it is\nnever closed for that reason. We will most likely ask the same questions there\nand it will take longer.\n\n## Standing projects\n\nThese are always open, and a pull request needs no issue first:\n\n- **Translations.** You can improve the terminal language packs, the translated\n  READMEs in `docs/i18n/`, the website, and the translated docs in\n  `docs/zh_hans/` and `docs/id/`. The [localization matrix](docs/LOCALIZATION.md)\n  shows what each language has and how to add one, and the\n  [Chinese localization group](https://github.com/codewhale-hq/CodeWhale/issues/6804)\n  is recruiting.\n- **Providers and models.** You can add a missing provider or a new model, or\n  correct a price, a context size or a provider quirk.\n  [Providers](docs/PROVIDERS.md) lists what ships, and\n  `python3 scripts/check-provider-registry.py` checks your change.\n- **Documentation.** You can fix a typo, an unclear step, a missing example or\n  an instruction that no longer matches the product.\n- **Bug fixes.** If there is no issue yet, describe the bug in the pull request\n  by saying what you ran, what happened and what you expected.\n- **Planned work.** You can take anything in a\n  [milestone](docs/ROADMAP.md#releases) or anything labelled\n  [help wanted](https://github.com/codewhale-hq/CodeWhale/issues?q=is%3Aissue+is%3Aopen+label%3A%22help+wanted%22).\n  Comment on the issue to claim it so two people do not build the same thing.\n\nLarge changes to planned work are easier in layers; see\n[Layered and EPIC-Sized Work](#layered-and-epic-sized-work).\n\n### What a planned-work issue gives you\n\nEvery issue a maintainer opens for planned work, and every release tracking issue, has the same parts, so you can pick one up without asking first:\n\n- **Status:** what is on `main` already, with commit links, and what is still missing.\n- **Ways to help:** each piece open to contributors, with the files to read, the command to test it locally (usually `scripts/dev-test.sh <crate> <filter>Line truncated
       },
       {
         "path": "docs/AGENT_ETHOS.md",
@@ -387,8 +387,8 @@
       },
       {
         "path": "docs/CONTRIBUTORS.md",
-        "sha256": "bedd1cc62969108856f75d823d13bde9f7217a936d4b0e7af3d2e6232cad5770",
-        "content": "# Contributors\n\nCodewhale is built in the open with a growing community of contributors. Every\nissue report and pull request is real project work — welcome at any experience\nlevel. This is the full per-PR contributor record, grouped into release/time\nbands with the newest band first so it stays scannable. Expand any band to see\neveryone.\n\nFor the live list, see the\n[GitHub contributors page](https://github.com/codewhale-hq/CodeWhale/graphs/contributors),\n[`AUTHOR_MAP`](https://github.com/codewhale-hq/CodeWhale/blob/main/.github/AUTHOR_MAP),\nand [CHANGELOG.md](../CHANGELOG.md).\n\n<!-- contributor-wall:start -->\n## Everyone\n\n[![433 people built Codewhale](assets/contributors.png)](#everyone)\n\n433 people have code, co-authored commits, carried-forward pull requests or credited reports in Codewhale. GitHub's contributor list shows only commit authors whose email is linked to their account, so it counts 212 of them. If your commits are missing there, add your commit email to your GitHub account and they will link.