docs(release): 准备 v1.25.2 中英更新日志 (#8895 )

* docs(release): prepare v1.25.2 notes Summary: Generate a bilingual, product-focused draft from merged pull request metadata. Reuse the selected release-bound PR when one is available. Verification: Validate the catalog, citations, bilingual fields, and rendered GitHub release notes before committing. * docs(release): prepare v1.25.2 notes Summary: Generate a bilingual, product-focused draft from merged pull request metadata. Reuse the selected release-bound PR when one is available. Verification: Validate the catalog, citations, bilingual fields, and rendered GitHub release notes before committing. * docs(release): review v1.25.2 notes Problem: Regenerating the notes after the release-blocking test fix made the open notes PR self-referential, added github-actions[bot] as a contributor, and dropped reviewed user-facing changes and guide links. Fix: Restore the comprehensive reviewed bilingual record. The merged test-only PR does not change the product outcomes described by the notes. Verification: Validate the release catalog, run the release-note tests, render v1.25.2, and scan the rendered public text for private data. --------- Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com> Co-authored-by: SivanCola <[email protected]> Co-authored-by: SivanCola <[email protected]>
test(frontend): cover expanded reasoning display mode (#8896 )
2026-11-10 01:51:40 -10:00 · 2026-08-14 23:26:58 +08:00 · 2026-08-14 23:21:35 +08:00 · 2026-08-14 14:45:30 +00:00 · 2026-08-14 21:13:25 +08:00 · 2026-08-14 19:09:57 +08:00
7674 changed files with 746486 additions and 627158 deletions
@@ -2,7 +2,6 @@
 .gitignore text eol=lf
 .npmrc text eol=lf
 *.astro text eol=lf
-*.bat text eol=crlf
 *.css text eol=lf
 *.go text eol=lf
 *.html text eol=lf
@@ -27,4 +26,3 @@
 *.yml text eol=lf
 .githooks/** text eol=lf
 benchmarks/**/*.sh text eol=lf
-internal/runtime/agent/testdata/perseveration/** -text
@@ -1,43 +1,14 @@
-# Ownership for the studio line. Standard: docs/DOCS_STANDARD.md.
-#
-# Each rule lists exactly two owners: primary first, backup second.
-# GitHub requests review from both. The last matching rule wins.
-# repolint (codeowners, doc-owner) fails a rule for a missing path, a rule
-# without a backup, and a document header that names different owners.
+# Every pull request requests review from the maintainers below.
+# Either maintainer can review and merge any change.
+* @SivanCola @esengine

-# Default: kernel, Studio shell, tooling, standing instructions.
-*                                         @esengine @SivanCola
-
-# User-facing guides and references.
-/docs/                                    @SivanCola @esengine
-/sdk/                                     @SivanCola @esengine
-/README.md                                @SivanCola @esengine
-/README.zh-CN.md                          @SivanCola @esengine
-/SECURITY.md                              @SivanCola @esengine
-
-# Contracts and processes the Studio line runs on.
-/docs/DOCS_STANDARD.md                    @esengine @SivanCola
-/docs/STUDIO_*.md                         @esengine @SivanCola
-/docs/SESSION_MEMORY_RETRIEVAL.md         @esengine @SivanCola
-/docs/*_CONTRACT.md                       @esengine @SivanCola
-/docs/TOOL_APPROVAL_MODES.md              @esengine @SivanCola
-/docs/CAPABILITY_DIAGNOSTICS.md           @esengine @SivanCola
-/docs/CODE_SIGNING_POLICY.md            @esengine @SivanCola
-/docs/CHECKPOINTS.md                      @esengine @SivanCola
-/docs/EXTENSION_*.md                      @esengine @SivanCola
-/docs/SUBAGENT_*.md                       @esengine @SivanCola
-/docs/ROADMAP.md                          @esengine @SivanCola
-/docs/MIGRATING.md                        @esengine @SivanCola
-/docs/design/                             @esengine @SivanCola
-/release-notes/                           @esengine @SivanCola
-
-# Release signing: provider policy, its contract and the workflow that uses it.
-/.signpath/                               @esengine @SivanCola
-/.github/workflows/release-studio.yml     @esengine @SivanCola
-/.github/workflows/studio-certum-signing-smoke.yml @esengine @SivanCola
-/.github/actions/setup-certum/            @esengine @SivanCola
-/cmd/signpath-contract/                   @esengine @SivanCola
-/scripts/sign-certum.ps1                  @esengine @SivanCola
-/scripts/verify-windows-authenticode.ps1  @esengine @SivanCola
-/scripts/verify-windows-package.ps1       @esengine @SivanCola
-/scripts/windows-signing-lib.ps1          @esengine @SivanCola
+# Keep the provider-side signing policy and its GitHub control plane under the
+# same explicit review boundary as the machine-readable contract.
+/.signpath/ @SivanCola @esengine
+/.github/workflows/release-stable.yml @SivanCola @esengine
+/.github/workflows/release-stable-trigger.yml @SivanCola @esengine
+/.github/workflows/release-desktop.yml @SivanCola @esengine
+/scripts/complete-signpath-request.ps1 @SivanCola @esengine
+/scripts/package-windows-desktop.sh @SivanCola @esengine
+/scripts/verify-windows-authenticode.ps1 @SivanCola @esengine
+/cmd/signpath-contract/ @SivanCola @esengine
@@ -7,25 +7,23 @@ body:
    attributes:
      value: |
        Thanks for filing a bug. Pick which version line you're on below — it gets
-        labeled automatically, no maintainer action needed. Not sure which line
-        you run? See the [version roadmap](https://github.com/esengine/DeepSeek-Reasonix/discussions/10748).
+        labeled (`v1`/`v2`) automatically, no maintainer action needed.
  - type: dropdown
    id: version-line
    attributes:
      label: Version line
      description: Which line are you running?
      options:
-        - "2.x — Reasonix Studio, studio branch (active development)"
-        - "1.x — main-v2 branch (maintenance / stable)"
-        - "0.x — legacy TypeScript (maintenance only)"
+        - "v2 — Go rewrite (1.x), main-v2 (active development)"
+        - "v1 — Legacy TypeScript (0.x), maintenance only"
    validations:
      required: true
  - type: input
    id: version
    attributes:
      label: Exact version
-      description: Output of `reasonix --version`, or the version shown in Studio's settings.
-      placeholder: e.g. 2.20.0 or 1.39.0
+      description: Output of `reasonix --version`.
+      placeholder: e.g. 1.0.0
    validations:
      required: true
  - type: textarea
@@ -1,8 +1,5 @@
 blank_issues_enabled: false
 contact_links:
-  - name: Version roadmap (1.x / 2.x)
-    url: https://github.com/esengine/DeepSeek-Reasonix/discussions/10748
-    about: Which line to use, which branch takes which changes, and where Reasonix is heading.
  - name: Questions & discussions
    url: https://github.com/esengine/DeepSeek-Reasonix/discussions
    about: For usage questions, ideas, and general discussion — please use Discussions instead of opening an issue.
@@ -9,9 +9,8 @@ body:
      label: Version line
      description: Which line is this for?
      options:
-        - "2.x — Reasonix Studio, studio branch (active development)"
-        - "1.x — main-v2 branch (maintenance / stable)"
-        - "0.x — legacy TypeScript (maintenance only)"
+        - "v2 — Go rewrite (1.x), main-v2 (active development)"
+        - "v1 — Legacy TypeScript (0.x), maintenance only"
    validations:
      required: true
  - type: textarea
@@ -1,35 +0,0 @@
-name: Connect to Certum cloud signing
-description: Verify the pinned SimplySign installer and connect to the configured certificate.
-inputs:
-  username:
-    required: true
-    description: SimplySign account
-  otp-uri:
-    required: true
-    description: TOTP provisioning URI
-  thumbprint:
-    required: true
-    description: Expected SHA-1 certificate thumbprint
-runs:
-  using: composite
-  steps:
-    - name: Verify SimplySign installer
-      shell: pwsh
-      run: |
-        $ErrorActionPreference = 'Stop'
-        Invoke-WebRequest 'https://files.certum.eu/software/SimplySignDesktop/Windows/9.4.3.90/SimplySignDesktop-9.4.3.90-64-bit-en.msi' -OutFile SimplySignDesktop.msi
-        $hash = (Get-FileHash -Algorithm SHA256 SimplySignDesktop.msi).Hash
-        if ($hash -ne 'CE2B38EF0124A574BC08E5558726099CAAC79A7C58B8D3D712F629371FE1BCB7') {
-          throw "SimplySign installer SHA-256 $hash does not match the pinned 9.4.3.90 build"
-        }
-        $signature = Get-AuthenticodeSignature SimplySignDesktop.msi
-        if ($signature.Status -ne 'Valid' -or $signature.SignerCertificate.Subject -notmatch 'Asseco Data Systems') {
-          throw 'SimplySign installer signature or publisher verification failed'
-        }
-    - name: Authenticate cloud certificate
-      uses: dismine/windows-app-signing-setup-action@89ae3b032d4bc7a5b98d1a42a34e61ecb6faad64
-      with:
-        certum-username: ${{ inputs.username }}
-        certum-otp-uri: ${{ inputs.otp-uri }}
-        certum-key-id: ${{ inputs.thumbprint }}
-        capture-diagnostics: 'false'
@@ -1,6 +1,5 @@
 version: 2
 updates:
-  # 2.x (studio): the development line.
  - package-ecosystem: "gomod"
    directory: "/"
    schedule:
@@ -22,7 +21,7 @@ updates:
        update-types: ["minor", "patch"]

  - package-ecosystem: "npm"
-    directory: "/desktop/frontend-next"
+    directory: "/desktop/frontend"
    schedule:
      interval: "weekly"
    open-pull-requests-limit: 5
@@ -32,7 +31,7 @@ updates:
        update-types: ["minor", "patch"]

  - package-ecosystem: "npm"
-    directory: "/desktop/electron"
+    directory: "/site"
    schedule:
      interval: "weekly"
    open-pull-requests-limit: 5
@@ -48,61 +47,3 @@ updates:
    groups:
      actions:
        patterns: ["*"]
-
-  # 1.x (main-v2): maintenance. Dependabot reads this file from the default
-  # branch only, so the 1.x line is named here; it takes patch updates alone.
-  - package-ecosystem: "gomod"
-    directory: "/"
-    target-branch: "main-v2"
-    schedule:
-      interval: "weekly"
-    open-pull-requests-limit: 3
-    ignore:
-      - dependency-name: "*"
-        update-types: ["version-update:semver-major", "version-update:semver-minor"]
-    groups:
-      go-1x:
-        patterns: ["*"]
-        update-types: ["patch"]
-
-  - package-ecosystem: "gomod"
-    directory: "/desktop"
-    target-branch: "main-v2"
-    schedule:
-      interval: "weekly"
-    open-pull-requests-limit: 3
-    ignore:
-      - dependency-name: "*"
-        update-types: ["version-update:semver-major", "version-update:semver-minor"]
-    groups:
-      go-1x:
-        patterns: ["*"]
-        update-types: ["patch"]
-
-  - package-ecosystem: "npm"
-    directory: "/desktop/frontend"
-    target-branch: "main-v2"
-    schedule:
-      interval: "weekly"
-    open-pull-requests-limit: 3
-    ignore:
-      - dependency-name: "*"
-        update-types: ["version-update:semver-major", "version-update:semver-minor"]
-    groups:
-      npm-1x:
-        patterns: ["*"]
-        update-types: ["patch"]
-
-  - package-ecosystem: "npm"
-    directory: "/site"
-    target-branch: "main-v2"
-    schedule:
-      interval: "weekly"
-    open-pull-requests-limit: 3
-    ignore:
-      - dependency-name: "*"
-        update-types: ["version-update:semver-major", "version-update:semver-minor"]
-    groups:
-      npm-1x:
-        patterns: ["*"]
-        update-types: ["patch"]
@@ -6,36 +6,36 @@
 agent:
  - changed-files:
      - any-glob-to-any-file:
-          - 'internal/runtime/agent/**'
-          - 'internal/session/control/**'
+          - 'internal/agent/**'
+          - 'internal/control/**'

 mcp:
  - changed-files:
      - any-glob-to-any-file:
-          - 'internal/ext/plugin/**'
+          - 'internal/plugin/**'
          - 'internal/codegraph/**'

 config:
  - changed-files:
      - any-glob-to-any-file:
-          - 'internal/contract/config/**'
-          - 'internal/assembly/boot/**'
+          - 'internal/config/**'
+          - 'internal/boot/**'

 provider:
  - changed-files:
      - any-glob-to-any-file:
-          - 'internal/contract/provider/**'
+          - 'internal/provider/**'

 skills:
  - changed-files:
      - any-glob-to-any-file:
-          - 'internal/ext/skill/**'
-          - 'internal/contract/tool/**'
+          - 'internal/skill/**'
+          - 'internal/tool/**'

 tui:
  - changed-files:
      - any-glob-to-any-file:
-          - 'internal/frontend/cli/**'
+          - 'internal/cli/**'

 desktop:
  - changed-files:
@@ -1,9 +1,3 @@
-<!--
-Target branch: 2.x features and fixes go to `studio`. 1.x takes bug fixes,
-provider/API compatibility, release/updater, security and platform stability
-only, on `main-v2`.
-->
-
 ## Summary

 -
@@ -26,3 +20,25 @@ the source-to-target contract here: easing syntax, time units, callbacks,
 cancellation, reduced-motion behavior, and failure fallback. Verification must
 assert that the target API was actually called; a mock that silently skips it
 does not count.
+
+## Documentation impact
+
+Documentation-impact: TODO
+
+For changes to user-visible CLI, Desktop, configuration, provider, permission,
+or tool behavior, use one of:
+
+- `Documentation-impact: updated - <what changed>` and update `docs/*.md`.
+- `Documentation-impact: none - <why the embedded documentation remains correct>`.
+
+## Cache impact
+
+Cache-impact: TODO
+Cache-guard: TODO
+System-prompt-review: N/A
+
+For cache-sensitive changes, fill these lines before requesting review:
+
+- `Cache-impact`: `none`, `low`, `medium`, or `high`, plus the reason.
+- `Cache-guard`: the focused guard test/command added or run, or why an existing guard covers the change.
+- `System-prompt-review`: required reviewer/approval note when provider-visible system prompt, memory prefix, output style, or skill index behavior changes.
@@ -0,0 +1,49 @@
+name: Cache impact guard
+
+on:
+  pull_request_target:
+    branches: [main-v2]
+    types: [opened, synchronize, reopened, edited, ready_for_review]
+    # The guard only cares about the prompt/tool surfaces listed in
+    # scripts/check-cache-impact.sh; PRs confined to clearly unrelated paths
+    # can never trip it, so skip the run entirely. Keep the ignore list
+    # conservative: when in doubt, run.
+    paths-ignore:
+      - 'docs/**'
+      - 'site/**'
+      - 'release-notes/**'
+      - 'workers/**'
+      - 'benchmarks/**'
+      - 'npm/**'
+      - '*.md'
+
+permissions:
+  contents: read
+  pull-requests: read
+
+concurrency:
+  group: cache-impact-${{ github.event.pull_request.number }}
+  cancel-in-progress: true
+
+jobs:
+  cache-impact:
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v7
+        with:
+          ref: ${{ github.event.pull_request.base.sha }}
+
+      - name: Collect PR changed files
+        env:
+          GH_TOKEN: ${{ github.token }}
+          PR_NUMBER: ${{ github.event.pull_request.number }}
+          REPOSITORY: ${{ github.repository }}
+        run: |
+          gh api --paginate "repos/${REPOSITORY}/pulls/${PR_NUMBER}/files" \
+            --jq '.[].filename' > "$RUNNER_TEMP/cache-impact-files.txt"
+
+      - name: Check cache-sensitive PR metadata
+        env:
+          PR_BODY: ${{ github.event.pull_request.body }}
+          CACHE_IMPACT_CHANGED_FILES_FILE: ${{ runner.temp }}/cache-impact-files.txt
+        run: bash scripts/check-cache-impact.sh
@@ -1,14 +1,10 @@
 name: CI

-# studio is a parallel release line, not a feature branch: it carries the
-# kernel changes Studio is built on, so it needs the same kernel gates main-v2
-# has. Studio's own surface (desktop/next, desktop/frontend-next) exists only
-# there and is covered by studio.yml.
 on:
  push:
-    branches: [main-v2, studio]
+    branches: [main-v2]
  pull_request:
-    branches: [main-v2, studio]
+    branches: [main-v2]

 permissions:
  contents: read
@@ -18,7 +14,7 @@ concurrency:
  cancel-in-progress: true

 jobs:
-  # Cheap path gate for pull requests. PRs confined to docs/release-notes
+  # Cheap path gate for pull requests. PRs confined to docs/site/release-notes
  # (or top-level Markdown) skip the heavy Go jobs, and PRs that cannot affect
  # the desktop module skip the desktop jobs. Job-level `if` reports skipped,
  # which satisfies the required status checks (lint, race, test) — a
@@ -32,15 +28,15 @@ jobs:
    outputs:
      code: ${{ steps.filter.outputs.code }}
      desktop: ${{ steps.filter.outputs.desktop }}
+      site: ${{ steps.filter.outputs.site }}
      sdk: ${{ steps.filter.outputs.sdk }}
-      scripts: ${{ steps.filter.outputs.scripts }}
    steps:
-      - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7
+      - uses: actions/checkout@v7
        with:
          fetch-depth: 0
      - id: filter
        run: |
-          code=true; desktop=true; sdk=true; scripts=true
+          code=true; desktop=true; site=true; sdk=true
          base=""
          if [ "${{ github.event_name }}" = "pull_request" ]; then
            base="${{ github.event.pull_request.base.sha }}"
@@ -50,37 +46,21 @@ jobs:
          if [ -n "$base" ] && git cat-file -e "$base^{commit}" 2>/dev/null; then
            files=$(git diff --name-only "$base" HEAD)
            if [ -n "$files" ]; then
-              code=false; desktop=false; sdk=false; scripts=false
+              code=false; desktop=false; site=false; sdk=false
              # Root-module CI: desktop/ is a separate module, so only the
              # clearly unrelated paths below can skip it.
-              if echo "$files" | grep -qvE '^(docs/|release-notes/|desktop/|workers/|[^/]+\.md$)'; then code=true; fi
+              if echo "$files" | grep -qvE '^(docs/|site/|release-notes/|desktop/|workers/|[^/]+\.md$)'; then code=true; fi
              # desktop/ imports the root kernel via `replace reasonix => ../`,
              # so only this clearly-unrelated set is safe to skip.
-              if echo "$files" | grep -qvE '^(docs/|release-notes/|workers/|benchmarks/|npm/|[^/]+\.md$)'; then desktop=true; fi
+              if echo "$files" | grep -qvE '^(docs/|site/|release-notes/|workers/|benchmarks/|npm/|[^/]+\.md$)'; then desktop=true; fi
+              if echo "$files" | grep -q '^site/'; then site=true; fi
              # sdk/go is a nested module invisible to root `go test ./...`;
-              # its DTOs are generated from internal/ext/extension/protocol, so
+              # its DTOs are generated from internal/extension/protocol, so
              # both paths must trigger the sdk job.
-              if echo "$files" | grep -qE '^(sdk/|internal/ext/extension/)'; then sdk=true; fi
-              # The release-note tests render every committed Studio note; npm/ holds the publish tests.
-              if echo "$files" | grep -qE '^(scripts/|npm/|release-notes/studio/)'; then scripts=true; fi
+              if echo "$files" | grep -qE '^(sdk/|internal/extension/)'; then sdk=true; fi
            fi
          fi
-          { echo "code=$code"; echo "desktop=$desktop"; echo "sdk=$sdk"; echo "scripts=$scripts"; } >> "$GITHUB_OUTPUT"
-
-  release-notes-scripts:
-    needs: changes
-    if: always() && (github.event_name != 'pull_request' || needs.changes.outputs.scripts != 'false')
-    runs-on: ubuntu-latest
-    permissions:
-      contents: read
-    steps:
-      - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7
-        with:
-          persist-credentials: false
-      - uses: actions/setup-node@820762786026740c76f36085b0efc47a31fe5020 # v7
-        with:
-          node-version: "22"
-      - run: node --test scripts/release-credits.test.mjs scripts/markdown-refs.test.mjs scripts/studio-release-notes.test.mjs npm/publish.test.mjs
+          { echo "code=$code"; echo "desktop=$desktop"; echo "site=$site"; echo "sdk=$sdk"; } >> "$GITHUB_OUTPUT"

  # The ruleset requires the per-OS check names (test (ubuntu-latest) etc.).
  # A matrix job skipped at job level reports no per-leg checks at all, so
@@ -101,10 +81,10 @@ jobs:
      RUN_STEPS: ${{ github.event_name != 'pull_request' || needs.changes.outputs.code != 'false' }}
    steps:
      - if: env.RUN_STEPS == 'true'
-        uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7
+        uses: actions/checkout@v7

      - if: env.RUN_STEPS == 'true'
-        uses: actions/setup-go@b7ad1dad31e06c5925ef5d2fc7ad053ef454303e # v7
+        uses: actions/setup-go@v7
        with:
          go-version-file: go.mod
          cache: true
@@ -144,21 +124,6 @@ jobs:
        if: env.RUN_STEPS == 'true'
        run: go build ./...