\n\n[0gl20shk0sbt36](https://github.com/0gl20shk0sbt36) · [1Git2Clone](https://github.com/1Git2Clone) · [20bytes](https://github.com/20bytes) · [47Cid](https://github.com/47Cid) · [7jrxt42BxFZo4iAnN4CX](https://github.com/7jrxt42BxFZo4iAnN4CX) · [8sero](https://github.com/8sero) · [A-Corner](https://github.com/A-Corner) · [aboimpinto](https://github.com/aboimpinto) · [Aboysky](https://github.com/Aboysky) · [adity982](https://github.com/adity982) · [AdityaVG13](https://github.com/AdityaVG13) · [AGSaturn](https://github.com/AGSaturn) · [ai-dengwsh](https://github.com/ai-dengwsh) · [aiapienthusiast](https://github.com/aiapienthusiast) · [Aitensa](https://github.com/Aitensa) · [AiurArtanis](https://github.com/AiurArtanis) · [alitvak69](https://github.com/alitvak69) · [all-lopezg](https://github.com/all-lopezg) · [alvin1](https://github.com/alvin1) · [amuthantamil](https://github.com/amuthantamil) · [AnaheimEX](https://github.com/AnaheimEX) · [Andrea-Bruno](https://github.com/Andrea-Bruno) · [Angel-Hair](https://github.com/Angel-Hair) · [angziii](https://github.com/angziii) · [AnonymousUser443](https://github.com/AnonymousUser443) · [antarikshraya](https://github.com/antarikshraya) · [Anyexyz](https://github.com/Anyexyz) · [Apeiron0w0](https://github.com/Apeiron0w0) · [aqilaziz](https://github.com/aqilaziz) · [AresNing](https://github.com/AresNing) · [ArronAI007](https://github.com/ArronAI007) · [Artenx](https://github.com/Artenx) · [asdfg314284230](https://github.com/asdfg314284230) · [Asteria1028](https://github.com/Asteria1028) · [asto18089](https://github.com/asto18089) · [atmosuwiryo](https://github.com/atmosuwiryo) · [axobase001](https://github.com/axobase001) · [aznikline](https://github.com/aznikline) · [baendlorel](https://github.com/baendlorel) · [banqii](https://github.com/banqii) · [Beltran12138](https://github.com/Beltran12138) · [bevis-wong](https://github.com/bevis-wong) · [BH8GCJ](https://github.com/BH8GCJ) · [BigBenLabs](https://github.com/BigBenLabs) · [Biilow](https://github.com/Biilow) · [bistack](https://github.com/bistack) · [BrathonBai](https://github.com/BrathonBai) · [bruce6135](https://github.com/bruce6135) · [BryonGo](https://github.com/BryonGo) · [buiducnhat](https://github.com/buiducnhat) · [buko](https://github.com/buko) · [BX166](https://github.com/BX166) · [c020627](https://github.com/c020627) · [cacdcaecawae](https://github.com/cacdcaecawae) · [caeserchen](https://github.com/caeserchen) · [caiyilian](https://github.com/caiyilian) · [CCChisato](https://github.com/CCChisato) · [ccomma](https://github.com/ccomma) · [cenab](https://github.com/cenab) · [ChaceLyee2101](https://github.com/ChaceLyee2101) · [chatman-media](https://github.com/chatman-media) · [chennest](https://github.com/chennest) · [chinaqy110](https://github.com/chinaqy110) · [chnjames](https://github.com/chnjames) · [chuntseevolving](https://github.com/chuntseevolving) · [ci4ic4](https://github.com/ci4ic4) · [cian-lien](https://github.com/cian-lien) · [cmdcorp6534](https://github.com/cmdcorp6534) · [cmyyy](https://github.com/cmyyy) · [codepgq](https://github.com/codepgq) · [Colorful-glassblock](https://github.com/Colorful-glassblock) · [ComeFromTheMars](https://github.com/ComeFromTheMars) · [CrepuscularIRIS](https://github.com/CrepuscularIRIS) · [ct-jaryn](https://github.com/ct-jaryn) · [ctxyao](https://github.com/ctxyao) · [cy2311](https://github.com/cy2311) · [cyq1017](https://github.com/cyq1017) · [czf0718](https://github.com/czf0718) · [dajiaohuang](https://github.com/dajiaohuang) · [DarrellThomas](https://github.com/DarrellThomas) · [ddlw](https://github.com/ddlw) · [demian-welt](https://github.com/demian-welt) · [Dev-next-gen](https://github.com/Dev-next-gen) · [dfwqdyl-ui](https://github.com/dfwqdyl-ui) · [DI-HUO-MING-YI](https://github.com/DI-HUO-MING-YI) · [DieMoe233](https://github.Line truncated
+        "sha256": "4368939ea860f6db9484fd86e613574b72eca71361f08a1130ec87c5f5c66c78",
+        "content": "# Contributors\n\nCodewhale is built in the open with a growing community of contributors. Every\nissue report and pull request is real project work — welcome at any experience\nlevel. This is the full per-PR contributor record, grouped into release/time\nbands with the newest band first so it stays scannable. Expand any band to see\neveryone.\n\nFor the live list, see the\n[GitHub contributors page](https://github.com/codewhale-hq/CodeWhale/graphs/contributors),\n[`AUTHOR_MAP`](https://github.com/codewhale-hq/CodeWhale/blob/main/.github/AUTHOR_MAP),\nand [CHANGELOG.md](../CHANGELOG.md).\n\n<!-- contributor-wall:start -->\n## Everyone\n\n[![433 people built Codewhale](assets/contributors.png)](#everyone)\n\n433 people have code, co-authored commits, carried-forward pull requests or credited reports in Codewhale. GitHub's contributor list shows only commit authors whose email is linked to their account, so it counts 212 of them. If your commits are missing there, add your commit email to your GitHub account and they will link.\n\n[0gl20shk0sbt36](https://github.com/0gl20shk0sbt36) · [1Git2Clone](https://github.com/1Git2Clone) · [20bytes](https://github.com/20bytes) · [47Cid](https://github.com/47Cid) · [7jrxt42BxFZo4iAnN4CX](https://github.com/7jrxt42BxFZo4iAnN4CX) · [8sero](https://github.com/8sero) · [A-Corner](https://github.com/A-Corner) · [aboimpinto](https://github.com/aboimpinto) · [Aboysky](https://github.com/Aboysky) · [adity982](https://github.com/adity982) · [AdityaVG13](https://github.com/AdityaVG13) · [AGSaturn](https://github.com/AGSaturn) · [ai-dengwsh](https://github.com/ai-dengwsh) · [aiapienthusiast](https://github.com/aiapienthusiast) · [Aitensa](https://github.com/Aitensa) · [AiurArtanis](https://github.com/AiurArtanis) · [alitvak69](https://github.com/alitvak69) · [all-lopezg](https://github.com/all-lopezg) · [alvin1](https://github.com/alvin1) · [amuthantamil](https://github.com/amuthantamil) · [AnaheimEX](https://github.com/AnaheimEX) · [Andrea-Bruno](https://github.com/Andrea-Bruno) · [Angel-Hair](https://github.com/Angel-Hair) · [angziii](https://github.com/angziii) · [AnonymousUser443](https://github.com/AnonymousUser443) · [antarikshraya](https://github.com/antarikshraya) · [Anyexyz](https://github.com/Anyexyz) · [Apeiron0w0](https://github.com/Apeiron0w0) · [aqilaziz](https://github.com/aqilaziz) · [AresNing](https://github.com/AresNing) · [ArronAI007](https://github.com/ArronAI007) · [Artenx](https://github.com/Artenx) · [asdfg314284230](https://github.com/asdfg314284230) · [Asteria1028](https://github.com/Asteria1028) · [asto18089](https://github.com/asto18089) · [atmosuwiryo](https://github.com/atmosuwiryo) · [axobase001](https://github.com/axobase001) · [aznikline](https://github.com/aznikline) · [baendlorel](https://github.com/baendlorel) · [banqii](https://github.com/banqii) · [Beltran12138](https://github.com/Beltran12138) · [bevis-wong](https://github.com/bevis-wong) · [BH8GCJ](https://github.com/BH8GCJ) · [BigBenLabs](https://github.com/BigBenLabs) · [Biilow](https://github.com/Biilow) · [bistack](https://github.com/bistack) · [BrathonBai](https://github.com/BrathonBai) · [bruce6135](https://github.com/bruce6135) · [BryonGo](https://github.com/BryonGo) · [buiducnhat](https://github.com/buiducnhat) · [buko](https://github.com/buko) · [BX166](https://github.com/BX166) · [c020627](https://github.com/c020627) · [cacdcaecawae](https://github.com/cacdcaecawae) · [caeserchen](https://github.com/caeserchen) · [caiyilian](https://github.com/caiyilian) · [CCChisato](https://github.com/CCChisato) · [ccomma](https://github.com/ccomma) · [cenab](https://github.com/cenab) · [ChaceLyee2101](https://github.com/ChaceLyee2101) · [chatman-media](https://github.com/chatman-media) · [chennest](https://github.com/chennest) · [chinaqy110](https://github.com/chinaqy110) · [chnjames](https://github.com/chnjames) · [chuntseevolving](https://github.com/chuntseevolving) · [ci4ic4](https://github.com/ci4ic4) · [cian-lien](https://github.com/cian-lien) · [cmdcorp6534](https://github.com/cmdcorp6534) · [cmyyy](https://github.com/cmyyy) · [codepgq](https://github.com/codepgq) · [Colorful-glassblock](https://github.com/Colorful-glassblock) · [ComeFromTheMars](https://github.com/ComeFromTheMars) · [CrepuscularIRIS](https://github.com/CrepuscularIRIS) · [ct-jaryn](https://github.com/ct-jaryn) · [ctxyao](https://github.com/ctxyao) · [cy2311](https://github.com/cy2311) · [cyq1017](https://github.com/cyq1017) · [czf0718](https://github.com/czf0718) · [dajiaohuang](https://github.com/dajiaohuang) · [DarrellThomas](https://github.com/DarrellThomas) · [ddlw](https://github.com/ddlw) · [demian-welt](https://github.com/demian-welt) · [Dev-next-gen](https://github.com/Dev-next-gen) · [dfwqdyl-ui](https://github.com/dfwqdyl-ui) · [DI-HUO-MING-YI](https://github.com/DI-HUO-MING-YI) · [DieMoe233](https://github.Line truncated