-      # The trajectory analyzer is the last consumer of the todo_progress frame,
-      # so a rename anywhere upstream has to fail somewhere. Standard library
-      # only; the runner's python3 is enough.
-      - name: test (trajectory tools)
-        if: env.RUN_STEPS == 'true' && runner.os != 'Windows'
-        run: python3 -m unittest discover -s tools/trajectory -t tools/trajectory
-
-      # A hardware exception on a Windows host whose CPU saves a large XSTATE
-      # (Intel AMX) can corrupt the Go heap (golang/go#81238). The runner pool
-      # mixes hosts, so a runtime crash is only attributable with the CPU on record.
-      - name: runner CPU
-        if: env.RUN_STEPS == 'true' && runner.os == 'Windows'
-        shell: pwsh
-        run: Get-CimInstance Win32_Processor | Select-Object -ExpandProperty Name
-
      - name: test
        if: env.RUN_STEPS == 'true' && runner.os != 'Windows'
        env:
@@ -176,52 +141,45 @@ jobs:
      # remains independently bounded by Go's per-package timeout below.
      #
      # That timeout is 8m, not 3m: across four consecutive runs of the same
-      # code, internal/assembly/boot measured 133s, 146s, 162s and then hit 180s, and
-      # internal/runtime/agent 82s, 92s, 144s and 180s — a 2.2x spread from runner
+      # code, internal/boot measured 133s, 146s, 162s and then hit 180s, and
+      # internal/agent 82s, 92s, 144s and 180s — a 2.2x spread from runner
      # variance alone, with four heavy packages sharing four vCPUs. At 3m the
      # step failed on slow runners without a code change; a hang still trips
-      # the 15 minute job budget above.
-      #
-      # Every root-module `go test` that runs on Windows passes -count=1. With
-      # the result cache on, go test records each file and environment access
-      # a test binary makes, and before and after the run resolves every
-      # recorded path outside the module through filepath.EvalSymlinks.
-      # internal/assembly/boot records about 1.9M accesses; on a Windows runner
-      # that bookkeeping costs minutes per pass while nothing is printed.
-      # internal/base/testenv holds every such step to it.
+      # the 15 minute job budget above. internal/control runs in its own
+      # Windows job below so its durable-session I/O does not compete here.
      - name: test (Windows smoke)
        if: env.RUN_STEPS == 'true' && runner.os == 'Windows' && github.event_name == 'pull_request'
        timeout-minutes: 15
        env:
          REASONIX_RELEASE_CACHE_GUARD: "1"
          WINDOWS_SANDBOX_WAIT_MS: "20000"
-        run: go test -count=1 -p 4 -timeout=8m ./internal/runtime/agent/... ./internal/platform/appidentity/... ./internal/assembly/boot/... ./internal/state/checkpoint/... ./internal/frontend/cli/... ./internal/session/control/... ./internal/base/filelock/... ./internal/base/fileutil/... ./internal/ext/hook/... ./internal/state/instruction/... ./internal/ext/mcplaunch/... ./internal/platform/notify/... ./internal/base/proc/... ./internal/platform/remote/... ./internal/platform/repair/... ./internal/safety/sandbox/... ./internal/base/sysproxy/... ./internal/state/workspacelease/... ./cmd/...
+        run: go test -p 4 -timeout=8m ./internal/agent/... ./internal/appidentity/... ./internal/boot/... ./internal/checkpoint/... ./internal/cli/... ./internal/desktoplauncher/... ./internal/filelock/... ./internal/fileutil/... ./internal/hook/... ./internal/instruction/... ./internal/mcplaunch/... ./internal/notify/... ./internal/proc/... ./internal/remote/... ./internal/repair/... ./internal/sandbox/... ./internal/sysproxy/... ./internal/workspacelease/... ./cmd/...

-      # The general Windows PR smoke list intentionally omits internal/contract/tool.
+      # The general Windows PR smoke list intentionally omits internal/tool.
      # Keep the session-temp portability contract covered without widening the
      # leg to every tool test: two real PowerShell launches must share the
      # injected TMPDIR/TMP/TEMP directory.
      - name: test (Windows session temp)
        if: env.RUN_STEPS == 'true' && runner.os == 'Windows' && github.event_name == 'pull_request'
        timeout-minutes: 3
-        run: go test -count=1 -timeout=2m -run '^TestBashSharesSessionTempAcrossCalls$' ./internal/tools/builtin
+        run: go test -timeout=2m -run '^TestBashSharesSessionTempAcrossCalls$' ./internal/tool/builtin

      # Shell execution contract: PowerShell identity, Chinese workspace paths,
      # UTF-8 output, and ExitCode retention must gate PRs on native Windows.
-      # Keep this focused (not full ./internal/contract/tool) so the smoke budget holds.
+      # Keep this focused (not full ./internal/tool) so the smoke budget holds.
      - name: test (Windows shell execution contract)
        if: env.RUN_STEPS == 'true' && runner.os == 'Windows' && github.event_name == 'pull_request'
        timeout-minutes: 5
-        run: go test -count=1 -timeout=3m -run '^TestBashPowerShellExecuteDetailedContract$|^TestBashPowerShell51PreflightRejectsAndAndDetailed$|^TestBashPowerShellOutputIsUTF8$|^TestBashPowerShellSurfacesNonZeroExit$|^TestBashPowerShellRejectsChaining$' ./internal/tools/builtin
+        run: go test -timeout=3m -run '^TestBashPowerShellExecuteDetailedContract$|^TestBashPowerShell51PreflightRejectsAndAndDetailed$|^TestBashPowerShellOutputIsUTF8$|^TestBashPowerShellSurfacesNonZeroExit$|^TestBashPowerShellRejectsChaining$' ./internal/tool/builtin

      # Full push suite still uses the same 8m package budget as Windows smoke:
      # agent/boot/control already hit 180s package timeouts under runner load,
-      # so 3m fails on slow machines without a code regression. It runs with
-      # -count=1 for the reason given at the smoke step. A hang still dies here
-      # rather than on a package.
+      # so 3m fails on slow machines without a code regression. Keep the step
+      # above normal full-suite wall time (~8-12m) so a hang still dies on the
+      # job budget rather than a false package timeout.
      - name: test (full)
        if: env.RUN_STEPS == 'true' && runner.os == 'Windows' && github.event_name != 'pull_request'
-        timeout-minutes: 30
+        timeout-minutes: 20
        env:
          # Run the prompt-cache prefix-stability guard (TestCacheHit*) in CI: a
          # regression there silently tanks the cache hit rate the project is
@@ -230,16 +188,46 @@ jobs:
          # Bound sandbox helper children in Windows tests so a failed OS-level
          # launch cannot pin the Actions step after Go's package timeout fires.
          WINDOWS_SANDBOX_WAIT_MS: "20000"
-        run: go test -count=1 -p 4 -timeout=8m ./...
+        run: go test -p 4 -timeout=8m ./...
+
+      - name: test (Scoop desktop launch)
+        if: env.RUN_STEPS == 'true' && runner.os == 'Windows'
+        timeout-minutes: 10
+        shell: powershell
+        run: .\scripts\test-scoop-desktop-launch.ps1
+
+  windows-control:
+    needs: changes
+    if: always()
+    runs-on: windows-latest
+    env:
+      RUN_STEPS: ${{ github.event_name != 'pull_request' || needs.changes.outputs.code != 'false' }}
+    steps:
+      - if: env.RUN_STEPS == 'true'
+        uses: actions/checkout@v7
+
+      - if: env.RUN_STEPS == 'true'
+        uses: actions/setup-go@v7
+        with:
+          go-version-file: go.mod
+          cache: true
+
+      - name: test
+        if: env.RUN_STEPS == 'true'
+        timeout-minutes: 10
+        env:
+          REASONIX_RELEASE_CACHE_GUARD: "1"
+          WINDOWS_SANDBOX_WAIT_MS: "20000"
+        run: go test -p 1 -timeout=8m ./internal/control/...

  race:
    needs: changes
    if: always() && (github.event_name != 'pull_request' || needs.changes.outputs.code != 'false')
    runs-on: ubuntu-latest
    steps:
-      - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7
+      - uses: actions/checkout@v7

-      - uses: actions/setup-go@b7ad1dad31e06c5925ef5d2fc7ad053ef454303e # v7
+      - uses: actions/setup-go@v7
        with:
          go-version-file: go.mod
          cache: true
@@ -266,7 +254,7 @@ jobs:
        if: github.event_name == 'pull_request'
        env:
          REASONIX_RELEASE_CACHE_GUARD: "1"
-        run: go test -race ./internal/runtime/agent/... ./internal/ext/plugin/... ./internal/tools/jobs/... ./internal/base/proc/... ./internal/safety/sandbox/... ./internal/base/filelock/... ./internal/contract/eventwire/... ./internal/platform/remote/... ./internal/ext/extension/... ./internal/assembly/boot/... ./internal/session/control/... ./internal/contract/tool/...
+        run: go test -race ./internal/agent/... ./internal/plugin/... ./internal/jobs/... ./internal/proc/... ./internal/sandbox/... ./internal/filelock/... ./internal/eventwire/... ./internal/remote/... ./internal/extension/... ./internal/boot/... ./internal/control/... ./internal/tool/...

      - name: test -race (full)
        if: github.event_name != 'pull_request'
@@ -275,7 +263,7 @@ jobs:
        run: go test -race ./...

  # sdk/go is a nested stdlib-only module invisible to root `go test ./...`.
-  # Its DTOs are generated from internal/ext/extension/protocol, and the
+  # Its DTOs are generated from internal/extension/protocol, and the
  # host-side conformance tests spawn the SDK example, so the job runs the
  # same three-OS matrix as the root tests.
  sdk:
@@ -293,10 +281,10 @@ jobs:
        working-directory: sdk/go
    steps:
      - if: env.RUN_STEPS == 'true'
-        uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7
+        uses: actions/checkout@v7

      - if: env.RUN_STEPS == 'true'
-        uses: actions/setup-go@b7ad1dad31e06c5925ef5d2fc7ad053ef454303e # v7.0.0
+        uses: actions/setup-go@v6
        with:
          go-version-file: sdk/go/go.mod

@@ -341,14 +329,30 @@ jobs:
      run:
        working-directory: desktop
    steps:
-      - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7
+      - uses: actions/checkout@v7

-      - uses: actions/setup-go@b7ad1dad31e06c5925ef5d2fc7ad053ef454303e # v7
+      - uses: actions/setup-go@v7
        with:
          go-version-file: desktop/go.mod
          cache: true
          cache-dependency-path: desktop/go.sum

+      - uses: pnpm/[email protected]
+        with:
+          version: 10
+          run_install: false
+
+      - uses: actions/setup-node@v7
+        with:
+          node-version: "24"
+          cache: pnpm
+          cache-dependency-path: desktop/frontend/pnpm-lock.yaml
+
+      - name: Install Wails CLI
+        run: |
+          echo "$(go env GOPATH)/bin" >> "$GITHUB_PATH"
+          go install "github.com/wailsapp/wails/v2/cmd/wails@$(cat "$GITHUB_WORKSPACE/.wails-version")"
+
      - name: gofmt
        run: |
          unformatted=$(gofmt -l .)
@@ -367,11 +371,37 @@ jobs:
            exit 1
          fi

+      # WebKitGTK 4.0 toolchain (pinned to ubuntu-22.04; no webkit2_41 tag).
+      - name: Install Linux build deps
+        run: |
+          sudo apt-get update
+          sudo apt-get install -y gcc libgtk-3-dev libwebkit2gtk-4.0-dev xvfb
+
+      - name: Build frontend
+        run: |
+          wails generate module
+          pnpm --dir frontend install --frozen-lockfile
+          PLAYWRIGHT_BROWSERS_PATH=.pw-browsers pnpm --dir frontend exec playwright install --with-deps chromium
+          pnpm --dir frontend build
+
+      - name: Test desktop frontend
+        run: |
+          pnpm --dir frontend test:terminal
+          pnpm --dir frontend test:task-monitor
+          pnpm --dir frontend test:workspace
+          pnpm --dir frontend test:motion
+          pnpm --dir frontend test:motion-browser
+          pnpm --dir frontend test:transcript
+          pnpm --dir frontend test:transcript-browser
+
+      - name: Test usage statistics frontend
+        run: pnpm --dir frontend test:usage-stats
+
      - name: vet
        run: go vet ./...

      - name: golangci-lint
-        uses: golangci/golangci-lint-action@ba0d7d2ec06a0ea1cb5fa41b2e4a3ab91d21278a # v9
+        uses: golangci/golangci-lint-action@v9
        with:
          version: v2.12.2
          working-directory: desktop
@@ -383,13 +413,46 @@ jobs:
      - name: test
        run: go test ./...

-      # The update helper replaces a running application under pkexec, so its
-      # own concurrency is CI-blocked rather than author-verified locally.
+      - name: WebKitGTK 4.0 native recovery smoke
+        run: xvfb-run -a env WEBKIT_DISABLE_COMPOSITING_MODE=1 go test -tags reasonix_webkit_smoke . -run '^TestWebKitNativeRecoverySmoke$'
+
+      # The extension work added new shared-state paths to the desktop
+      # runtime (tab/rebuild fences, extension UI). Race-sweep the module so
+      # regressions are CI-blocked, not just author-verified locally.
      - name: test -race
        run: go test -race ./...

+  # Ubuntu 24.04 ships WebKitGTK 4.1 only. Keep this capability build separate
+  # from the 22.04/4.0 leg so neither runtime becomes a distro whitelist.
+  desktop-linux-webkit41:
+    needs: changes
+    if: always() && (github.event_name != 'pull_request' || needs.changes.outputs.desktop != 'false')
+    runs-on: ubuntu-24.04
+    defaults:
+      run:
+        working-directory: desktop
+    steps:
+      - uses: actions/checkout@v7
+
+      - uses: actions/setup-go@v7
+        with:
+          go-version-file: desktop/go.mod
+          cache: true
+          cache-dependency-path: desktop/go.sum
+
+      - name: Install WebKitGTK 4.1 build dependencies
+        run: |
+          sudo apt-get update
+          sudo apt-get install -y gcc libgtk-3-dev libwebkit2gtk-4.1-dev xvfb
+
+      - name: Build and test WebKitGTK 4.1 observer
+        run: go test -tags webkit2_41 ./...
+
+      - name: WebKitGTK 4.1 native recovery smoke
+        run: xvfb-run -a env WEBKIT_DISABLE_COMPOSITING_MODE=1 go test -tags 'webkit2_41 reasonix_webkit_smoke' . -run '^TestWebKitNativeRecoverySmoke$'
+
  # desktop/ is a separate module, so the root macOS matrix above does not
-  # compile it against a real macOS toolchain.
+  # compile or exercise the native PTY implementation.
  desktop-macos:
    needs: changes
    if: always() && (github.event_name != 'pull_request' || needs.changes.outputs.desktop != 'false')
@@ -398,26 +461,31 @@ jobs:
      run:
        working-directory: desktop
    steps:
-      - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7
+      - uses: actions/checkout@v7

-      - uses: actions/setup-go@b7ad1dad31e06c5925ef5d2fc7ad053ef454303e # v7
+      - uses: actions/setup-go@v7
        with:
          go-version-file: desktop/go.mod
          cache: true
          cache-dependency-path: desktop/go.sum

-      - name: test
-        run: go test -race ./...
+      - name: Test integrated terminal and PTY lifecycle
+        run: go test -race -run 'Test(ResolveTerminal|Terminal|EmptyTerminal|UnixTerminalProcess)' .

+      # desktop/*_darwin.go is the one build-tag set the lint job cannot reach:
+      # the main-thread watchdog is cgo, so it only type-checks with a real
+      # macOS toolchain.
      - name: golangci-lint
-        uses: golangci/golangci-lint-action@ba0d7d2ec06a0ea1cb5fa41b2e4a3ab91d21278a # v9
+        uses: golangci/golangci-lint-action@v9
        with:
          version: v2.12.2
          working-directory: desktop
          args: --timeout=5m

-  # The update helper's versioned-install and uninstall paths are Windows-only
-  # (//go:build windows), so the ubuntu leg above never compiles them.
+  # Desktop project-root matching is case-insensitive only on Windows
+  # (sameDesktopPath folds case when os.PathSeparator is '\'), so the
+  # regression tests for that contract are named *OnWindows and can never
+  # run on the ubuntu leg above.
  desktop-windows:
    needs: changes
    if: always() && (github.event_name != 'pull_request' || needs.changes.outputs.desktop != 'false')
@@ -427,18 +495,85 @@ jobs:
        shell: bash
        working-directory: desktop
    steps:
-      - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7
+      - uses: actions/checkout@v7

-      - uses: actions/setup-go@b7ad1dad31e06c5925ef5d2fc7ad053ef454303e # v7
+      - uses: actions/setup-go@v7
        with:
          go-version-file: desktop/go.mod
          cache: true
          cache-dependency-path: desktop/go.sum

-      - name: test (update helper and uninstaller)
+      - uses: pnpm/[email protected]
+        with:
+          version: 10
+          run_install: false
+
+      - uses: actions/setup-node@v7
+        with:
+          node-version: "24"
+          cache: pnpm
+          cache-dependency-path: desktop/frontend/pnpm-lock.yaml
+
+      - name: Install Wails CLI
+        run: |
+          echo "$(go env GOPATH)/bin" >> "$GITHUB_PATH"
+          go install "github.com/wailsapp/wails/v2/cmd/wails@$(cat "$GITHUB_WORKSPACE/.wails-version")"
+
+      # go:embed of frontend/dist needs a built frontend before the package
+      # compiles, same as the ubuntu desktop leg.
+      - name: Build frontend
+        run: |
+          wails generate module
+          pnpm --dir frontend install --frozen-lockfile
+          pnpm --dir frontend build
+
+      - name: Test native motion contracts
+        run: pnpm --dir frontend test:motion
+
+      - name: Test WebView2 native smoke state machine
+        shell: pwsh
+        run: ../scripts/test-webview2-native-smoke.ps1 -SelfTest
+
+      # Build the real Wails executable on every desktop PR and exercise the
+      # exact production startup path in Microsoft's installed WebView2 runtime.
+      # The frontend and bindings were already built above, so this skips only
+      # duplicate build work, not the native shell.
+      - name: Build Wails executable for native startup smoke
+        timeout-minutes: 10
+        run: wails build -clean -s -skipbindings -nopackage -platform windows/amd64 -webview2 embed
+
+      - name: Smoke-test Wails/WebView2 native startup
+        shell: pwsh
+        timeout-minutes: 2
+        run: |
+          ../scripts/test-webview2-native-smoke.ps1 `
+            -ExecutablePath "build/bin/reasonix-desktop.exe" `
+            -TimeoutSeconds 60 `
+            -HealthySeconds 5
+
+      - name: test (Windows desktop and update helper)
        timeout-minutes: 15
        run: go test ./...