       },
       {
         "path": "docs/RELEASE_CHECKLIST.md",
@@ -403,8 +403,8 @@
     "sources": [
       {
         "path": "docs/RUNTIME_API.md",
-        "sha256": "c4f348c27735a7df47a79faf11a6d5e3284a1e67a76c9372e76a8809236af45a",
-        "content": "# Codewhale Engine Runtime API & Integration Contract\n\n> 阅读简体中文版：[zh_hans/RUNTIME_API.md](zh_hans/RUNTIME_API.md)。\n\n`codewhale app-server --http` exposes the [Codewhale Engine](ARCHITECTURE.md) through\nits canonical local Runtime API and control plane.\nLocal SDKs, mobile/remote-control clients, and editor integrations talk to it\ninstead of screen-scraping terminal output. It serves the full HTTP/SSE runtime\nAPI (`/v1/*`), a JSON-RPC control transport over stdio, and the phone-friendly\nmobile page. `codewhale doctor --json` provides machine-readable health, and\n`codewhale serve --acp` speaks the Agent Client Protocol over stdio for editors\nsuch as Zed.\n\n`codewhale serve --http` / `serve --mobile` remain as **compatibility aliases**\nfor `codewhale app-server --http` / `--mobile`; both launch the identical\nserver. Desktop, web, SDK and custom integrations should target\n`app-server --http`. `--stdio` and `--socket` expose compatibility controls.\nBare `app-server` selects initialized JSONL control over stdio. Local stdio,\nsocket and ACP owners do not bind a hidden TCP listener. Explicit `--http`\nretains HTTP/SSE and compatibility routes on the same Runtime owner.\n\n`codewhale exec` is the separate one-shot headless worker path (stream-json,\nfleet worker subprocess, CI primitive). It is not part of this API, but it\nshares the same runtime, provider/model resolution, permission profiles, and\nevent vocabulary.\n\nThis document is the stable integration contract for native workbench\napplications (and other local supervisors) that embed the Codewhale Engine.\n\n## Architecture\n\n```\nlocal supervisor / SDK / automation harness\n        │\n        ├─ codewhale app-server --http     → HTTP/SSE runtime API (/v1/*)        [canonical]\n        ├─ codewhale app-server --mobile   → runtime API + mobile control page\n        ├─ codewhale app-server --stdio    → JSON-RPC control transport over stdio\n        ├─ codewhale app-server --socket   → same JSON-RPC over a unix domain socket (desktop daemon)\n        ├─ codewhale doctor --json         → machine-readable health & capability\n        ├─ codewhale serve --acp           → ACP stdio agent for editors such as Zed\n        ├─ codewhale serve --mcp           → MCP stdio server\n        ├─ codewhale serve --http/--mobile → legacy aliases for `app-server --http/--mobile`\n        └─ codewhale exec [args]           → one-shot headless worker (stream-json)\n```\n\nThe engine runs as a local-only process. All APIs bind to `localhost` by\ndefault. No hosted relay, no provider-token custody, no secret leakage.\n\nFor the read-only record of what a thread or turn did, see\n[`docs/RECEIPTS.md`](RECEIPTS.md): `codewhale receipts` on the CLI and the\n`/receipt` routes under **Threads** below.