+      # A cold nested-module cache can spend several minutes compiling the
+      # real WebView2 cookie integration test. Keep its budget independent of
+      # the already-complete desktop suite and retain a Go timeout stack dump.
+      - name: test (vendored WebView2 edge)
+        timeout-minutes: 10
+        run: (cd third_party/go-webview2 && go test -timeout=8m ./pkg/edge)
+
+      # Installer packaging is packaging validation, not a code gate: run it
+      # on pushes to main-v2 (the release pipeline builds installers again
+      # anyway), but let pull requests stop after the test step.
+      - name: Install NSIS
+        if: github.event_name != 'pull_request'
+        run: pwsh -NoProfile -File ../scripts/install-nsis.ps1
+
+      - name: Build Windows installer and portable archive
+        if: github.event_name != 'pull_request'
+        timeout-minutes: 20
+        run: ../scripts/desktop-build.sh windows/amd64 v0.0.0-ci canary
+
  # repolint scans desktop/ and sdk/ too, so this job also runs for diffs the
  # `code` filter would otherwise skip.
  lint:
@@ -446,20 +581,39 @@ jobs:
    if: always() && (github.event_name != 'pull_request' || needs.changes.outputs.code != 'false' || needs.changes.outputs.desktop != 'false' || needs.changes.outputs.sdk != 'false')
    runs-on: ubuntu-latest
    steps:
-      - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7
+      - uses: actions/checkout@v7

-      - uses: actions/setup-go@b7ad1dad31e06c5925ef5d2fc7ad053ef454303e # v7
+      - uses: actions/setup-go@v7
        with:
          go-version-file: go.mod
          cache: true

+      - if: github.event_name != 'pull_request' || needs.changes.outputs.desktop != 'false'
+        uses: pnpm/[email protected]
+        with:
+          version: 10
+          run_install: false
+
+      - if: github.event_name != 'pull_request' || needs.changes.outputs.desktop != 'false'
+        uses: actions/setup-node@v7
+        with:
+          node-version: "24"
+          cache: pnpm
+          cache-dependency-path: desktop/frontend/pnpm-lock.yaml
+
+      - name: required native motion contracts
+        if: github.event_name != 'pull_request' || needs.changes.outputs.desktop != 'false'
+        run: |
+          pnpm --dir desktop/frontend install --frozen-lockfile
+          pnpm --dir desktop/frontend test:motion
+
      - name: repo standards
        run: go run ./tools/repolint

-      # The paths the project declared sensitive, held to the coverage they
-      # already have. It runs only those packages' tests, not the full suite.
-      - name: sensitive-path coverage
-        run: go run ./tools/covergate
+      - name: wails version pin
+        run: |
+          bash scripts/check-wails-pin.sh
+          bash scripts/check-wails-pin.test.sh

      # `make lint-install` reads the same file, so a local run and this job
      # cannot drift onto different linter versions.
@@ -467,7 +621,7 @@ jobs:
        run: echo "version=$(cat .golangci-version)" >> "$GITHUB_OUTPUT"

      - name: golangci-lint
-        uses: golangci/golangci-lint-action@ba0d7d2ec06a0ea1cb5fa41b2e4a3ab91d21278a # v9
+        uses: golangci/golangci-lint-action@v9
        with:
          version: ${{ steps.golangci.outputs.version }}
          args: --timeout=5m
@@ -491,9 +645,38 @@ jobs:
        run: |
          go run github.com/rhysd/actionlint/cmd/[email protected] \
            -ignore 'label "windows-11-arm" is unknown' \
-            .github/workflows/release-studio.yml \
-            .github/workflows/studio-certum-signing-smoke.yml \
+            .github/workflows/release-stable.yml \
+            .github/workflows/prepare-release-notes.yml \
+            .github/workflows/release-stable-trigger.yml \
+            .github/workflows/release.yml \
+            .github/workflows/release-npm.yml \
+            .github/workflows/release-desktop.yml \
+            .github/workflows/release-verify-issues.yml \
+            .github/workflows/docs-impact.yml \
            .github/workflows/ci.yml
+          bash scripts/release-workflows.test.sh
+          node scripts/check-single-release-public-contract.mjs
+
+  # The site/ auth client has security-sensitive redirect-validation logic
+  # (safeNext) covered by node:test unit tests. Those tests use only Node
+  # builtins, so no `npm install` is needed — run them directly on every PR so
+  # a regression in redirect validation fails the build instead of shipping.
+  site:
+    needs: changes
+    if: always() && (github.event_name != 'pull_request' || needs.changes.outputs.site != 'false')
+    runs-on: ubuntu-latest
+    defaults:
+      run:
+        working-directory: site
+    steps:
+      - uses: actions/checkout@v7
+
+      - uses: actions/setup-node@v7
+        with:
+          node-version: "24"
+
+      - name: test
+        run: npm test

  govulncheck:
    needs: changes
@@ -501,9 +684,9 @@ jobs:
    runs-on: ubuntu-latest
    continue-on-error: true # informational — stdlib vulns need a Go patch release
    steps:
-      - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7
+      - uses: actions/checkout@v7

-      - uses: actions/setup-go@b7ad1dad31e06c5925ef5d2fc7ad053ef454303e # v7
+      - uses: actions/setup-go@v7
        with:
          go-version-file: go.mod
          cache: true
@@ -519,9 +702,9 @@ jobs:
    if: always() && (github.event_name != 'pull_request' || needs.changes.outputs.code != 'false')
    runs-on: ubuntu-latest
    steps:
-      - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7
+      - uses: actions/checkout@v7

-      - uses: actions/setup-go@b7ad1dad31e06c5925ef5d2fc7ad053ef454303e # v7
+      - uses: actions/setup-go@v7
        with:
          go-version-file: go.mod
          cache: true
@@ -530,7 +713,7 @@ jobs:
        run: go test -coverprofile=coverage.out -covermode=atomic ./...

      - name: upload coverage
-        uses: actions/upload-artifact@043fb46d1a93c77aae656e7c1c64a875d1fc6a0a # v7
+        uses: actions/upload-artifact@v7
        with:
          name: coverage-report
          path: coverage.out
@@ -1,35 +0,0 @@
-name: Close linked issues
-
-# GitHub closes an issue named by "Fixes #N" only when the pull request merges
-# into the default branch. This does the same for the 1.x line. It reads the
-# merged pull request's description and never checks out its code.
-on:
-  pull_request_target:
-    types: [closed]
-    branches: [main-v2]
-
-permissions:
-  issues: write
-
-jobs:
-  close:
-    if: github.event.pull_request.merged == true
-    runs-on: ubuntu-latest
-    steps:
-      - uses: actions/github-script@3a2844b7e9c422d3c10d287c895573f7108da1b3 # v9
-        with:
-          script: |
-            const pr = context.payload.pull_request;
-            const { owner, repo } = context.repo;
-            // GitHub's own closing-keyword grammar, applied to this repository's issues only.
-            const keyword = /\b(?:close[sd]?|fix(?:e[sd])?|resolve[sd]?)\s*:?\s+#(\d+)\b/gi;
-            const numbers = [...new Set([...(pr.body ?? '').matchAll(keyword)].map((m) => Number(m[1])))];
-            const sha = (pr.merge_commit_sha ?? '').slice(0, 9);
-            for (const issue_number of numbers) {
-              const { data } = await github.rest.issues.get({ owner, repo, issue_number });
-              if (data.pull_request || data.state !== 'open') continue;
-              await github.rest.issues.createComment({ owner, repo, issue_number,
-                body: `Fixed on the 1.x line by #${pr.number} (${sha}). It ships in the next 1.x release.` });
-              await github.rest.issues.update({ owner, repo, issue_number, state: 'closed', state_reason: 'completed' });
-              core.info(`closed #${issue_number}`);
-            }
@@ -4,25 +4,23 @@ on:
  # Docs/release-notes/root-Markdown changes carry no analyzable code; the
  # weekly cron keeps the full scan cadence regardless.
  push:
-    branches: ["main-v2", "studio"]
+    branches: ["main-v2"]
    paths-ignore:
      - 'docs/**'
      - 'release-notes/**'
      - '*.md'
  pull_request:
-    branches: ["main-v2", "studio"]
+    branches: ["main-v2"]
    paths-ignore:
      - 'docs/**'
      - 'release-notes/**'
      - '*.md'
-  # A schedule runs only from the default branch, so the weekly scan checks out
-  # both lines itself: 1.x is in maintenance, and security is part of it.
  schedule:
    - cron: "27 3 * * 1"

 jobs:
  analyze:
-    name: Analyze (${{ matrix.language }}${{ matrix.branch && format(', {0}', matrix.branch) || '' }})
+    name: Analyze (${{ matrix.language }})
    runs-on: ubuntu-latest
    permissions:
      security-events: write
@@ -32,32 +30,21 @@ jobs:
    strategy:
      fail-fast: false
      matrix:
-        language: [go, javascript-typescript, actions]
-        # Empty means "the ref this event is about"; a scheduled run names each line.
-        branch: ${{ fromJSON(github.event_name == 'schedule' && '["main-v2", "studio"]' || '[""]') }}
+        include:
+          - language: go
+            build-mode: autobuild
+          - language: javascript-typescript
+            build-mode: none
+          - language: actions
+            build-mode: none
    steps:
-      - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7
-        with:
-          ref: ${{ matrix.branch }}
-      - id: head
-        if: matrix.branch != ''
-        run: echo "sha=$(git rev-parse HEAD)" >> "$GITHUB_OUTPUT"
+      - uses: actions/checkout@v7
      - name: Initialize CodeQL
-        uses: github/codeql-action/init@2892aa5e19bbd11bc0cff5427e3b750a04d9e3c2 # v4.38.2
+        uses: github/codeql-action/init@v4.37.4
        with:
          languages: ${{ matrix.language }}
-          build-mode: ${{ matrix.language == 'go' && 'autobuild' || 'none' }}
+          build-mode: ${{ matrix.build-mode }}
      - name: Perform CodeQL Analysis
-        if: matrix.branch == ''
-        uses: github/codeql-action/analyze@2892aa5e19bbd11bc0cff5427e3b750a04d9e3c2 # v4.38.2
+        uses: github/codeql-action/[email protected]
        with:
          category: "/language:${{ matrix.language }}"
-      # Results from a checked-out line are filed under that line's ref, not the
-      # default branch the schedule happened to run from.
-      - name: Perform CodeQL Analysis (${{ matrix.branch }})
-        if: matrix.branch != ''
-        uses: github/codeql-action/analyze@2892aa5e19bbd11bc0cff5427e3b750a04d9e3c2 # v4.38.2
-        with:
-          category: "/language:${{ matrix.language }}"
-          ref: refs/heads/${{ matrix.branch }}
-          sha: ${{ steps.head.outputs.sha }}
@@ -19,12 +19,12 @@ jobs:
  deploy:
    runs-on: ubuntu-latest
    steps:
-      - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7
-      - uses: pnpm/action-setup@ea17c68df8912ef543352723c149a84f56e3d413 # v6.1.0
+      - uses: actions/checkout@v7
+      - uses: pnpm/action-setup@v6.0.9
        with:
          version: 10
          run_install: false
-      - uses: actions/setup-node@820762786026740c76f36085b0efc47a31fe5020 # v7
+      - uses: actions/setup-node@v7
        with:
          node-version: "24"
          cache: pnpm
@@ -19,8 +19,8 @@ jobs:
  deploy:
    runs-on: ubuntu-latest
    steps:
-      - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7
-      - uses: actions/setup-node@820762786026740c76f36085b0efc47a31fe5020 # v7
+      - uses: actions/checkout@v7
+      - uses: actions/setup-node@v7
        with:
          node-version: '22'
          cache: npm
@@ -19,12 +19,12 @@ jobs:
  deploy:
    runs-on: ubuntu-latest
    steps:
-      - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7
-      - uses: pnpm/action-setup@ea17c68df8912ef543352723c149a84f56e3d413 # v6.1.0
+      - uses: actions/checkout@v7
+      - uses: pnpm/action-setup@v6.0.9
        with:
          version: 10
          run_install: false
-      - uses: actions/setup-node@820762786026740c76f36085b0efc47a31fe5020 # v7
+      - uses: actions/setup-node@v7
        with:
          node-version: "24"
          cache: pnpm
@@ -1,49 +0,0 @@
-name: Deploy Web Studio
-
-on:
-  push:
-    branches: [studio]
-    paths:
-      - "desktop/frontend-next/**"
-      - ".github/workflows/deploy-web-studio.yml"
-
-permissions:
-  contents: read
-
-concurrency:
-  group: web-studio-production
-  cancel-in-progress: true
-
-jobs:
-  deploy:
-    runs-on: ubuntu-latest
-    defaults:
-      run:
-        working-directory: desktop/frontend-next
-    steps:
-      - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7
-
-      - uses: pnpm/action-setup@0ebf47130e4866e96fce0953f49152a61190b271 # v6.0.9
-        with:
-          version: 10
-          run_install: false
-
-      - uses: actions/setup-node@820762786026740c76f36085b0efc47a31fe5020 # v7
-        with:
-          node-version: "24"
-          cache: pnpm
-          cache-dependency-path: desktop/frontend-next/pnpm-lock.yaml
-
-      - name: Install
-        run: pnpm install --frozen-lockfile
-
-      - name: Build
-        env:
-          VITE_REMOTE_HOME: https://reasonix.io/remote/
-        run: pnpm build
-
-      - name: Deploy
-        env:
-          CLOUDFLARE_API_TOKEN: ${{ secrets.CLOUDFLARE_API_TOKEN }}
-          CLOUDFLARE_ACCOUNT_ID: ${{ secrets.R2_ACCOUNT_ID }}
-        run: npx --yes [email protected] deploy --config wrangler.jsonc
@@ -0,0 +1,39 @@
+name: Documentation impact guard
+
+on:
+  pull_request_target:
+    branches: [main-v2]
+    types: [opened, synchronize, reopened, edited, ready_for_review]
+
+permissions:
+  contents: read
+  pull-requests: read
+
+concurrency:
+  group: docs-impact-${{ github.event.pull_request.number }}
+  cancel-in-progress: true
+
+jobs:
+  docs-impact:
+    runs-on: ubuntu-latest
+    steps:
+      # Execute only the trusted base-branch checker; never run contributor code
+      # from a pull_request_target workflow.
+      - uses: actions/checkout@v7
+        with:
+          ref: ${{ github.event.pull_request.base.sha }}
+
+      - name: Collect PR changed files
+        env:
+          GH_TOKEN: ${{ github.token }}
+          PR_NUMBER: ${{ github.event.pull_request.number }}
+          REPOSITORY: ${{ github.repository }}
+        run: |
+          gh api --paginate "repos/${REPOSITORY}/pulls/${PR_NUMBER}/files" \
+            --jq '.[].filename' > "$RUNNER_TEMP/docs-impact-files.txt"
+
+      - name: Check documentation impact declaration
+        env:
+          PR_BODY: ${{ github.event.pull_request.body }}
+          DOCS_IMPACT_CHANGED_FILES_FILE: ${{ runner.temp }}/docs-impact-files.txt
+        run: bash scripts/check-docs-impact.sh
@@ -32,7 +32,7 @@ jobs:
    environment: e2e-bot
    steps:
      - name: Acknowledge
-        uses: actions/github-script@3a2844b7e9c422d3c10d287c895573f7108da1b3 # v9
+        uses: actions/github-script@v9
        with:
          script: |
            await github.rest.reactions.createForIssueComment({
@@ -42,16 +42,16 @@ jobs:

      # Default-branch checkout: this is where the harness (cmd/e2ebench), the
      # suite, and a run --metrics-capable agent live.
-      - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7
+      - uses: actions/checkout@v7
        with:
          fetch-depth: 0

-      - uses: actions/setup-go@b7ad1dad31e06c5925ef5d2fc7ad053ef454303e # v7
+      - uses: actions/setup-go@v7
        with:
          go-version-file: go.mod
          cache: true

-      - uses: actions/setup-python@5fda3b95a4ea91299a34e894583c3862153e4b97 # v7
+      - uses: actions/setup-python@v7
        with:
          python-version: '3.12'

@@ -20,7 +20,7 @@ jobs:
  classify:
    runs-on: ubuntu-latest
    steps:
-      - uses: actions/github-script@3a2844b7e9c422d3c10d287c895573f7108da1b3 # v9
+      - uses: actions/github-script@v9
        env:
          DEEPSEEK_API_KEY: ${{ secrets.DEEPSEEK_API_KEY }}
        with:
@@ -35,7 +35,7 @@ jobs:
            const body = (issue.body || '').slice(0, 4000);

            const system = [
-              'You categorize GitHub issues for Reasonix, a Go-based AI coding agent with an Electron desktop app and a terminal UI.',
+              'You categorize GitHub issues for Reasonix, a Go-based AI coding agent with a Wails desktop app and a terminal UI.',
              'Pick labels ONLY from these fixed sets. Never invent labels.',
              'area (0-2, the affected subsystem):',
              '  agent: core agent loop / tool-calling / reasoning',
@@ -43,7 +43,7 @@ jobs:
              '  config: configuration, setup wizard, .toml/.env',
              '  updater: auto-update, installer, release packaging',
              '  provider: model providers, model selection/switching',
-              '  desktop: Electron desktop GUI',
+              '  desktop: Wails desktop GUI',
              '  tui: terminal UI / CLI',
              '  skills: skills system',
              '  rendering: terminal rendering / flicker / repaint',
@@ -1,116 +0,0 @@
-name: Issue replies
-
-# Posts maintainer replies as github-actions[bot], so a batch of triage answers
-# spends the repository's quota instead of a person's content-creation limit.
-on:
-  workflow_dispatch:
-    inputs:
-      replies:
-        description: 'JSON array of {"issue": 123, "body": "...", "close": true | "not_planned"} or {"discussion": 123, "body": "...", "close": "resolved" | "outdated" | "duplicate"}'
-        required: true
-        type: string
-      interval:
-        description: Seconds between entries
-        required: false
-        default: '20'
-        type: string
-
-permissions:
-  issues: write
-  pull-requests: write
-  discussions: write
-
-concurrency:
-  group: issue-replies
-  cancel-in-progress: false
-
-jobs:
-  reply:
-    runs-on: win-gpu
-    timeout-minutes: 180
-    steps:
-      - uses: actions/github-script@3a2844b7e9c422d3c10d287c895573f7108da1b3 # v9
-        env:
-          REPLIES: ${{ inputs.replies }}
-          INTERVAL: ${{ inputs.interval }}
-        with:
-          script: |
-            const replies = JSON.parse(process.env.REPLIES);
-            if (!Array.isArray(replies)) throw new Error('replies must be a JSON array');
-            const { owner, repo } = context.repo;
-            const sleep = (ms) => new Promise((resolve) => setTimeout(resolve, ms));
-            const failed = [];
-            // Only a rate limit is worth waiting out; any other refusal is final.
-            const rateLimited = (error) => {
-              const headers = error.response?.headers ?? {};
-              return error.status === 429 ||
-                (error.status === 403 && (headers['retry-after'] !== undefined || headers['x-ratelimit-remaining'] === '0'));
-            };
-            // Discussions have no REST comment API; they are addressed by node id.
-            const replyToDiscussion = async (number, body) => {
-              const q = await github.graphql(
-                `query($owner:String!,$repo:String!,$n:Int!){repository(owner:$owner,name:$repo){discussion(number:$n){id comments(last:100){nodes{body author{login}}}}}}`,
-                { owner, repo, n: number });
-              const d = q.repository.discussion;
-              if (!d) throw new Error('no such discussion');
-              if (d.comments.nodes.some((c) => c.author?.login === 'github-actions' && c.body === body)) return false;
-              await github.graphql(`mutation($id:ID!,$body:String!){addDiscussionComment(input:{discussionId:$id,body:$body}){comment{id}}}`, { id: d.id, body });
-              return true;
-            };
-            const closeReasons = { resolved: 'RESOLVED', outdated: 'OUTDATED', duplicate: 'DUPLICATE' };
-            const closeDiscussion = async (number, reason) => {
-              const q = await github.graphql(
-                `query($owner:String!,$repo:String!,$n:Int!){repository(owner:$owner,name:$repo){discussion(number:$n){id closed}}}`,
-                { owner, repo, n: number });
-              const d = q.repository.discussion;
-              if (d.closed) return;
-              await github.graphql(`mutation($id:ID!,$r:DiscussionCloseReason!){closeDiscussion(input:{discussionId:$id,reason:$r}){discussion{id}}}`, { id: d.id, r: reason });
-            };
-            for (const { issue, discussion, body, close } of replies) {
-              if (Number.isInteger(discussion) && typeof body === 'string' && body.trim() !== '') {
-                try {
-                  const posted = await replyToDiscussion(discussion, body);
-                  if (close !== undefined && close !== false) {
-                    const reason = closeReasons[close];
-                    if (!reason) throw new Error(`unknown close reason ${close}`);
-                    await closeDiscussion(discussion, reason);
-                  }
-                  core.info(`discussion #${discussion}: ${posted ? 'replied' : 'already replied'}${close ? ', closed' : ''}`);
-                } catch (error) {
-                  failed.push(`discussion ${discussion}: ${error.message}`);
-                }
-                await sleep(Math.max(1, Number(process.env.INTERVAL) || 20) * 1000);
-                continue;
-              }
-              if (!Number.isInteger(issue) || typeof body !== 'string' || body.trim() === '') {
-                failed.push(`${issue}: malformed entry`);
-                continue;
-              }
-              try {
-                const comments = await github.paginate(github.rest.issues.listComments, { owner, repo, issue_number: issue, per_page: 100 });
-                const posted = comments.some((c) => c.user?.type === 'Bot' && c.body === body);
-                if (!posted) {
-                  for (let attempt = 0; ; attempt++) {
-                    try {
-                      await github.rest.issues.createComment({ owner, repo, issue_number: issue, body });
-                      break;
-                    } catch (error) {
-                      if (attempt >= 5 || !rateLimited(error)) throw error;
-                      core.warning(`#${issue}: ${error.message}; retrying in 2 minutes`);
-                      await sleep(120_000);
-                    }
-                  }
-                }
-                if (close === true || close === 'not_planned') {
-                  const { data } = await github.rest.issues.get({ owner, repo, issue_number: issue });
-                  if (data.state === 'open') {
-                    await github.rest.issues.update({ owner, repo, issue_number: issue, state: 'closed', state_reason: close === 'not_planned' ? 'not_planned' : 'completed' });
-                  }
-                }
-                core.info(`#${issue}: ${posted ? 'already replied' : 'replied'}${close ? ', closed' : ''}`);
-              } catch (error) {
-                failed.push(`${issue}: ${error.message}`);
-              }
-              await sleep(Math.max(1, Number(process.env.INTERVAL) || 20) * 1000);
-            }
-            if (failed.length) core.setFailed(`not done:\n${failed.join('\n')}`);
@@ -2,8 +2,7 @@ name: Label issue version

 # Issue forms can't let a reporter set a label directly (that needs triage rights),
 # so the bug/feature forms carry a "Version line" dropdown and this workflow reads
-# the submitted value and applies the matching label on their behalf:
-# 0.x -> v1 (TypeScript), 1.x -> v2 (main-v2), 2.x -> v3 (studio).
+# the submitted value and applies the matching v1/v2 label on their behalf.
 on:
  issues:
    types: [opened, edited]
@@ -19,46 +18,46 @@ jobs:
  label:
    runs-on: ubuntu-latest
    steps:
-      - uses: actions/github-script@3a2844b7e9c422d3c10d287c895573f7108da1b3 # v9
+      - uses: actions/github-script@v9
        with:
          script: |
            const body = context.payload.issue.body || '';
            const sectionOf = (name) =>
              body.split(/^###\s+/m).find((s) => new RegExp(`^${name}`, 'i').test(s)) || '';
-            const labelOf = (major) => ({ 0: 'v1', 1: 'v2', 2: 'v3' })[major];

-            const m = sectionOf('Version line').match(/\b([012])\.x\b/);
+            const m = sectionOf('Version line').match(/\bv([12])\b/i);
            if (!m) {
              core.info('No version line found; nothing to label.');
              return;
            }
-            let choice = labelOf(Number(m[1]));
+            let choice = 'v' + m[1];