\n\n## Runtime API entrypoints\n\n| Entry | Transport | Use |\n|---|---|---|\n| `codewhale web [--port 7878]` | HTTP/SSE on `127.0.0.1:7878` + embedded client | First-class loopback-only browser client; opens the default browser |\n| `codewhale app-server --http` | HTTP/SSE on `127.0.0.1:7878` | Full `/v1/*` runtime API (canonical) |\n| `codewhale app-server --mobile` | HTTP/SSE on loopback + `/mobile` | Runtime API + local mobile control page |\n| `codewhale app-server --stdio` | Initialized JSONL over stdio | Compatibility controls and explicit tool-free turns through the held manager; no hidden TCP listener or full Codex interaction channel. |\n| `codewhale app-server --socket [--socket-path P]` | RPC text messages over WebSocket on a `0600` Unix domain socket (macOS/Linux) or an owner-only named pipe (Windows) | Desktop daemon: multi-client, kernel peer identity checked, `daemon/attach` claim handshake. Windows native-app qualification remains open. |\n| `codewhale app-server` | Initialized JSONL over stdio | Same selection as `--stdio`; HTTP requires explicit `--http`. |\n| `codewhale serve --http` / `--mobile` | same server as `app-server --http`/`--mobile` | Compatibility aliases |\n\n`app-server --http` and `--mobile` launch the same mature runtime API server\nhistorically reached through `serve --http` — no routes or behavior changed, so\nevery endpoint documented below is identical across both entrypoints. The\nruntime API token is read from `--auth-token`, then `CODEWHALE_RUNTIME_TOKEN`,\nthen `DEEPSEEK_RUNTIME_TOKEN`; use `--insecure-no-auth` only with a loopback\nbind. The `serve` compatibility aliases keep their `--insecure` flag.\nBare `app-server` does not open an HTTP port. Selected attached HTTP listeners\nretain their explicit bearer and non-loopback security checks.\n\nInitialized stdio/private-socket controls support `thread/loaded/list` with\n`{cursor?: string | null, limit?: number | null}`. The result is\n`{data: string[], nextCursor: string | null}` from the held manager's loaded\nengine snapshot, including idle and archived engines still in memory. It dLine truncated
+        "sha256": "6f3b8885978ef2e1b33d0fdf66ad6ccf5c25464c1951b59658aec00745390db2",
+        "content": "# Codewhale Engine Runtime API & Integration Contract\n\n> 阅读简体中文版：[zh_hans/RUNTIME_API.md](zh_hans/RUNTIME_API.md)。\n\n`codewhale app-server --http` exposes the [Codewhale Engine](ARCHITECTURE.md) through\nits canonical local Runtime API and control plane.\nLocal SDKs, mobile/remote-control clients, and editor integrations talk to it\ninstead of screen-scraping terminal output. It serves the full HTTP/SSE runtime\nAPI (`/v1/*`), a JSON-RPC control transport over stdio, and the phone-friendly\nmobile page. `codewhale doctor --json` provides machine-readable health, and\n`codewhale serve --acp` speaks the Agent Client Protocol over stdio for editors\nsuch as Zed.\n\n`codewhale serve --http` / `serve --mobile` remain as **compatibility aliases**\nfor `codewhale app-server --http` / `--mobile`; both launch the identical\nserver. Desktop, web, SDK and custom integrations should target\n`app-server --http`. `--stdio` and `--socket` expose compatibility controls.\nBare `app-server` selects initialized JSONL control over stdio. Local stdio,\nsocket and ACP owners do not bind a hidden TCP listener. Explicit `--http`\nretains HTTP/SSE and compatibility routes on the same Runtime owner.\n\n`codewhale exec` is the separate one-shot headless worker path (stream-json,\nfleet worker subprocess, CI primitive). It is not part of this API, but it\nshares the same runtime, provider/model resolution, permission profiles, and\nevent vocabulary.\n\nThis document is the stable integration contract for native workbench\napplications (and other local supervisors) that embed the Codewhale Engine.