-            // The dropdown is a choice, and its first entry gets picked by
-            // reporters who never looked. `Exact version` is pasted from
-            // `reasonix --version` or Studio, so when it parses it is the better
-            // evidence. Anything that does not parse (a commit sha, a branch
-            // name) or names no known line leaves the dropdown to decide.
+            // The dropdown is a choice, and v2 sits first, so it gets picked by
+            // reporters who are actually on 0.x. `Exact version` is pasted from
+            // `reasonix --version`, so when it parses it is the better evidence:
+            // 0.x is the TypeScript line, 1.x the Go rewrite. Anything that does
+            // not parse (a commit sha, "main-v2") leaves the dropdown to decide.
            const exact = sectionOf('Exact version').match(/\b(\d+)\.(\d+)\.(\d+)/);
-            const fromVersion = exact && labelOf(Number(exact[1]));
-            if (fromVersion && fromVersion !== choice) {
-              core.info(`Version line says ${choice} but ${exact[0]} says ${fromVersion}; trusting the version.`);
-              choice = fromVersion;
+            if (exact) {
+              const fromVersion = Number(exact[1]) === 0 ? 'v1' : 'v2';
+              if (fromVersion !== choice) {
+                core.info(`Version line says ${choice} but ${exact[0]} says ${fromVersion}; trusting the version.`);
+                choice = fromVersion;
+              }
            }
+            const other = choice === 'v2' ? 'v1' : 'v2';
            await github.rest.issues.addLabels({
              ...context.repo,
              issue_number: context.issue.number,
              labels: [choice],
            });
-            // Keep the line labels mutually exclusive in case an edit flipped the choice.
-            for (const other of ['v1', 'v2', 'v3'].filter((l) => l !== choice)) {
-              try {
-                await github.rest.issues.removeLabel({
-                  ...context.repo,
-                  issue_number: context.issue.number,
-                  name: other,
-                });
-              } catch (e) {
-                core.info(`No ${other} label to remove.`);
-              }
+            // Keep v1/v2 mutually exclusive in case an edit flipped the choice.
+            try {
+              await github.rest.issues.removeLabel({
+                ...context.repo,
+                issue_number: context.issue.number,
+                name: other,
+              });
+            } catch (e) {
+              core.info(`No ${other} label to remove.`);
            }
@@ -0,0 +1,47 @@
+name: Deploy site
+
+on:
+  push:
+    branches: [main-v2]
+    paths: ['site/**', 'release-notes/**', '.github/workflows/pages.yml']
+  workflow_dispatch:
+
+permissions:
+  contents: read
+  pages: write
+  id-token: write
+
+concurrency:
+  group: pages
+  cancel-in-progress: false
+
+jobs:
+  build:
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v7
+      - uses: actions/setup-node@v7
+        with:
+          node-version: '22'
+          cache: npm
+          cache-dependency-path: site/package-lock.json
+      - name: Build
+        working-directory: site
+        env:
+          GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
+        run: |
+          npm ci
+          npm run build
+      - uses: actions/upload-pages-artifact@v5
+        with:
+          path: site/dist
+
+  deploy:
+    needs: build
+    runs-on: ubuntu-latest
+    environment:
+      name: github-pages
+      url: ${{ steps.deploy.outputs.page_url }}
+    steps:
+      - id: deploy
+        uses: actions/deploy-pages@v5
@@ -23,6 +23,6 @@ jobs:
  label:
    runs-on: ubuntu-latest
    steps:
-      - uses: actions/labeler@bf12e9b00b37c5c0ca2b87b79b2daf7891dbda13 # v7
+      - uses: actions/labeler@v7
        with:
          sync-labels: false
@@ -20,7 +20,7 @@ jobs:
  label:
    runs-on: ubuntu-latest
    steps:
-      - uses: actions/github-script@3a2844b7e9c422d3c10d287c895573f7108da1b3 # v9
+      - uses: actions/github-script@v9
        with:
          script: |
            const base = context.payload.pull_request.base.ref;
@@ -0,0 +1,101 @@
+name: Prepare release
+
+on:
+  workflow_dispatch:
+    inputs:
+      version:
+        description: "Official version without a tag prefix (for example 1.20.0)"
+        required: true
+        type: string
+
+permissions:
+  contents: write
+  pull-requests: write
+
+jobs:
+  draft:
+    name: generate reviewable bilingual draft
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v7
+        with:
+          fetch-depth: 0
+
+      - uses: actions/setup-node@v7
+        with:
+          node-version: "22"
+
+      - name: Select the review branch
+        env:
+          GH_TOKEN: ${{ github.token }}
+          VERSION: ${{ inputs.version }}
+        run: |
+          set -euo pipefail
+          git config user.name "github-actions[bot]"
+          git config user.email "41898282+github-actions[bot]@users.noreply.github.com"
+          git fetch origin main-v2
+
+          if [[ ! "$VERSION" =~ ^(0|[1-9][0-9]*)\.(0|[1-9][0-9]*)\.(0|[1-9][0-9]*)$ ]]; then
+            echo "::error::version must be MAJOR.MINOR.PATCH"
+            exit 1
+          fi
+          branch="release-notes/v${VERSION}"
+          if git ls-remote --exit-code --heads origin "$branch" >/dev/null 2>&1; then
+            git fetch origin "refs/heads/$branch:refs/remotes/origin/$branch"
+            git checkout -b "$branch" --track "origin/$branch"
+            git merge --no-edit origin/main-v2
+          else
+            git checkout -b "$branch"
+          fi
+          echo "RELEASE_NOTES_BRANCH=$branch" >> "$GITHUB_ENV"
+
+      - name: Generate and validate release notes
+        env:
+          DEEPSEEK_API_KEY: ${{ secrets.DEEPSEEK_API_KEY }}
+          DEEPSEEK_MODEL: deepseek-v4-pro
+          GH_TOKEN: ${{ github.token }}
+          VERSION: ${{ inputs.version }}
+        run: |
+          set -euo pipefail
+          node scripts/generate-release-notes.mjs --version "$VERSION"
+          node scripts/release-notes.mjs validate
+          node --test scripts/release-notes.test.mjs
+          node scripts/release-notes.mjs render --version "$VERSION" --output /tmp/release-notes.md
+          {
+            echo "## Review draft for v${VERSION#v}"
+            echo
+            cat /tmp/release-notes.md
+          } >> "$GITHUB_STEP_SUMMARY"
+
+      - name: Open or update the review PR
+        env:
+          GH_TOKEN: ${{ github.token }}
+          VERSION: ${{ inputs.version }}
+        run: |
+          set -euo pipefail
+          git add release-notes/releases.json
+          if ! git diff --cached --quiet; then
+            git commit -m "docs(release): prepare v${VERSION#v} notes" \
+              -m "Summary:
+          Generate a bilingual, product-focused draft from merged pull request metadata. Reuse the selected release-bound PR when one is available.
+
+          Verification:
+          Validate the catalog, citations, bilingual fields, and rendered GitHub release notes before committing."
+          fi
+          git push -u origin "$RELEASE_NOTES_BRANCH"
+          if ! gh pr list --head "$RELEASE_NOTES_BRANCH" --state open --json number --jq 'length > 0' | grep -q true; then
+            if ! gh pr create \
+              --base main-v2 \
+              --head "$RELEASE_NOTES_BRANCH" \
+              --title "docs(release): 准备 v${VERSION#v} 中英更新日志" \
+              --body-file /tmp/release-notes.md; then
+              echo "::warning::GitHub Actions could not open the PR; the reviewed branch is preserved."
+              {
+                echo "## Manual PR handoff"
+                echo
+                echo '```sh'
+                echo "gh pr create --repo ${{ github.repository }} --base main-v2 --head $RELEASE_NOTES_BRANCH --fill"
+                echo '```'
+              } >> "$GITHUB_STEP_SUMMARY"
+            fi
+          fi
@@ -0,0 +1,26 @@
+name: Release notes
+
+on:
+  pull_request:
+    paths:
+      - "release-notes/**"
+      - "scripts/release-notes*.mjs"
+      - "site/src/pages/changelog/**"
+      - "site/src/components/ChangelogPage.astro"
+      - "desktop/frontend/src/App.tsx"
+      - "desktop/frontend/src/components/SettingsPanel.tsx"
+      - "desktop/frontend/src/components/UpdateBanner.tsx"
+
+permissions:
+  contents: read
+
+jobs:
+  validate:
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v7
+      - uses: actions/setup-node@v7
+        with:
+          node-version: "22"
+      - run: node scripts/release-notes.mjs validate
+      - run: node --test scripts/release-notes.test.mjs
@@ -0,0 +1,206 @@
+name: Release npm
+
+# npm line. Stable releases are called by release-stable.yml after its single
+# GitHub environment approval. Direct prerelease tag publication is disabled;
+# manual workflow_dispatch recovers only an approved Stable version from an
+# existing tag. Historical canary and next package identities remain readable,
+# but normal publication advances only the official release.
+on:
+  workflow_dispatch:
+    inputs:
+      channel:
+        description: "Standalone npm recovery channel"
+        required: true
+        default: stable
+        type: choice
+        options:
+          - stable
+      base_version:
+        description: "Stable version to recover exactly"
+        required: true
+        type: string
+      tag:
+        description: "stable: existing npm tag to publish (for example npm-v1.18.0)"
+        required: false
+        type: string
+  workflow_call:
+    inputs:
+      channel:
+        description: "Publish channel selected by the approved release orchestrator"
+        required: true
+        type: string
+      base_version:
+        description: "Version to publish"
+        required: true
+        type: string
+      tag:
+        description: "Existing npm tag to check out for stable publication"
+        required: false
+        default: ""
+        type: string
+      approved_cli_tag:
+        description: "Stable CLI tag recorded by the approved orchestrator"
+        required: true
+        type: string
+      approved_sha:
+        description: "Immutable commit recorded by the approved orchestrator"
+        required: true
+        type: string
+      orchestrated:
+        description: "True only when called by an approved release orchestrator"
+        required: false
+        default: false
+        type: boolean
+      orchestrator:
+        description: "Trusted release orchestrator (legacy Preview calls remain readable)"
+        required: false
+        default: stable
+        type: string
+      preview_number:
+        description: "Legacy Preview ordinal for old workflow-call compatibility"
+        required: false
+        default: ""
+        type: string
+
+permissions:
+  contents: read
+
+concurrency:
+  # Serialize every publisher for one npm dist-tag. Historical canary calls
+  # remain ordered even though normal publication now uses Stable only.
+  group: release-npm-${{ inputs.channel || 'next' }}
+  cancel-in-progress: false
+
+jobs:
+  orchestration-guard:
+    name: verify approved orchestrator
+    if: ${{ inputs.orchestrated }}
+    runs-on: ubuntu-latest
+    permissions:
+      contents: read
+    steps:
+      - uses: actions/checkout@v7
+        with:
+          fetch-depth: 0
+          ref: ${{ github.sha }}
+      - name: Verify caller and approved release ref
+        env:
+          ACTUAL_CALLER_WORKFLOW_REF: ${{ github.workflow_ref }}
+          EXPECTED_CALLER_WORKFLOW_REF: ${{ format('{0}/.github/workflows/release-{1}.yml@{2}', github.repository, inputs.orchestrator, github.ref) }}
+          CALLER_EVENT_NAME: ${{ github.event_name }}
+          CALLER_REF: ${{ github.ref }}
+          CALLER_REF_PROTECTED: ${{ github.ref_protected }}
+          CALLER_WORKFLOW_SHA: ${{ github.workflow_sha }}
+          CALLER_SHA: ${{ github.sha }}
+          APPROVED_CLI_TAG: ${{ inputs.approved_cli_tag }}
+          APPROVED_SHA: ${{ inputs.approved_sha }}
+          APPROVED_CHANNEL: ${{ inputs.orchestrator }}
+          RELEASE_TAG: ${{ inputs.approved_cli_tag }}
+          VERIFY_RELEASE_CHECKOUT: false
+        run: |
+          bash scripts/verify-release-authorization.sh
+          bash scripts/verify-release-tag.sh
+
+  release-gate:
+    name: approve standalone npm release
+    if: ${{ !inputs.orchestrated }}
+    runs-on: ubuntu-latest
+    environment: release
+    steps:
+      - env:
+          RELEASE_TAG: ${{ inputs.tag || github.ref_name }}
+        run: echo "Approved standalone npm release $RELEASE_TAG"
+
+  cache-guard:
+    name: cache hit guard
+    needs: [orchestration-guard, release-gate]
+    if: ${{ always() && !cancelled() && ((inputs.orchestrated && needs.orchestration-guard.result == 'success') || (!inputs.orchestrated && needs.release-gate.result == 'success')) }}
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v7
+        with:
+          ref: ${{ inputs.approved_sha || inputs.tag || github.ref }}
+      - uses: actions/setup-go@v7
+        with:
+          go-version-file: go.mod
+          cache: true
+      - run: ./scripts/cache-guard.sh
+      - name: Verify embedded documentation identity
+        env:
+          DOCS_BUILD_VERSION: v${{ inputs.base_version }}
+          DOCS_SOURCE_REVISION: ${{ inputs.approved_sha }}
+        run: |
+          if [ ! -f scripts/verify-embedded-docs.sh ]; then
+            echo "Legacy candidate predates the embedded docs contract; skipping."
+            exit 0
+          fi
+          revision="$DOCS_SOURCE_REVISION"
+          if [ -z "$revision" ]; then revision="$(git rev-parse HEAD)"; fi
+          bash scripts/verify-embedded-docs.sh "$DOCS_BUILD_VERSION" "$revision"
+
+  npm:
+    name: publish npm packages
+    needs: cache-guard
+    if: ${{ always() && !cancelled() && needs.cache-guard.result == 'success' }}
+    runs-on: ubuntu-latest
+    # Orchestrated releases have already passed their GitHub environment
+    # approval. Direct prereleases and manual Stable recovery pass release-gate.
+    steps:
+      - uses: actions/checkout@v7
+        with:
+          ref: ${{ inputs.approved_sha || inputs.tag || github.ref }}
+      - name: Load approved standalone recovery control plane
+        if: ${{ !inputs.orchestrated }}
+        env:
+          RECOVERY_CONTROL_SHA: ${{ github.workflow_sha }}
+        run: |
+          set -euo pipefail
+          git fetch --no-tags --depth=1 origin "$RECOVERY_CONTROL_SHA"
+          git restore --source="$RECOVERY_CONTROL_SHA" -- \
+            npm/publish.mjs \
+            scripts/finalize-npm-official-release.mjs
+      - uses: actions/setup-go@v7
+        with:
+          go-version-file: go.mod
+          cache: true
+      - uses: actions/setup-node@v7
+        with:
+          node-version: '22'
+          registry-url: 'https://registry.npmjs.org'
+      # Stable publication uses the exact npm-v* tag. build.mjs strips the
+      # leading `npm-`/`v`; historical Preview inputs remain only for old
+      # workflow-call compatibility and are not reachable from a public entry.
+      - name: Resolve version
+        id: ver
+        env:
+          EVENT_NAME: ${{ github.event_name }}
+          IN_ORCHESTRATED: ${{ inputs.orchestrated }}
+          IN_CHANNEL: ${{ inputs.channel }}
+          IN_BASE_VERSION: ${{ inputs.base_version }}
+          IN_TAG: ${{ inputs.tag }}
+          REF_NAME: ${{ github.ref_name }}
+          RUN_NUMBER: ${{ github.run_number }}
+          IN_PREVIEW_NUMBER: ${{ inputs.preview_number }}
+        run: bash scripts/resolve-npm-release.sh
+      - name: Revalidate approved release ref
+        if: ${{ inputs.orchestrated }}
+        env:
+          RELEASE_TAG: ${{ inputs.approved_cli_tag }}
+          APPROVED_SHA: ${{ inputs.approved_sha }}
+        run: bash scripts/verify-release-tag.sh
+      - name: Publish or recover immutable npm packages
+        env:
+          VERSION_ARG: ${{ steps.ver.outputs.arg }}
+          NODE_AUTH_TOKEN: ${{ secrets.NPM_TOKEN }}
+        run: node npm/build.mjs "$VERSION_ARG" --publish
+      - name: Align legacy aliases with the official release
+        if: ${{ inputs.channel == 'stable' }}
+        env:
+          VERSION: ${{ inputs.base_version }}
+          APPROVED_SHA: ${{ inputs.approved_sha }}
+          NODE_AUTH_TOKEN: ${{ secrets.NPM_TOKEN }}
+        run: |
+          set -euo pipefail
+          candidate="$APPROVED_SHA"
+          if [ -z "$candidate" ]; then candidate="$(git rev-parse HEAD)"; fi
+          EXPECTED_SHA="$candidate" node scripts/finalize-npm-official-release.mjs "$VERSION"
@@ -0,0 +1,42 @@
+name: Relay stable release
+run-name: Relay stable release ${{ github.ref_name }}
+
+# The tag event itself has a tag-shaped origin. Relay the immutable tag name to
+# release-stable.yml on protected main-v2 so the production SignPath policy can
+# allow exactly one branch name without trusting wildcard tag-like branches.
+on:
+  push:
+    tags:
+      - 'v*'
+      - '!v*-*'
+
+permissions:
+  actions: write
+  contents: read
+
+jobs:
+  relay:
+    name: dispatch protected stable control plane
+    runs-on: ubuntu-latest
+    steps:
+      - name: Dispatch release-stable.yml on main-v2
+        uses: actions/github-script@v9
+        env:
+          RELEASE_TAG: ${{ github.ref_name }}
+          CONTROL_PLANE_REF: ${{ github.event.repository.default_branch }}
+        with:
+          script: |
+            if (process.env.CONTROL_PLANE_REF !== 'main-v2') {
+              core.setFailed(`Unexpected default branch: ${process.env.CONTROL_PLANE_REF}`);
+              return;
+            }
+            await github.rest.actions.createWorkflowDispatch({
+              owner: context.repo.owner,
+              repo: context.repo.repo,
+              workflow_id: 'release-stable.yml',
+              ref: process.env.CONTROL_PLANE_REF,
+              inputs: {
+                tag: process.env.RELEASE_TAG,
+                allow_recovery: 'false',
+              },
+            });
@@ -0,0 +1,278 @@
+name: Release stable
+run-name: Release stable ${{ inputs.tag || github.ref_name }}
+
+# Stable publication has exactly one GitHub environment gate. Push the CLI,
+# npm, and desktop tags atomically; release-stable-trigger.yml relays the vX.Y.Z
+# tag to this workflow on protected main-v2. Preflight verifies that all three
+# tags point to the reviewed Notes candidate on main-v2 history, and one GitHub
+# approval releases every surface. After approval, a zero-publication Windows
+# preflight verifies both architectures and signing stages before publication.
+# Keeping the control-plane ref on main-v2 lets SignPath restrict
+# production signing to that one protected origin instead of trusting wildcard
+# tag-like branch names. Manual recovery uses the same fixed control plane while
+# preserving an older tagged candidate on main-v2 history.
+on:
+  workflow_dispatch:
+    inputs:
+      tag:
+        description: "Existing stable CLI tag to recover (for example v1.18.0)"
+        required: true
+        type: string
+      publish_cli:
+        description: "Recover the CLI/Homebrew channel"
+        required: false
+        default: true
+        type: boolean
+      publish_npm:
+        description: "Recover the npm channel"
+        required: false
+        default: true
+        type: boolean
+      publish_desktop:
+        description: "Recover the Desktop/R2 channel"
+        required: false
+        default: true
+        type: boolean
+      allow_recovery:
+        description: "Allow an existing stable tag on main-v2 history (manual recovery only)"
+        required: false
+        default: true
+        type: boolean
+
+concurrency:
+  group: stable-release-${{ inputs.tag || github.ref_name }}
+  cancel-in-progress: false
+
+# Reusable release workflows can only reduce caller permissions, so the
+# orchestrator grants the union needed by CLI/Desktop publication.
+permissions:
+  actions: write
+  contents: write
+
+jobs:
+  preflight:
+    name: validate stable release set
+    runs-on: ubuntu-latest
+    permissions:
+      actions: read
+      contents: read
+    outputs:
+      version: ${{ steps.release.outputs.version }}
+      cli_tag: ${{ steps.release.outputs.cli_tag }}
+      npm_tag: ${{ steps.release.outputs.npm_tag }}
+      desktop_tag: ${{ steps.release.outputs.desktop_tag }}
+      sha: ${{ steps.release.outputs.sha }}
+    steps:
+      - uses: actions/checkout@v7
+        with:
+          fetch-depth: 0
+          # A recovery dispatch uses the fixed workflow/scripts from protected
+          # main-v2. Publishers still check out the immutable approved tag SHA.
+          ref: ${{ github.sha }}
+      - uses: actions/setup-go@v7
+        with:
+          go-version-file: go.mod
+          cache: true
+      - uses: actions/setup-node@v7
+        with:
+          node-version: "22"
+      - name: Resolve stable release
+        id: release
+        env:
+          RELEASE_TAG: ${{ inputs.tag || github.ref_name }}
+          ALLOW_STABLE_RECOVERY: ${{ inputs.allow_recovery }}
+        run: bash scripts/resolve-stable-release.sh
+      - name: Revalidate normal release candidate and exact push CI
+        if: ${{ !inputs.allow_recovery }}
+        env:
+          GH_TOKEN: ${{ github.token }}
+          RELEASE_REPOSITORY: ${{ github.repository }}
+          RELEASE_CI_WAIT_SECONDS: 60
+          RELEASE_VERSION: ${{ steps.release.outputs.version }}
+          RELEASE_SHA: ${{ steps.release.outputs.sha }}
+        run: |
+          bash scripts/validate-stable-candidate.sh "$RELEASE_VERSION" "$RELEASE_SHA"
+          bash scripts/verify-release-push-ci.sh "$RELEASE_SHA"
+      - name: Validate reviewed release notes
+        run: node scripts/release-notes.mjs render --version "${{ steps.release.outputs.cli_tag }}" --output /tmp/release-notes.md
+      # Recovery builds deliberately check out the immutable tagged candidate,
+      # which can predate its reviewed release-note entry. Carry the exact file
+      # validated by this protected control-plane job into both publishers.
+      - name: Upload reviewed release notes
+        uses: actions/upload-artifact@v7
+        with:
+          name: orchestrator-reviewed-release-notes
+          path: /tmp/release-notes.md
+          if-no-files-found: error
+          retention-days: 1
+      - name: Cache hit guard
+        run: ./scripts/cache-guard.sh
+
+  authorize:
+    name: approve stable release
+    needs: preflight
+    runs-on: ubuntu-latest
+    environment: release
+    permissions:
+      contents: read
+    outputs:
+      version: ${{ steps.approved.outputs.version }}
+      cli_tag: ${{ steps.approved.outputs.cli_tag }}
+      npm_tag: ${{ steps.approved.outputs.npm_tag }}
+      desktop_tag: ${{ steps.approved.outputs.desktop_tag }}
+      sha: ${{ steps.approved.outputs.sha }}
+    steps:
+      - name: Record approved release
+        id: approved
+        env:
+          VERSION: ${{ needs.preflight.outputs.version }}
+          CLI_TAG: ${{ needs.preflight.outputs.cli_tag }}
+          NPM_TAG: ${{ needs.preflight.outputs.npm_tag }}
+          DESKTOP_TAG: ${{ needs.preflight.outputs.desktop_tag }}
+          RELEASE_SHA: ${{ needs.preflight.outputs.sha }}
+        run: |
+          {
+            echo "version=$VERSION"
+            echo "cli_tag=$CLI_TAG"
+            echo "npm_tag=$NPM_TAG"
+            echo "desktop_tag=$DESKTOP_TAG"
+            echo "sha=$RELEASE_SHA"
+          } >> "$GITHUB_OUTPUT"
+          echo "Approved stable release $VERSION at $RELEASE_SHA"
+
+  cli:
+    name: publish CLI and Homebrew
+    needs: [authorize, signpath-preflight]
+    if: ${{ always() && !cancelled() && needs.authorize.result == 'success' && (needs.signpath-preflight.result == 'success' || needs.signpath-preflight.result == 'skipped') && (github.event_name != 'workflow_dispatch' || inputs.publish_cli) }}
+    uses: ./.github/workflows/release.yml
+    with:
+      tag: ${{ needs.authorize.outputs.cli_tag }}
+      approved_cli_tag: ${{ needs.authorize.outputs.cli_tag }}
+      approved_sha: ${{ needs.authorize.outputs.sha }}
+      orchestrated: true
+    secrets: inherit
+
+  npm:
+    name: publish npm
+    needs: [authorize, signpath-preflight]
+    if: ${{ always() && !cancelled() && needs.authorize.result == 'success' && (needs.signpath-preflight.result == 'success' || needs.signpath-preflight.result == 'skipped') && (github.event_name != 'workflow_dispatch' || inputs.publish_npm) }}
+    uses: ./.github/workflows/release-npm.yml
+    with:
+      channel: stable
+      base_version: ${{ needs.authorize.outputs.version }}
+      tag: ${{ needs.authorize.outputs.npm_tag }}
+      approved_cli_tag: ${{ needs.authorize.outputs.cli_tag }}
+      approved_sha: ${{ needs.authorize.outputs.sha }}
+      orchestrated: true
+    secrets: inherit
+
+  signpath-preflight:
+    name: verify stable SignPath control plane
+    needs: authorize
+    if: ${{ github.event_name != 'workflow_dispatch' || inputs.publish_desktop }}
+    uses: ./.github/workflows/release-desktop.yml
+    with:
+      channel: stable
+      tag: ${{ needs.authorize.outputs.desktop_tag }}
+      approved_cli_tag: ${{ needs.authorize.outputs.cli_tag }}
+      approved_sha: ${{ needs.authorize.outputs.sha }}
+      orchestrated: true
+      signing_preflight: true
+    secrets: inherit
+
+  desktop:
+    name: publish desktop
+    needs: [authorize, signpath-preflight]
+    if: ${{ always() && !cancelled() && needs.authorize.result == 'success' && needs.signpath-preflight.result == 'success' && (github.event_name != 'workflow_dispatch' || inputs.publish_desktop) }}
+    uses: ./.github/workflows/release-desktop.yml
+    with:
+      channel: stable
+      tag: ${{ needs.authorize.outputs.desktop_tag }}
+      approved_cli_tag: ${{ needs.authorize.outputs.cli_tag }}
+      approved_sha: ${{ needs.authorize.outputs.sha }}
+      orchestrated: true
+      signing_preflight_verified: true
+    secrets: inherit
+
+  postflight:
+    name: verify stable release artifacts
+    needs: [authorize, cli, npm, desktop]
+    if: ${{ always() && !cancelled() }}
+    runs-on: ubuntu-latest
+    permissions:
+      actions: write
+      contents: write
+    steps:
+      - name: Require every publisher to succeed
+        env:
+          PUBLISH_CLI: ${{ github.event_name != 'workflow_dispatch' || inputs.publish_cli }}
+          PUBLISH_NPM: ${{ github.event_name != 'workflow_dispatch' || inputs.publish_npm }}
+          PUBLISH_DESKTOP: ${{ github.event_name != 'workflow_dispatch' || inputs.publish_desktop }}
+          CLI_RESULT: ${{ needs.cli.result }}
+          NPM_RESULT: ${{ needs.npm.result }}
+          DESKTOP_RESULT: ${{ needs.desktop.result }}
+        run: |
+          set -euo pipefail
+          for channel in cli npm desktop; do
+            selected_var="PUBLISH_${channel^^}"
+            result_var="${channel^^}_RESULT"
+            selected="${!selected_var}"
+            result="${!result_var}"
+            if [ "$selected" != "true" ]; then
+              echo "$channel recovery skipped; public postflight will still verify it"
+              continue
+            fi
+            if [ "$result" != "success" ]; then
+              echo "::error::$channel stable publisher result is $result, expected success"
+              exit 1
+            fi
+          done
+      - uses: actions/checkout@v7
+        with:
+          # Postflight belongs to the trusted control plane, not the old build
+          # candidate, which may predate this verifier.
+          ref: ${{ github.sha }}
+      - uses: actions/setup-node@v7
+        with:
+          node-version: "22"
+      - name: Verify public artifacts and npm latest
+        env:
+          GH_TOKEN: ${{ github.token }}
+          RELEASE_REPOSITORY: ${{ github.repository }}
+          RELEASE_VERSION: ${{ needs.authorize.outputs.version }}
+          CLI_TAG: ${{ needs.authorize.outputs.cli_tag }}
+          DESKTOP_TAG: ${{ needs.authorize.outputs.desktop_tag }}
+        run: bash scripts/verify-stable-release-artifacts.sh
+      - name: Publish exact Stable release record
+        env:
+          GH_TOKEN: ${{ github.token }}
+          VERSION: ${{ needs.authorize.outputs.version }}
+          CLI_TAG: ${{ needs.authorize.outputs.cli_tag }}
+          RELEASE_SHA: ${{ needs.authorize.outputs.sha }}
+        run: |
+          set -euo pipefail
+          if ! node -e '
+            const catalog = require("./release-notes/releases.json");
+            const release = catalog.releases.find((item) => item.version === process.env.VERSION);
+            process.exit(release?.status === "reviewed" ? 0 : 1);
+          '; then
+            echo "Legacy Stable notes do not require a publication marker"
+            exit 0
+          fi
+          node scripts/release-event.mjs generate \
+            --version "$VERSION" --sha "$RELEASE_SHA" \
+            --published-at "$(gh api "repos/${{ github.repository }}/releases/tags/$CLI_TAG" --jq .published_at)" \
+            --output /tmp/release-event.json
+          existing="$(mktemp -d)"
+          if gh release download "$CLI_TAG" --pattern release-event.json --dir "$existing" 2>/dev/null; then
+            cmp -s /tmp/release-event.json "$existing/release-event.json" || {
+              echo "::error::published release-event.json differs from the approved Stable event"
+              exit 1
+            }
+          else
+            gh release upload "$CLI_TAG" /tmp/release-event.json
+          fi
+      - name: Refresh public changelog
+        env:
+          GH_TOKEN: ${{ github.token }}
+        run: gh workflow run pages.yml --ref main-v2
@@ -0,0 +1,62 @@
+name: Request verification on released issues
+
+# A PR that writes "Fixes #N" auto-closes its issue on merge. One that writes
+# "Refs"/"Related"/"Addresses" does not, so the report stays open waiting for a
+# follow-up nobody remembers to send after the release. This sends it.
+
+on:
+  workflow_dispatch:
+    inputs:
+      tag:
+        description: Release tag to announce (e.g. desktop-v1.18.0)
+        required: true
+        type: string
+      dry_run:
+        description: Print the targets without commenting
+        required: false
+        default: true
+        type: boolean
+  release:
+    types: [published]
+  push:
+    branches:
+      - main-v2
+    paths:
+      - '.github/workflows/release-verify-issues.yml'
+      - 'scripts/release-verify-issues.mjs'
+      - 'scripts/release-verify-issues.test.mjs'
+
+permissions:
+  contents: read
+  issues: write
+
+concurrency:
+  group: release-verify-issues-${{ github.event.inputs.tag || github.event.release.tag_name || github.ref }}
+  cancel-in-progress: false
+
+jobs:
+  notify:
+    if: github.repository == 'esengine/DeepSeek-Reasonix'
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v7
+        with:
+          fetch-depth: 0
+      - uses: actions/setup-node@v7
+        with:
+          node-version: '22'
+      - name: Test notifier
+        run: node --test scripts/release-verify-issues.test.mjs
+      # A push that only touched the script has no release to announce; the test
+      # above is the whole point of that trigger.
+      - name: Request verification
+        if: github.event_name != 'push'
+        env:
+          GH_TOKEN: ${{ secrets.GITHUB_TOKEN }}
+          TAG: ${{ github.event.inputs.tag || github.event.release.tag_name }}
+          DRY_RUN: ${{ github.event.inputs.dry_run || 'false' }}
+        run: |
+          set -euo pipefail
+          args=(--tag "$TAG")
+          if [ "$DRY_RUN" = "true" ]; then args+=(--dry-run); fi
+          node scripts/release-verify-issues.mjs "${args[@]}"
@@ -0,0 +1,581 @@
+name: Release
+
+# Native CLI binary line. Official releases are called by the protected Stable
+# orchestrator. Manual dispatch remains available only for official recovery;
+# historical Preview inputs below are workflow-call compatibility, not a public
+# publication entrypoint.
+on:
+  workflow_dispatch:
+    inputs:
+      channel:
+        description: "Standalone CLI recovery channel"
+        required: true
+        default: stable
+        type: choice
+        options: [stable]
+      tag:
+        description: "Existing official CLI tag (for example v1.18.0)"
+        required: true
+        type: string
+  workflow_call:
+    inputs:
+      channel:
+        description: "Native CLI release channel selected by the approved orchestrator"
+        required: false
+        default: stable
+        type: string
+      tag:
+        description: "Existing CLI tag selected by the approved release orchestrator"
+        required: true
+        type: string
+      approved_cli_tag:
+        description: "Stable CLI tag recorded by the approved orchestrator"
+        required: true
+        type: string
+      approved_sha:
+        description: "Immutable commit recorded by the approved orchestrator"
+        required: true
+        type: string
+      orchestrated:
+        description: "True only when called by an approved release orchestrator"
+        required: false
+        default: false
+        type: boolean
+      orchestrator:
+        description: "Trusted release orchestrator (legacy Preview calls remain readable)"
+        required: false
+        default: stable
+        type: string
+      allow_preview_recovery:
+        description: "Legacy compatibility for already-created Preview runs"
+        required: false
+        default: false
+        type: boolean
+
+permissions:
+  contents: write # create the release and upload archives
+
+concurrency:
+  # A channel pointer is a monotonic public state machine. Serialize all
+  # publishers for the same channel so an older recovery run cannot pass its
+  # read-before-write window after a newer release has published.
+  group: release-cli-${{ inputs.channel || 'stable' }}
+  cancel-in-progress: false
+
+jobs:
+  resolve:
+    name: resolve CLI release
+    runs-on: ubuntu-latest
+    outputs:
+      tag: ${{ steps.release.outputs.tag }}
+      version: ${{ steps.release.outputs.version }}
+      base_version: ${{ steps.release.outputs.base_version }}
+      notes_version: ${{ steps.release.outputs.notes_version }}
+      channel: ${{ steps.release.outputs.channel }}
+      prerelease: ${{ steps.release.outputs.prerelease }}
+      sha: ${{ steps.candidate.outputs.sha }}
+    steps:
+      - uses: actions/checkout@v7
+        with:
+          fetch-depth: 0
+          ref: ${{ github.sha }}
+      - name: Resolve channel and tag
+        id: release
+        env:
+          EVENT_NAME: ${{ github.event_name }}
+          IN_ORCHESTRATED: ${{ inputs.orchestrated }}
+          IN_CHANNEL: ${{ inputs.channel }}
+          IN_TAG: ${{ inputs.tag }}
+          REF_NAME: ${{ github.ref_name }}
+          CALLER_REF: ${{ github.ref }}
+          CALLER_REF_PROTECTED: ${{ github.ref_protected }}
+        run: bash scripts/resolve-cli-release.sh
+      - name: Record immutable candidate
+        id: candidate
+        env:
+          RELEASE_TAG: ${{ steps.release.outputs.tag }}
+          RELEASE_CHANNEL: ${{ steps.release.outputs.channel }}
+          IN_ORCHESTRATED: ${{ inputs.orchestrated }}
+          IN_ORCHESTRATOR: ${{ inputs.orchestrator }}
+          ALLOW_PREVIEW_RECOVERY: ${{ inputs.allow_preview_recovery }}
+        run: |
+          set -euo pipefail
+          git fetch origin main-v2
+          sha="$(git rev-parse "$RELEASE_TAG^{commit}")"
+          if ! git merge-base --is-ancestor "$sha" origin/main-v2; then
+            echo "::error::$RELEASE_TAG points to $sha, which is not on main-v2 history"
+            exit 1
+          fi
+          if [ "$ALLOW_PREVIEW_RECOVERY" = "true" ]; then
+            if [ "$IN_ORCHESTRATED" != "true" ] || [ "$IN_ORCHESTRATOR" != "preview" ] || [ "$RELEASE_CHANNEL" != "preview" ]; then
+              echo "::error::Preview recovery requires the approved Preview orchestrator"
+              exit 1
+            fi
+          elif [ "$RELEASE_CHANNEL" = "preview" ] && [ "$sha" != "$(git rev-parse origin/main-v2)" ]; then
+              echo "::error::CLI Preview must tag current main-v2; $RELEASE_TAG points to $sha"
+              exit 1
+          fi
+          echo "sha=$sha" >> "$GITHUB_OUTPUT"
+      - name: Verify existing protected tag
+        env:
+          RELEASE_TAG: ${{ steps.release.outputs.tag }}
+          APPROVED_SHA: ${{ steps.candidate.outputs.sha }}
+          VERIFY_RELEASE_CHECKOUT: false
+        run: bash scripts/verify-release-tag.sh
+
+  orchestration-guard:
+    name: verify approved orchestrator
+    needs: resolve
+    if: ${{ inputs.orchestrated }}
+    runs-on: ubuntu-latest
+    permissions:
+      contents: read
+    steps:
+      - uses: actions/checkout@v7
+        with:
+          fetch-depth: 0
+          ref: ${{ github.sha }}
+      - name: Verify caller and approved release ref
+        env:
+          ACTUAL_CALLER_WORKFLOW_REF: ${{ github.workflow_ref }}
+          EXPECTED_CALLER_WORKFLOW_REF: ${{ format('{0}/.github/workflows/release-{1}.yml@{2}', github.repository, inputs.orchestrator, github.ref) }}
+          CALLER_EVENT_NAME: ${{ github.event_name }}
+          CALLER_REF: ${{ github.ref }}
+          CALLER_REF_PROTECTED: ${{ github.ref_protected }}
+          CALLER_WORKFLOW_SHA: ${{ github.workflow_sha }}
+          CALLER_SHA: ${{ github.sha }}
+          APPROVED_CLI_TAG: ${{ inputs.approved_cli_tag }}
+          APPROVED_SHA: ${{ inputs.approved_sha }}
+          APPROVED_CHANNEL: ${{ inputs.orchestrator }}
+          RELEASE_TAG: ${{ inputs.tag }}
+          VERIFY_RELEASE_CHECKOUT: false
+        run: |
+          bash scripts/verify-release-authorization.sh
+          bash scripts/verify-release-tag.sh
+
+  release-gate:
+    name: approve standalone CLI release
+    needs: resolve
+    if: ${{ !inputs.orchestrated }}
+    runs-on: ubuntu-latest
+    environment: release
+    steps:
+      - env:
+          RELEASE_TAG: ${{ needs.resolve.outputs.tag }}
+          RELEASE_CHANNEL: ${{ needs.resolve.outputs.channel }}
+        run: echo "Approved standalone CLI $RELEASE_CHANNEL release $RELEASE_TAG"
+
+  cache-guard:
+    name: cache hit guard
+    needs: [resolve, orchestration-guard, release-gate]
+    if: ${{ always() && !cancelled() && needs.resolve.result == 'success' && ((inputs.orchestrated && needs.orchestration-guard.result == 'success') || (!inputs.orchestrated && needs.release-gate.result == 'success')) }}
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v7
+        with:
+          ref: ${{ needs.resolve.outputs.sha }}
+      - uses: actions/setup-go@v7
+        with:
+          go-version-file: go.mod
+          cache: true
+      - run: ./scripts/cache-guard.sh
+      - name: Verify embedded documentation identity
+        env:
+          DOCS_BUILD_VERSION: ${{ needs.resolve.outputs.tag }}
+          DOCS_SOURCE_REVISION: ${{ needs.resolve.outputs.sha }}
+        run: |
+          if [ ! -f scripts/verify-embedded-docs.sh ]; then
+            echo "Legacy candidate predates the embedded docs contract; skipping."
+            exit 0
+          fi
+          bash scripts/verify-embedded-docs.sh "$DOCS_BUILD_VERSION" "$DOCS_SOURCE_REVISION"
+
+  goreleaser:
+    name: archives + checksums + homebrew tap
+    needs: [resolve, cache-guard]
+    if: ${{ always() && !cancelled() && needs.cache-guard.result == 'success' }}
+    runs-on: ubuntu-latest
+    # The Stable caller has already passed the single GitHub release approval.
+    # Official standalone recovery passes release-gate above. This job therefore
+    # must not add a second GitHub environment approval.
+    steps:
+      - uses: actions/checkout@v7
+        with:
+          fetch-depth: 0
+          ref: ${{ needs.resolve.outputs.sha }}
+      # Recovery may build an immutable tag that predates the current recovery
+      # policy. Keep product sources pinned above, but execute publication
+      # decisions from the protected workflow commit.
+      - uses: actions/checkout@v7
+        with:
+          fetch-depth: 0
+          path: release-control
+          ref: ${{ github.workflow_sha }}
+      - name: Isolate release-control checkout from product git state
+        run: |
+          set -euo pipefail
+          git_common_dir="$(git rev-parse --path-format=absolute --git-common-dir)"
+          exclude_file="$git_common_dir/info/exclude"
+          if ! grep -qxF '/release-control/' "$exclude_file"; then
+            printf '%s\n' '/release-control/' >> "$exclude_file"
+          fi
+          git check-ignore -q release-control/
+          dirty="$(git status --porcelain --untracked-files=all)"
+          if [ -n "$dirty" ]; then
+            printf 'product checkout is dirty before release:\n%s\n' "$dirty" >&2
+            exit 1
+          fi
+      - uses: actions/setup-go@v7
+        with:
+          go-version-file: go.mod
+          cache: true
+      - uses: actions/setup-node@v7
+        with:
+          node-version: "22"
+      - name: Download orchestrator-reviewed release notes
+        if: ${{ inputs.orchestrated }}
+        uses: actions/download-artifact@v8
+        with:
+          name: orchestrator-reviewed-release-notes
+          path: /tmp/orchestrator-reviewed-release-notes
+      - name: Use orchestrator-reviewed release notes
+        if: ${{ inputs.orchestrated }}
+        run: |
+          test -s /tmp/orchestrator-reviewed-release-notes/release-notes.md
+          cp /tmp/orchestrator-reviewed-release-notes/release-notes.md /tmp/release-notes.md
+      - name: Render reviewed release notes
+        if: ${{ !inputs.orchestrated }}
+        env:
+          RELEASE_TAG: ${{ needs.resolve.outputs.notes_version }}
+        run: node scripts/release-notes.mjs render --version "$RELEASE_TAG" --output /tmp/release-notes.md
+      - name: Revalidate approved release ref
+        env:
+          RELEASE_TAG: ${{ needs.resolve.outputs.tag }}
+          APPROVED_SHA: ${{ needs.resolve.outputs.sha }}
+        run: bash scripts/verify-release-tag.sh
+      - name: Decide whether CLI artifacts need publication
+        id: publication
+        env:
+          GH_TOKEN: ${{ github.token }}
+          TAG: ${{ needs.resolve.outputs.tag }}
+          CHANNEL: ${{ needs.resolve.outputs.channel }}
+          PRERELEASE: ${{ needs.resolve.outputs.prerelease }}
+        run: |
+          set -euo pipefail
+          validation_channel="$CHANNEL"
+          if [ "$CHANNEL" = "stable" ] && [ "$PRERELEASE" = "true" ]; then
+            validation_channel=any
+          fi
+          release_json=/tmp/existing-cli-release.json
+          release_error=/tmp/existing-cli-release.error
+          checksums=/tmp/existing-cli-release-SHA256SUMS
+          if gh api "repos/${{ github.repository }}/releases/tags/$TAG" \
+            >"$release_json" 2>"$release_error"; then
+            gh release download "$TAG" -R "${{ github.repository }}" \
+              --pattern SHA256SUMS --output "$checksums"
+            decision="$(
+              bash release-control/scripts/decide-cli-release-publication.sh \
+                "$validation_channel" "$TAG" "${{ github.repository }}" \
+                "$release_json" "$checksums"
+            )"
+            echo "existing CLI release $TAG is complete and checksum-bound; reusing it"
+          elif grep -Eiq 'HTTP 404|Not Found' "$release_error"; then
+            decision="$(
+              bash release-control/scripts/decide-cli-release-publication.sh \
+                "$validation_channel" "$TAG" "${{ github.repository }}" - -
+            )"
+            echo "CLI release $TAG does not exist; GoReleaser will publish it"
+          else
+            cat "$release_error" >&2
+            exit 1
+          fi
+          test "$decision" = "publish" -o "$decision" = "reuse"
+          echo "decision=$decision" >> "$GITHUB_OUTPUT"
+      - uses: goreleaser/goreleaser-action@v7
+        if: ${{ steps.publication.outputs.decision == 'publish' }}
+        with:
+          version: '~> v2'
+          args: release --clean
+        env:
+          GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
+          HOMEBREW_TAP_TOKEN: ${{ needs.resolve.outputs.channel == 'stable' && secrets.HOMEBREW_TAP_TOKEN || '' }}
+          # workflow_dispatch recovery runs have a branch-shaped GITHUB_REF even
+          # though checkout is on the release tag. Pin GoReleaser explicitly, and
+          # avoid ambiguity from the three release tags sharing one commit.
+          GORELEASER_CURRENT_TAG: ${{ needs.resolve.outputs.tag }}
+
+      - name: Publish product release notes
+        env:
+          GH_TOKEN: ${{ github.token }}
+          TAG: ${{ needs.resolve.outputs.tag }}
+        run: gh release edit "$TAG" --notes-file /tmp/release-notes.md
+
+      - name: Publish CLI release metadata to R2
+        env:
+          GH_TOKEN: ${{ github.token }}
+          TAG: ${{ needs.resolve.outputs.tag }}
+          NOTES_TAG: ${{ needs.resolve.outputs.notes_version }}
+          HAS_R2: ${{ secrets.R2_ACCESS_KEY_ID != '' && secrets.R2_SECRET_ACCESS_KEY != '' && secrets.R2_ACCOUNT_ID != '' && secrets.R2_BUCKET != '' }}
+          AWS_ACCESS_KEY_ID: ${{ secrets.R2_ACCESS_KEY_ID }}
+          AWS_SECRET_ACCESS_KEY: ${{ secrets.R2_SECRET_ACCESS_KEY }}
+          AWS_DEFAULT_REGION: auto
+          R2_ACCOUNT_ID: ${{ secrets.R2_ACCOUNT_ID }}
+          R2_BUCKET: ${{ secrets.R2_BUCKET }}
+        run: |
+          set -euo pipefail
+          if [ "$HAS_R2" != "true" ]; then
+            echo "R2 secrets not configured; skipping CLI release metadata"
+            exit 0
+          fi
+
+          channel=""
+          if [[ "$TAG" =~ ^v(0|[1-9][0-9]*)\.(0|[1-9][0-9]*)\.(0|[1-9][0-9]*)$ ]]; then
+            channel="stable"
+          elif [[ "$TAG" =~ ^v(0|[1-9][0-9]*)\.(0|[1-9][0-9]*)\.(0|[1-9][0-9]*)-preview\.(0|[1-9][0-9]*)$ ]]; then
+            channel="preview"
+          else
+            echo "internal CLI release $TAG; publishing only immutable metadata"
+          fi
+
+          required_assets='[
+            "reasonix-darwin-amd64.tar.gz",
+            "reasonix-darwin-arm64.tar.gz",
+            "reasonix-linux-amd64.tar.gz",
+            "reasonix-linux-arm64.tar.gz",
+            "reasonix-windows-amd64.zip",
+            "reasonix-windows-arm64.zip",
+            "SHA256SUMS"
+          ]'
+          gh api "repos/${{ github.repository }}/releases/tags/$TAG" > /tmp/cli-release.raw.json
+          jq --arg tag "$TAG" --arg notes_tag "$NOTES_TAG" --argjson required "$required_assets" '
+            if .tag_name != $tag then error("release tag mismatch") else . end |
+            if .draft then error("draft release cannot be published") else . end |
+            . as $release |
+            ($release.assets | map({key: .name, value: .}) | from_entries) as $assets |
+            if ($required | all(. as $name | $assets[$name] != null))
+            then {
+              tag_name: $release.tag_name,
+              prerelease: $release.prerelease,
+              html_url: $release.html_url,
+              release_notes_url: ("https://reasonix.io/changelog/" + $notes_tag + "/"),
+              assets: [
+                $required[] as $name |
+                $assets[$name] |
+                {
+                  name: .name,
+                  browser_download_url: .browser_download_url,
+                  size: .size
+                }
+              ]
+            }
+            else error("release is missing one or more required CLI assets")
+            end
+          ' /tmp/cli-release.raw.json > /tmp/cli-release.json
+          if [ -n "$channel" ]; then
+            bash scripts/validate-cli-release-manifest.sh \
+              "$channel" "$TAG" "${{ github.repository }}" /tmp/cli-release.json "$NOTES_TAG"
+          else
+            bash scripts/validate-cli-release-manifest.sh \
+              any "$TAG" "${{ github.repository }}" /tmp/cli-release.json "$NOTES_TAG"
+          fi
+
+          endpoint="https://${R2_ACCOUNT_ID}.r2.cloudflarestorage.com"
+          validation_channel="${channel:-any}"
+          immutable_key="cli/releases/${TAG}/latest.json"
+          immutable_error="$(mktemp)"
+          if aws s3 cp "s3://${R2_BUCKET}/${immutable_key}" /tmp/cli-release.immutable.json \
+            --endpoint-url "$endpoint" 2>"$immutable_error"; then
+            bash scripts/validate-cli-release-manifest.sh \
+              "legacy-${validation_channel}" "$TAG" "${{ github.repository }}" \
+              /tmp/cli-release.immutable.json "$NOTES_TAG"
+            if ! bash scripts/compare-cli-release-manifests.sh \
+              /tmp/cli-release.json /tmp/cli-release.immutable.json; then
+              echo "::error::immutable CLI release metadata for $TAG already exists with different content"
+              exit 1
+            fi
+            echo "immutable CLI release metadata for $TAG already exists; preserving it"
+          elif grep -Eiq '404|NoSuchKey|Not Found' "$immutable_error"; then
+            aws s3 cp /tmp/cli-release.json "s3://${R2_BUCKET}/${immutable_key}" \
+              --endpoint-url "$endpoint" \
+              --content-type "application/json; charset=utf-8" \
+              --cache-control "public, max-age=31536000, immutable"
+          else
+            cat "$immutable_error" >&2
+            exit 1
+          fi
+          rm -f "$immutable_error"
+
+          aws s3 cp "s3://${R2_BUCKET}/${immutable_key}" /tmp/cli-release.immutable.json \
+            --endpoint-url "$endpoint"
+          bash scripts/validate-cli-release-manifest.sh \
+            "legacy-${validation_channel}" "$TAG" "${{ github.repository }}" \
+            /tmp/cli-release.immutable.json "$NOTES_TAG"
+          bash scripts/compare-cli-release-manifests.sh \
+            /tmp/cli-release.json /tmp/cli-release.immutable.json
+
+          if [ -z "$channel" ]; then
+            echo "internal CLI release $TAG; Stable and Preview pointers remain unchanged"
+            exit 0
+          fi
+
+          current_tag=""
+          pointer_error="$(mktemp)"
+          if aws s3 cp "s3://${R2_BUCKET}/cli/${channel}/latest.json" /tmp/cli-release.pointer.json \
+            --endpoint-url "$endpoint" 2>"$pointer_error"; then
+            current_tag="$(jq -er '.tag_name | strings' /tmp/cli-release.pointer.json)"
+            bash scripts/validate-cli-release-manifest.sh \
+              "legacy-${channel}" "$current_tag" "${{ github.repository }}" \
+              /tmp/cli-release.pointer.json "$current_tag"
+          elif grep -Eiq '404|NoSuchKey|Not Found' "$pointer_error"; then
+            echo "CLI $channel pointer does not exist yet"
+          else
+            cat "$pointer_error" >&2
+            exit 1
+          fi
+          rm -f "$pointer_error"
+
+          pointer_manifest=-
+          if [ -n "$current_tag" ]; then
+            pointer_manifest=/tmp/cli-release.pointer.json
+          fi
+          pointer_decision="$(
+            bash scripts/decide-cli-pointer-update.sh \
+              "$channel" /tmp/cli-release.json "$pointer_manifest"
+          )"
+          if [ "$pointer_decision" = "skip" ]; then
+            echo "CLI $channel pointer remains ${current_tag:-unset}; candidate $TAG is not newer and needs no repair"
+            exit 0
+          fi
+          aws s3 cp /tmp/cli-release.json "s3://${R2_BUCKET}/cli/${channel}/latest.json" \
+            --endpoint-url "$endpoint" \
+            --content-type "application/json; charset=utf-8" \
+            --cache-control "public, max-age=300, stale-if-error=86400"
+
+          aws s3 cp "s3://${R2_BUCKET}/cli/${channel}/latest.json" /tmp/cli-release.pointer.json \
+            --endpoint-url "$endpoint"
+          bash scripts/validate-cli-release-manifest.sh \
+            "$channel" "$TAG" "${{ github.repository }}" \
+            /tmp/cli-release.pointer.json "$NOTES_TAG"
+          cmp -s /tmp/cli-release.json /tmp/cli-release.pointer.json
+          echo "CLI $channel pointer -> $TAG"
+
+      - name: Attach desktop manifest compatibility asset
+        env:
+          GH_TOKEN: ${{ github.token }}
+          TAG: ${{ needs.resolve.outputs.tag }}
+          HAS_R2: ${{ secrets.R2_ACCESS_KEY_ID != '' && secrets.R2_SECRET_ACCESS_KEY != '' && secrets.R2_ACCOUNT_ID != '' && secrets.R2_BUCKET != '' }}
+          R2_ACCOUNT_ID: ${{ secrets.R2_ACCOUNT_ID }}
+          R2_BUCKET: ${{ secrets.R2_BUCKET }}
+        run: |
+          set -euo pipefail
+          case "$TAG" in
+            *-*)
+              echo "prerelease $TAG — GitHub latest does not move here; skipping desktop manifest compatibility asset"
+              exit 0
+              ;;
+          esac
+          if [ "$HAS_R2" != "true" ]; then
+            echo "R2 secrets not configured; skipping desktop manifest compatibility asset"
+            exit 0
+          fi
+          # dl.reasonix.io serves 403 to GitHub Actions egress IPs (Cloudflare bot
+          # protection), so read the manifest over the authenticated S3 API instead
+          # of the public edge.
+          aws configure set aws_access_key_id "${{ secrets.R2_ACCESS_KEY_ID }}"
+          aws configure set aws_secret_access_key "${{ secrets.R2_SECRET_ACCESS_KEY }}"
+          aws configure set region auto
+          aws s3 cp "s3://${R2_BUCKET}/latest/latest.json" latest.raw.json \
+            --endpoint-url "https://${R2_ACCOUNT_ID}.r2.cloudflarestorage.com"
+          jq '.download_page = "https://reasonix.io/?download=desktop#start"' latest.raw.json > latest.json
+          jq -e '
+            ([.platforms[] | (.url, .sig)] |
+              all(type == "string" and startswith("https://dl.reasonix.io/") and (contains("/releases/latest/") | not)))
+          ' latest.json >/dev/null
+          gh release upload "$TAG" latest.json --clobber
+
+      # The compatibility asset exists for pre-v1.16 desktop updaters that
+      # still poll GitHub's repository-wide latest URL. Desktop releases now
+      # own that Latest badge, but this check still exercises the public fallback
+      # path exactly the way those clients fetch it: anonymously, over the public
+      # edge, with a Go client UA. Unlike dl.reasonix.io (whose bot protection
+      # 403s Actions egress — see the R2 note above), GitHub serves its own
+      # runners, so this can hard-fail. #5826/#5858 shipped a broken update check
+      # for weeks precisely because nothing exercised the public path. Retries
+      # cover the release CDN propagating the freshly uploaded asset.
+      - name: Smoke public compatibility manifest
+        env:
+          TAG: ${{ needs.resolve.outputs.tag }}
+          HAS_R2: ${{ secrets.R2_ACCESS_KEY_ID != '' && secrets.R2_SECRET_ACCESS_KEY != '' && secrets.R2_ACCOUNT_ID != '' && secrets.R2_BUCKET != '' }}
+        run: |
+          set -euo pipefail
+          case "$TAG" in
+            *-*)
+              echo "prerelease $TAG — no compatibility asset uploaded; skipping"
+              exit 0
+              ;;
+          esac
+          if [ "$HAS_R2" != "true" ]; then
+            echo "R2 secrets not configured; no compatibility asset uploaded; skipping"
+            exit 0
+          fi
+          url="https://github.com/${{ github.repository }}/releases/latest/download/latest.json"
+          for attempt in 1 2 3 4 5 6; do
+            if curl -fsSL -A "Go-http-client/2.0" -o /tmp/compat-latest.json "$url"; then
+              jq -e '(.version | type == "string") and (.platforms | type == "object")' /tmp/compat-latest.json >/dev/null
+              echo "public compatibility manifest OK (desktop version $(jq -r .version /tmp/compat-latest.json))"
+              exit 0
+            fi
+            echo "attempt $attempt failed; retrying in 10s"
+            sleep 10
+          done
+          echo "::error::public compatibility manifest unreachable at $url"
+          exit 1
+
+      # A stable CLI release must never leave the npm line behind: v1.17.5
+      # shipped as binaries/Homebrew while npm `latest` still pointed at 0.53.2
+      # (#5822) — every `npm update -g` user was silently downgraded to a
+      # months-old version, and nothing noticed because the npm line
+      # (release-npm.yml, `npm-vX.Y.Z` tags) is triggered independently and the
+      # stable npm tag was simply never pushed. release-npm.yml's own verify
+      # step only guards runs that happen; this guard catches the run that
+      # DIDN'T.
+      #
+      # Two distinct states, two responses (the approved orchestrator starts the
+      # CLI and npm reusable workflows concurrently, and npm dist-tags propagate
+      # asynchronously, so "tag pushed but latest not moved yet" is a NORMAL
+      # mid-release state, not a failure):
+      #   - npm-v<version> tag missing  -> hard fail. This is the #5822 gap:
+      #     nobody pushed the npm release at all.
+      #   - tag pushed, latest lagging  -> poll briefly, then WARN and pass.
+      #     The npm job may still be publishing; release-npm.yml's verify step
+      #     owns asserting the dist-tag lands.
+      - name: Check npm latest dist-tag freshness
+        env:
+          TAG: ${{ needs.resolve.outputs.tag }}
+        run: |
+          set -euo pipefail
+          case "$TAG" in
+            *-*)
+              echo "prerelease $TAG — npm latest does not move on prereleases; skipping"
+              exit 0
+              ;;
+          esac
+          version="${TAG#v}"
+          if ! git ls-remote --exit-code origin "refs/tags/npm-v$version" >/dev/null; then
+            echo "::error::the npm-v$version tag was never pushed — the npm channel is being left behind and 'npm update -g' users will be downgraded to the old 'latest'. Push it: git tag npm-v$version ${TAG} && git push origin npm-v$version (or 'npm dist-tag add reasonix@$version latest' for an already-published version)."
+            exit 1
+          fi
+          for attempt in 1 2 3 4 5 6; do
+            got="$(npm view reasonix dist-tags.latest 2>/dev/null || true)"
+            if [ -n "$got" ]; then
+              newest="$(printf '%s\n%s\n' "$got" "$version" | sort -V | tail -1)"
+              if [ "$newest" = "$got" ]; then
+                echo "npm latest -> $got (>= $version) OK"
+                exit 0
+              fi
+            fi
+            echo "npm latest -> ${got:-<unreadable>}, want >= $version (attempt $attempt)"
+            sleep 10
+          done
+          echo "::warning::npm-v$version is pushed but npm 'latest' is still ${got:-<unreadable>} — the concurrent npm publish is likely still running or propagating. Monitor the npm job; its verify step asserts the dist-tag lands."
+          exit 0
@@ -24,7 +24,7 @@ jobs:
    if: github.event.comment.user.type != 'Bot' && github.event.comment.author_association == 'NONE'
    runs-on: ubuntu-latest
    steps:
-      - uses: actions/github-script@3a2844b7e9c422d3c10d287c895573f7108da1b3 # v9
+      - uses: actions/github-script@v9
        with:
          script: |
            const EMOJI = /[\u{1F300}-\u{1FAFF}\u{2600}-\u{27BF}\u{2B00}-\u{2BFF}\u{FE0F}]/gu;
@@ -40,10 +40,10 @@ jobs:
    if: github.repository == 'esengine/DeepSeek-Reasonix'
    runs-on: ubuntu-latest
    steps:
-      - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7
+      - uses: actions/checkout@v7
        with:
          fetch-depth: 0
-      - uses: actions/setup-node@820762786026740c76f36085b0efc47a31fe5020 # v7
+      - uses: actions/setup-node@v7
        with:
          node-version: '22'
      - name: Test sweeper
@@ -1,117 +0,0 @@
-name: Studio Certum signing smoke test
-
-# Signs two throwaway probes with the certificate release-studio.yml would use,
-# and publishes nothing. It reports the signer's subject, which is the value
-# STUDIO_SIGNING_SUBJECT must hold.
-on:
-  workflow_dispatch:
-
-permissions:
-  contents: read
-
-concurrency:
-  group: certum-signing
-  cancel-in-progress: false
-
-jobs:
-  sign:
-    if: github.repository == 'esengine/DeepSeek-Reasonix' && github.ref == 'refs/heads/studio'
-    runs-on: windows-2022
-    environment: studio-release
-    timeout-minutes: 20
-    steps:
-      - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7
-        with:
-          persist-credentials: false
-
-      - uses: actions/setup-go@b7ad1dad31e06c5925ef5d2fc7ad053ef454303e # v7
-        with:
-          go-version-file: go.mod
-          cache: false
-
-      - name: Check signing configuration
-        shell: pwsh
-        env:
-          CERTUM_USERNAME: ${{ secrets.CERTUM_USERNAME }}
-          CERTUM_OTP_URI: ${{ secrets.CERTUM_OTP_URI }}
-          CERTUM_KEY_ID: ${{ secrets.CERTUM_KEY_ID }}
-        run: |
-          $ErrorActionPreference = 'Stop'
-          if ([string]::IsNullOrWhiteSpace($env:CERTUM_USERNAME)) { throw 'Missing secret CERTUM_USERNAME' }
-          if ($env:CERTUM_OTP_URI -notmatch '^otpauth://totp/') { throw 'Secret CERTUM_OTP_URI must be a complete TOTP URI' }
-          if ($env:CERTUM_KEY_ID -notmatch '^[0-9a-fA-F]{40}$') { throw 'Secret CERTUM_KEY_ID must be the 40-character SHA-1 certificate thumbprint' }
-
-      - name: Build open-source signing probes
-        shell: pwsh
-        run: |
-          $ErrorActionPreference = 'Stop'
-          # This minimal probe is covered by the repository's MIT license.
-          New-Item -ItemType Directory payload | Out-Null
-          'package main; func main() { println("Reasonix Studio Certum signing probe") }' | Set-Content probe.go
-          $env:GOOS = 'windows'
-          $env:GOARCH = 'amd64'
-          $env:CGO_ENABLED = '0'
-          go build -trimpath -o payload/probe.exe probe.go
-          if ($LASTEXITCODE -ne 0) { throw 'Probe compilation failed' }
-          'probe.exe' | Set-Content payload/signing-files.txt
-          Copy-Item payload/probe.exe installer-probe.exe
-
-      - name: Connect to SimplySign cloud certificate
-        uses: ./.github/actions/setup-certum
-        with:
-          username: ${{ secrets.CERTUM_USERNAME }}
-          otp-uri: ${{ secrets.CERTUM_OTP_URI }}
-          thumbprint: ${{ secrets.CERTUM_KEY_ID }}
-
-      - name: Sign and verify payload and installer probes
-        shell: pwsh
-        env:
-          CERTUM_KEY_ID: ${{ secrets.CERTUM_KEY_ID }}
-        run: |
-          ./scripts/sign-certum.ps1 -PayloadDirectory payload
-          ./scripts/sign-certum.ps1 -FilePath installer-probe.exe
-
-      # The session signs for any process on the runner while it is up.
-      - name: Disconnect from Certum
-        if: always()
-        shell: pwsh
-        run: Get-Process -Name '*SimplySign*' -ErrorAction SilentlyContinue | Stop-Process -Force
-
-      # With the switch on, an unset subject fails here as it would in a
-      # release; with it off, the value is reported so it can be recorded.
-      - name: Report and check the signer
-        shell: pwsh
-        env:
-          SIGNING_ENABLED: ${{ vars.STUDIO_SIGNING_ENABLED }}
-          EXPECTED_SUBJECT: ${{ vars.STUDIO_SIGNING_SUBJECT }}
-        run: |
-          $ErrorActionPreference = 'Stop'
-          $failed = $false
-          foreach ($file in @('payload/probe.exe', 'installer-probe.exe')) {
-            $signature = Get-AuthenticodeSignature -LiteralPath $file
-            $certificate = $signature.SignerCertificate
-            $stamp = if ($signature.TimeStamperCertificate) { $signature.TimeStamperCertificate.Subject } else { '(none)' }
-            @("### $file", '', "- subject: ``$($certificate.Subject)``", "- issuer: ``$($certificate.Issuer)``",
-              "- thumbprint: ``$($certificate.Thumbprint)``", "- timestamp: ``$stamp``", '') >> $env:GITHUB_STEP_SUMMARY
-            if (-not $signature.TimeStamperCertificate) {
-              Write-Output "::error title=studio-signing.timestamp-missing::$file carries no timestamp"
-              $failed = $true
-            }
-            if ([string]::IsNullOrWhiteSpace($certificate.Issuer)) {
-              Write-Output "::error title=studio-signing.issuer-missing::$file has a signer certificate without an issuer"
-              $failed = $true
-            }
-            if ([string]::IsNullOrWhiteSpace($env:EXPECTED_SUBJECT)) {
-              if ($env:SIGNING_ENABLED -eq 'true') {
-                Write-Output "::error title=studio-signing.subject-unset::STUDIO_SIGNING_SUBJECT is empty while STUDIO_SIGNING_ENABLED=true"
-                $failed = $true
-              } else {
-                Write-Output "::warning::STUDIO_SIGNING_SUBJECT is not set; record '$($certificate.Subject)'"
-              }
-            } elseif ($certificate.Subject -cne $env:EXPECTED_SUBJECT) {
-              Write-Output "::error title=studio-signing.unexpected-signer::STUDIO_SIGNING_SUBJECT is '$($env:EXPECTED_SUBJECT)' but $file is signed by '$($certificate.Subject)'"
-              $failed = $true
-            }
-          }
-          if ($failed) { exit 1 }
-          'Certum payload and installer signing, signer, trust and timestamps passed.' >> $env:GITHUB_STEP_SUMMARY
@@ -1,107 +0,0 @@
-name: Studio Windows self-test
-
-# An unsigned Windows installer for trying a branch on a real machine before it
-# is released. It runs on the intranet runner, so it is dispatch-only: the
-# repository is public, and nothing a pull request writes may run there.
-on:
-  workflow_dispatch:
-    inputs:
-      ref:
-        description: Branch, tag or commit to build
-        required: false
-        default: studio
-        type: string
-
-permissions:
-  contents: read
-
-concurrency:
-  group: studio-selftest-windows
-  cancel-in-progress: true
-
-jobs:
-  build:
-    runs-on: win-gpu
-    timeout-minutes: 90
-    steps:
-      # The machine's PATH finds WSL's bash before Git's, and a bash step run in
-      # WSL sees neither this checkout's toolchain nor Windows paths. cmd, since
-      # the machine's execution policy refuses the script a powershell step is.
-      - name: Put Git Bash first on PATH
-        shell: cmd
-        run: '>>"%GITHUB_PATH%" echo C:\Program Files\Git\bin'
-
-      - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7
-        with:
-          ref: ${{ inputs.ref }}
-
-      - uses: actions/setup-go@b7ad1dad31e06c5925ef5d2fc7ad053ef454303e # v7
-        with:
-          go-version-file: go.mod
-          cache: true
-          cache-dependency-path: go.sum
-
-      - uses: pnpm/action-setup@ea17c68df8912ef543352723c149a84f56e3d413 # v6.1.0
-        with:
-          version: 10
-          run_install: false
-
-      - uses: actions/setup-node@820762786026740c76f36085b0efc47a31fe5020 # v7
-        with:
-          node-version: "24"
-          cache: pnpm
-          cache-dependency-path: |
-            desktop/frontend-next/pnpm-lock.yaml
-            desktop/electron/pnpm-lock.yaml
-
-      # A patch above the newest release, as a prerelease: it ranks ahead of
-      # what is published, so the build does not offer itself that release as
-      # an update, and behind the next one, which it does offer.
-      - name: Choose the version
-        id: version
-        shell: bash
-        run: |
-          latest="$(git ls-remote --tags origin 'refs/tags/studio-v*' | sed -n 's#.*refs/tags/studio-v\([0-9.]*\)$#\1#p' | sort -V | tail -1)"
-          test -n "$latest" || { echo "::error::no studio-v* release tag"; exit 1; }
-          IFS=. read -r major minor patch <<< "$latest"
-          version="$major.$minor.$((patch + 1))-selftest.${{ github.run_number }}"
-          echo "version=$version" >> "$GITHUB_OUTPUT"
-          echo "Version: \`$version\` from \`$(git rev-parse --short HEAD)\`" >> "$GITHUB_STEP_SUMMARY"
-
-      # The intranet's route to the module proxy drops connections now and then.
-      # What arrives stays in this machine's module cache, so a retry resumes.
-      - name: Fetch Go modules
-        shell: bash
-        run: for i in 1 2 3 4 5; do go mod download && exit 0; sleep 10; done; exit 1
-
-      - name: Build the SPA
-        working-directory: desktop/frontend-next
-        shell: bash
-        run: pnpm install --frozen-lockfile && pnpm build
-
-      - name: Build the kernel
-        working-directory: desktop/electron
-        shell: bash
-        env:
-          REASONIX_VERSION: v${{ steps.version.outputs.version }}
-        run: pnpm install --frozen-lockfile && pnpm run build:host
-
-      - name: Build the bundle
-        working-directory: desktop/electron
-        shell: bash
-        run: pnpm exec electron-builder --dir -c.extraMetadata.version="${{ steps.version.outputs.version }}"
-
-      - name: Package the installer
-        working-directory: desktop/electron
-        shell: bash
-        run: |
-          pnpm exec electron-builder --prepackaged ../../dist/win-unpacked --win nsis \
-            -c.extraMetadata.version="${{ steps.version.outputs.version }}"
-          sha256sum ../../dist/ReasonixStudio-windows-amd64-installer.exe | tee -a "$GITHUB_STEP_SUMMARY"
-
-      - uses: actions/upload-artifact@043fb46d1a93c77aae656e7c1c64a875d1fc6a0a # v7.0.1
-        with:
-          name: studio-selftest-windows-${{ steps.version.outputs.version }}
-          path: dist/ReasonixStudio-windows-amd64-installer.exe
-          if-no-files-found: error
-          retention-days: 7
@@ -1,303 +0,0 @@
-name: Studio
-
-# Studio's own surface — desktop/frontend-next (the SPA) and the Electron shell
-# that serves it — exists only on the studio branch, so it cannot live in ci.yml
-# without failing on main-v2 where neither directory is present. ci.yml covers
-# the kernel; this file covers the frontend build, the desktop module's Go side,
-# and the packaging contract on real artifacts.
-on:
-  push:
-    branches: [studio]
-  pull_request:
-    branches: [studio]
-
-permissions:
-  contents: read
-
-concurrency:
-  group: studio-${{ github.ref }}
-  cancel-in-progress: true
-
-jobs:
-  frontend:
-    name: frontend-next
-    runs-on: ubuntu-latest
-    defaults:
-      run:
-        working-directory: desktop/frontend-next
-    steps:
-      - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7
-
-      - uses: pnpm/action-setup@ea17c68df8912ef543352723c149a84f56e3d413 # v6.1.0
-        with:
-          version: 10
-          run_install: false
-
-      - uses: actions/setup-node@820762786026740c76f36085b0efc47a31fe5020 # v7
-        with:
-          node-version: "24"
-          cache: pnpm
-          cache-dependency-path: desktop/frontend-next/pnpm-lock.yaml
-
-      - name: Install
-        run: pnpm install --frozen-lockfile
-
-      - name: typecheck
-        run: pnpm typecheck
-
-      # The reducer and the event port carry the two pieces of logic a type
-      # error cannot reach: what a card means when a turn is interrupted, and
-      # what the client does about a frame that never arrived.
-      - name: test
-        run: pnpm test
-
-      # What the UI census holds. The gate reads structure — which inputs reach
-      # a mutating endpoint, and which of those the product names an action for
-      # — so UNDECLARED_MUTATION == 0 is a fact about the tree, not a lint over
-      # how anything is spelled.
-      - name: census invariants
-        run: pnpm census > /dev/null
-
-      # And that the analyzer still judges the way it did. _fx is a corpus
-      # written to exercise the rules, so its reports move only when a judgement
-      # moves; the product's name every root by file and line and are stale the
-      # moment a button moves, which is why they are not frozen.
-      - name: census fixture reports have not drifted
-        run: pnpm census:verify
-
-      # `build` runs tsc -b before vite, so this also proves the project
-      # references resolve — typecheck alone uses --noEmit and does not.
-      - name: build
-        run: pnpm build
-
-      # The published bundle must not carry MockPort's scripted fixtures: a
-      # release that cannot reach the kernel has to say so, not act out a fake
-      # session. main.tsx guards the import behind import.meta.env.DEV, and this
-      # is what proves the guard still holds.
-      - name: no fixtures in the bundle
-        run: |
-          if grep -rl 网关瞬时态 dist/ 2>/dev/null; then
-            echo "::error::MockPort fixtures reached the production bundle - check the import.meta.env.DEV guard in main.tsx"
-            exit 1
-          fi
-
-  desktop:
-    name: desktop (${{ matrix.os }})
-    runs-on: ${{ matrix.os }}
-    strategy:
-      fail-fast: false
-      matrix:
-        os: [ubuntu-22.04, windows-latest, macos-latest]
-    defaults:
-      run:
-        working-directory: desktop
-    steps:
-      - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7
-
-      - uses: actions/setup-go@b7ad1dad31e06c5925ef5d2fc7ad053ef454303e # v7
-        with:
-          go-version-file: desktop/go.mod
-          cache: true
-          cache-dependency-path: desktop/go.sum
-
-      # The desktop module's Go tests: the elevated update helper, the
-      # release-side tools and the Windows uninstall planner. The code this
-      # gate covers is what replaces a running application on a user's disk.
-      # The module is CGO-free since the Wails shell retired, so no platform
-      # carries build dependencies of its own.
-      # Same command as `make studio-test`; this job is already in desktop/.
-      - name: go test
-        run: go test ./...
-
-      # The host sits in the main module and nothing else builds it, so its
-      # dependency closure is checked here: it must stay CGO-free and
-      # platform-neutral or the Electron bundle cannot ship it. Cross-compiling
-      # is host-independent, so once is enough.
-      - name: host stays pure Go and platform-neutral
-        if: runner.os == 'Linux'
-        working-directory: .
-        run: |
-          for os in windows linux darwin; do
-            CGO_ENABLED=0 GOOS=$os GOARCH=amd64 go build -o /dev/null ./cmd/reasonix-studio-host
-          done
-
-  # The packaging contract, on real artifacts rather than on the config that
-  # describes them. Unsigned on purpose: signing and notarization are a
-  # distribution concern and belong to the release workflow, and treating their
-  # absence as "packaging unverified" is what left macOS and Linux resting on a
-  # cross-compile while only Windows had ever produced a package.
-  package:
-    name: package (${{ matrix.os }})
-    runs-on: ${{ matrix.os }}
-    strategy:
-      fail-fast: false
-      matrix:
-        include:
-          - os: macos-latest
-            arch: arm64
-            platform: darwin
-          - os: ubuntu-22.04
-            arch: amd64
-            platform: linux
-          - os: windows-latest
-            arch: amd64
-            platform: windows
-    defaults:
-      run:
-        shell: bash
-    steps:
-      - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7
-
-      - uses: actions/setup-go@b7ad1dad31e06c5925ef5d2fc7ad053ef454303e # v7
-        with:
-          go-version-file: desktop/go.mod
-          cache: true
-          cache-dependency-path: desktop/go.sum
-
-      - uses: pnpm/action-setup@ea17c68df8912ef543352723c149a84f56e3d413 # v6.1.0
-        with:
-          version: 10
-          run_install: false
-
-      - uses: actions/setup-node@820762786026740c76f36085b0efc47a31fe5020 # v7
-        with:
-          node-version: "24"
-          cache: pnpm
-          cache-dependency-path: desktop/electron/pnpm-lock.yaml
-
-      # electron-builder shells out to dpkg-deb through fakeroot for the .deb.
-      - name: Install Linux packaging deps
-        if: runner.os == 'Linux'
-        run: sudo apt-get update && sudo apt-get install -y fakeroot
-
-      - name: Build the SPA
-        working-directory: desktop/frontend-next
-        run: pnpm install --frozen-lockfile && pnpm build
-
-      - name: Build the kernel the shell speaks to
-        working-directory: desktop/electron
-        run: pnpm install --frozen-lockfile && pnpm run build:host
-
-      # The .deb's absolute-path payload. A no-op off Linux, so it is not
-      # guarded here: the one place it must not be forgotten is the one place
-      # it does anything.
-      - name: Stage the Linux package payload
-        working-directory: desktop/electron
-        run: pnpm run build:linux-payload
-
-      # No identity on a runner, and none wanted here: auto-discovery would make
-      # this job fail for the one reason it is not testing.
-      - name: Package
-        working-directory: desktop/electron
-        env:
-          CSC_IDENTITY_AUTO_DISCOVERY: "false"
-        run: pnpm run build:dist
-
-      # The release vocabulary is GOARCH. electron-builder's own is not, and the
-      # rename that reconciles them runs here rather than in the manifest.
-      - name: Artifacts carry the release architecture
-        run: |
-          ls -la dist
-          if ls dist | grep -q -- '-x64'; then
-            echo "::error::an artifact still names x64; studio-manifest reads names as GOARCH"
-            exit 1
-          fi
-          if ! ls dist | grep -q -- "-${{ matrix.platform }}-${{ matrix.arch }}"; then
-            echo "::error::no artifact named for ${{ matrix.platform }}-${{ matrix.arch }}"
-            exit 1
-          fi
-
-      # The update channel is the signed manifest. A second answer to which
-      # version is current is not a convenience. The .blockmap is the same
-      # channel's differential index and carries the release prefix, so it also
-      # inflates the count the release uses to refuse a partial publish.
-      - name: No electron-updater metadata
-        run: |
-          if ls dist | grep -qiE '^(latest|beta|alpha).*\.yml$'; then
-            echo "::error::electron-updater metadata was published beside the manifest channel"
-            exit 1
-          fi
-          if ls dist | grep -q '\.blockmap$'; then
-            echo "::error::a .blockmap ships under the release prefix; set differentialPackage: false"
-            exit 1
-          fi
-
-      # A binary inside app.asar cannot be spawned and the kernel serves the SPA
-      # off disk, so both have to be resources. Found rather than asserted at a
-      # path, because the unpacked directory is named by the builder.
-      - name: The kernel and the page are resources, not archive members
-        run: |
-          host=$(find dist -type f -path '*/bin/reasonix-studio-host*' ! -path '*app.asar*' | head -1)
-          page=$(find dist -type f -path '*/frontend-next/dist/index.html' ! -path '*app.asar*' | head -1)
-          echo "kernel: ${host:-<missing>}"
-          echo "page:   ${page:-<missing>}"
-          test -n "$host" -a -n "$page"
-
-      # Read off the package rather than off the config that describes it. Both
-      # of these went wrong the same way once: electron-builder names a package
-      # after package.json unless told, and fpm records the owner it finds on
-      # disk. Either one leaves a Studio that cannot update itself on Linux and
-      # says nothing about it until a user asks.
-      - name: The .deb is Studio's own package and carries what updates it
-        if: runner.os == 'Linux'
-        run: |
-          deb=$(ls dist/ReasonixStudio-linux-*.deb)
-          name=$(dpkg-deb -f "$deb" Package)
-          if [ "$name" != "reasonix-studio" ]; then
-            echo "::error::the package is $name; dpkg holds that beside the installed reasonix-studio instead of upgrading it, and update.StudioLine asks for it by name"
-            exit 1
-          fi
-          dpkg-deb -c "$deb" > contents.txt
-          # The path is the polkit action's exec.path annotation, so it is the
-          # only one authorization is granted for; root because polkit will not
-          # run a helper the invoking user could have written. Either spelling
-          # of root: dpkg-deb prints the numeric ids it cannot map, and which
-          # one it prints is a property of the machine reading the package
-          # rather than of the package.
-          if ! grep -qE '^-rwxr-xr-x (root/root|0/0) +[0-9]+ .* \./usr/lib/reasonix-studio/reasonix-studio-update-helper$' contents.txt; then
-            echo "::error::no root-owned executable update helper at the path the polkit action names"
-            grep reasonix-studio-update-helper contents.txt || echo "(not in the package at all)"
-            exit 1
-          fi
-          if ! grep -q './usr/share/polkit-1/actions/io.reasonix.studio.update.policy$' contents.txt; then
-            echo "::error::the polkit action is not in the package, so pkexec has nothing to authorize against"
-            exit 1
-          fi
-
-      # The Electron shell sets the package's glibc floor; a Go binary linked
-      # with cgo would raise it to the runner's own.
-      - name: The .deb's Go binaries are statically linked
-        if: runner.os == 'Linux'
-        run: |
-          deb=$(ls dist/ReasonixStudio-linux-*.deb)
-          dpkg-deb -x "$deb" "$RUNNER_TEMP/deb"
-          host=$(find "$RUNNER_TEMP/deb" -type f -path '*/resources/bin/reasonix-studio-host')
-          test -n "$host"
-          node desktop/electron/packaging/elf.js "$host" "$RUNNER_TEMP/deb/usr/lib/reasonix-studio/reasonix-studio-update-helper"
-
-      # Recorded, not asserted: the real bundle layout, so later work argues
-      # from what electron-builder produced rather than from its documentation.
-      - name: Record the macOS bundle layout
-        if: runner.os == 'macOS'
-        run: |
-          app=$(find dist -maxdepth 2 -name '*.app' | head -1)
-          echo "app: $app"
-          ls "$app/Contents"
-          ls "$app/Contents/Resources" | head -20
-          ls "$app/Contents/Frameworks" | head -20
-
-      # The end of the contract: not that the name parses, but that the manifest
-      # keys the artifact under the platform this runner actually is. The tool
-      # writes latest.json into the directory and logs to stdout, so the check
-      # reads the file -- captured output would match its own progress line.
-      - name: studio-manifest resolves the platform this runner is
-        working-directory: desktop
-        env:
-          GITHUB_REPOSITORY: ${{ github.repository }}
-        run: |
-          go run ./cmd/studio-manifest ../dist v0.0.0-ci v0.0.0-ci
-          cat ../dist/latest.json
-          if ! grep -q '"${{ matrix.platform }}-${{ matrix.arch }}"' ../dist/latest.json; then
-            echo "::error::the manifest has no entry for ${{ matrix.platform }}-${{ matrix.arch }}"
-            exit 1
-          fi
@@ -13,8 +13,8 @@ jobs:
  update:
    runs-on: ubuntu-latest
    steps:
-      - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7
-      - uses: actions/setup-node@820762786026740c76f36085b0efc47a31fe5020 # v7
+      - uses: actions/checkout@v7
+      - uses: actions/setup-node@v7
        with:
          node-version: '22'
      - name: Update README contributor tables
@@ -22,10 +22,9 @@ jobs:
          GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
        run: node scripts/update-acknowledgments.mjs
      - name: Open pull request
-        uses: peter-evans/create-pull-request@5f6978faf089d4d20b00c7766989d076bb2fc7f1 # v8
+        uses: peter-evans/create-pull-request@v8
        with:
-          # GITHUB_TOKEN-authored pull requests never trigger CI, so required checks would never report.
-          token: ${{ secrets.ACKNOWLEDGMENTS_TOKEN || secrets.GITHUB_TOKEN }}
+          token: ${{ secrets.GITHUB_TOKEN }}
          branch: chore/update-acknowledgments
          delete-branch: true
          title: 'Update acknowledgments contributors / 更新致谢贡献者'
@@ -24,10 +24,10 @@ jobs:
    if: github.repository == 'esengine/DeepSeek-Reasonix'
    runs-on: ubuntu-latest
    steps:
-      - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7
+      - uses: actions/checkout@v7
        with:
          fetch-depth: 0
-      - uses: actions/setup-node@820762786026740c76f36085b0efc47a31fe5020 # v7
+      - uses: actions/setup-node@v7
        with:
          node-version: '22'
      - name: Test chart generator
@@ -1,29 +0,0 @@
-name: 1.x nightly monitoring
-
-# A schedule runs only from the default branch. These monitors measure the 1.x
-# desktop app, which lives on main-v2 alone, so their nightly runs are started
-# from here on that branch. Their push and pull_request gates run on main-v2 as
-# before.
-on:
-  schedule:
-    - cron: "17 3 * * *"
-    - cron: "43 4 * * *"
-  workflow_dispatch:
-
-permissions:
-  actions: write
-
-jobs:
-  dispatch:
-    runs-on: ubuntu-latest
-    steps:
-      - name: App memory screening
-        if: github.event_name == 'workflow_dispatch' || github.event.schedule == '17 3 * * *'
-        env:
-          GH_TOKEN: ${{ github.token }}
-        run: gh workflow run app-memory.yml --repo "$GITHUB_REPOSITORY" --ref main-v2
-      - name: Diagnostic overhead monitoring
-        if: github.event_name == 'workflow_dispatch' || github.event.schedule == '43 4 * * *'
-        env:
-          GH_TOKEN: ${{ github.token }}
-        run: gh workflow run diagnostic-overhead.yml --repo "$GITHUB_REPOSITORY" --ref main-v2
@@ -1,6 +1,4 @@
-# Build artifacts. A `go build` of one command writes its binary into the
-# working directory, so every main package in the root module can drop its own
-# name here; repolint's build-artifact rule holds the two lists together.
+# Build artifacts
 /bin/
 /dist/
 /stage/
@@ -9,22 +7,11 @@
 /reasonix.exe
 /e2ebench
 /e2ebench.exe
-/context-retrieval
-/compaction
-/context-maintenance-e2e
-/runtime-resume
-/extension-protocol-gen
-/reasonix-plugin-example
-/reasonix-studio-host
-/signpath-contract
-/studio-delta
-/covergate
-/pricecheck
-/repolint
-/workgroupassay
-/write_heartbeat
-/catalog-detector
-/tool-surface-cache
+
+# Website (Astro) build output
+/site/dist/
+/site/.astro/
+/site/_preview.png