\n\n## Architecture\n\n```\nlocal supervisor / SDK / automation harness\n        │\n        ├─ codewhale app-server --http     → HTTP/SSE runtime API (/v1/*)        [canonical]\n        ├─ codewhale app-server --mobile   → runtime API + mobile control page\n        ├─ codewhale app-server --stdio    → JSON-RPC control transport over stdio\n        ├─ codewhale app-server --socket   → same JSON-RPC over a unix domain socket (desktop daemon)\n        ├─ codewhale doctor --json         → machine-readable health & capability\n        ├─ codewhale serve --acp           → ACP stdio agent for editors such as Zed\n        ├─ codewhale serve --mcp           → MCP stdio server\n        ├─ codewhale serve --http/--mobile → legacy aliases for `app-server --http/--mobile`\n        └─ codewhale exec [args]           → one-shot headless worker (stream-json)\n```\n\nThe engine runs as a local-only process. All APIs bind to `localhost` by\ndefault. No hosted relay, no provider-token custody, no secret leakage.\n\nFor the read-only record of what a thread or turn did, see\n[`docs/RECEIPTS.md`](RECEIPTS.md): `codewhale receipts` on the CLI and the\n`/receipt` routes under **Threads** below.\n\n## Runtime API entrypoints\n\n| Entry | Transport | Use |\n|---|---|---|\n| `codewhale web [--port 7878]` | HTTP/SSE on `127.0.0.1:7878` + embedded client | First-class loopback-only browser client; opens the default browser |\n| `codewhale app-server --http` | HTTP/SSE on `127.0.0.1:7878` | Full `/v1/*` runtime API (canonical) |\n| `codewhale app-server --mobile` | HTTP/SSE on loopback + `/mobile` | Runtime API + local mobile control page |\n| `codewhale app-server --stdio` | Initialized JSONL over stdio | Compatibility controls and explicit tool-free turns through the held manager; no hidden TCP listener or full Codex interaction channel. |\n| `codewhale app-server --socket [--socket-path P]` | RPC text messages over WebSocket on a `0600` Unix domain socket (macOS/Linux) or an owner-only named pipe (Windows) | Desktop daemon: multi-client, kernel peer identity checked, `daemon/attach` claim handshake. Windows native-app qualification remains open. |\n| `codewhale app-server` | Initialized JSONL over stdio | Same selection as `--stdio`; HTTP requires explicit `--http`. |\n| `codewhale serve --http` / `--mobile` | same server as `app-server --http`/`--mobile` | Compatibility aliases |\n\n`app-server --http` and `--mobile` launch the same mature runtime API server\nhistorically reached through `serve --http` — no routes or behavior changed, so\nevery endpoint documented below is identical across both entrypoints. The\nruntime API token is read from `--auth-token`, then `CODEWHALE_RUNTIME_TOKEN`,\nthen `DEEPSEEK_RUNTIME_TOKEN`; use `--insecure-no-auth` only with a loopback\nbind. The `serve` compatibility aliases keep their `--insecure` flag.\nBare `app-server` does not open an HTTP port. Selected attached HTTP listeners\nretain their explicit bearer and non-loopback security checks.\n\nInitialized stdio/private-socket controls support `thread/loaded/list` with\n`{cursor?: string | null, limit?: number | null}`. The result is\n`{data: string[], nextCursor: string | null}` from the held manager's loaded\nengine snapshot, including idle and archived engines still in memory. It dLine truncated
       },
       {
         "path": "docs/RUNTIME_ROUTES.md",

web/data/features.generated.json

@@ -2,7 +2,7 @@
   "format": "codewhale.features.v1",
   "source": {
     "path": "docs/features.toml",
-    "sha256": "a63042dc25a658f626735cc40e7a7cf4c67eaa51bfbb037dc2820d083d10fe3a"
+    "sha256": "4c4dfe17cea716328253c687d9d0665e1b24fc3b59d00f2a81cf1157dfea7072"
   },
   "features": [
     {
@@ -1070,7 +1070,7 @@
     {
       "id": "live-thinking",
       "name": "Live thinking",
-      "summary": "Model thinking shows live in the transcript, then collapses under a Thought header; /verbose shows it in full.",
+      "summary": "/verbose toggles live model thinking in the transcript.",
       "status": "stable",
       "surfaces": {
         "tui": "stable"

web/lib/release-credits.ts

@@ -35,6 +35,7 @@ export const RELEASE_CONTRIBUTORS: string[] = [
   "@luochen211",
   "@gannp",
   "@Felixkw12",
+  "@MannXo",
 ];
 
 /**