 # Go test/build scratch
 *.test
@@ -76,6 +63,7 @@ desktop/**/.reasonix/
 Thumbs.db
 .idea/
 .vscode/
+/site/public/av/

 benchmarks/context-maintenance-e2e/run/

@@ -85,12 +73,7 @@ temp/
 # Official theme art intermediate outputs / caches
 scripts/official-theme-art/out/
 scripts/official-theme-art/__pycache__/
+site/.generated/

 # Committed pointer at REASONIX.md; overrides a global CLAUDE.md ignore rule.
 !/CLAUDE.md
-
-# Go build output left in the tree by `go build ./tools/...`. The ~ form is
-# what a linker leaves behind when it cannot replace a binary that is still
-# running, which on Windows is every rebuild while Studio is open.
-*.exe
-*.exe~
@@ -16,7 +16,7 @@ builds:
    ldflags:
      # Inject version identity for `reasonix version --verbose/--json` (Integration D/E).
      # buildTimeUTC is the goreleaser run time (UTC), not vcs.time.
-      - -s -w -X main.version={{ .Tag }} -X main.gitCommit={{ .ShortCommit }} -X main.buildTimeUTC={{ .Date }} -X reasonix/internal/tools/productdocs.linkedVersion={{ .Tag }} -X reasonix/internal/tools/productdocs.linkedRevision={{ .Commit }}
+      - -s -w -X main.version={{ .Tag }} -X main.gitCommit={{ .ShortCommit }} -X main.buildTimeUTC={{ .Date }} -X reasonix/internal/productdocs.linkedVersion={{ .Tag }} -X reasonix/internal/productdocs.linkedRevision={{ .Commit }}
    goos: [darwin, linux, windows]
    goarch: [amd64, arm64]

@@ -0,0 +1,13 @@
+<?xml version="1.0" encoding="utf-8"?>
+<!-- Canary-only configuration. SignPath test certificates intentionally do not
+     chain to a Windows trusted root, so authenticode-verify cannot be used for
+     the already test-signed payload. The workflow verifies that every payload
+     executable is signed and hash-identical to the rebuilt portable package.
+     Stable and RC releases use windows-installer-v2 instead. -->
+<artifact-configuration xmlns="http://signpath.io/artifact-configuration/v1">
+  <zip-file>
+    <pe-file path="*installer*.exe">
+      <authenticode-sign />
+    </pe-file>
+  </zip-file>
+</artifact-configuration>
@@ -0,0 +1,30 @@
+<?xml version="1.0" encoding="utf-8"?>
+<!-- Sign the finished NSIS container only when the exact payload used to build it
+     is present and already Authenticode-signed. GitHub Actions wraps this flat
+     bundle in a zip before SignPath receives it. Keep this versioned separately
+     so provisioning it cannot break releases still using windows-installer. -->
+<artifact-configuration xmlns="http://signpath.io/artifact-configuration/v1">
+  <zip-file>
+    <pe-file path="*installer*.exe">
+      <authenticode-sign />
+    </pe-file>
+    <pe-file path="reasonix-desktop.exe">
+      <authenticode-verify />
+    </pe-file>
+    <pe-file path="reasonix-guard.exe">
+      <authenticode-verify />
+    </pe-file>
+    <pe-file path="reasonix-launcher.exe">
+      <authenticode-verify />
+    </pe-file>
+    <pe-file path="reasonix-update-helper.exe">
+      <authenticode-verify />
+    </pe-file>
+    <pe-file path="reasonix-cli.exe">
+      <authenticode-verify />
+    </pe-file>
+    <pe-file path="reasonix-uninstall.exe">
+      <authenticode-verify />
+    </pe-file>
+  </zip-file>
+</artifact-configuration>
@@ -0,0 +1,13 @@
+<?xml version="1.0" encoding="utf-8"?>
+<!-- GitHub Actions wraps every uploaded artifact in a zip, so the installer arrives
+     zipped even though it is a single file — match and sign the .exe inside. -->
+<artifact-configuration xmlns="http://signpath.io/artifact-configuration/v1">
+  <zip-file>
+    <pe-file-set>
+      <include path="*installer*.exe" min-matches="1" max-matches="1" />
+      <for-each>
+        <authenticode-sign />
+      </for-each>
+    </pe-file-set>
+  </zip-file>
+</artifact-configuration>
@@ -0,0 +1,25 @@
+<?xml version="1.0" encoding="utf-8"?>
+<!-- Sign every executable that the NSIS installer writes to disk. GitHub Actions
+     wraps the flat payload directory in this zip before SignPath receives it. -->
+<artifact-configuration xmlns="http://signpath.io/artifact-configuration/v1">
+  <zip-file>
+    <pe-file path="reasonix-desktop.exe">
+      <authenticode-sign />
+    </pe-file>
+    <pe-file path="reasonix-guard.exe">
+      <authenticode-sign />
+    </pe-file>
+    <pe-file path="reasonix-launcher.exe">
+      <authenticode-sign />
+    </pe-file>
+    <pe-file path="reasonix-update-helper.exe">
+      <authenticode-sign />
+    </pe-file>
+    <pe-file path="reasonix-cli.exe">
+      <authenticode-sign />
+    </pe-file>
+    <pe-file path="reasonix-uninstall.exe">
+      <authenticode-sign />
+    </pe-file>
+  </zip-file>
+</artifact-configuration>
@@ -3,17 +3,17 @@ project_slug: DeepSeek-Reasonix
 signing_policy_slug: release-signing
 repository_url: https://github.com/esengine/DeepSeek-Reasonix.git
 allowed_branch_names:
-  - studio
+  - main-v2
 allowed_build_definitions:
-  - .github/workflows/release-studio.yml
-  - .github/workflows/studio-certum-signing-smoke.yml
+  - .github/workflows/release-stable.yml
+  - .github/workflows/release-desktop.yml
 fingerprint_files:
-  - .github/workflows/release-studio.yml
-  - .github/workflows/studio-certum-signing-smoke.yml
-  - .github/actions/setup-certum/action.yml
-  - scripts/sign-certum.ps1
+  - .github/workflows/release-stable.yml
+  - .github/workflows/release-stable-trigger.yml
+  - .github/workflows/release-desktop.yml
+  - .signpath/artifact-configurations/windows-payload.xml
+  - .signpath/artifact-configurations/windows-installer-v2.xml
+  - scripts/complete-signpath-request.ps1
+  - scripts/package-windows-desktop.sh
  - scripts/verify-windows-authenticode.ps1
-  - scripts/verify-windows-package.ps1
-  - scripts/windows-signing-lib.ps1
-  - desktop/electron/electron-builder.yml
  - cmd/signpath-contract/main.go
@@ -0,0 +1 @@
+v2.13.0
@@ -1,6 +0,0 @@
-# Reasonix
-
-The project's standing instructions live in `REASONIX.md` — one file, read by
-both Reasonix and Codex. Add rules there, never here.
-
-@REASONIX.md
@@ -0,0 +1,362 @@
+# Changelog
+
+All notable changes to the Go line (Reasonix 1.0+) are recorded here. The legacy
+`0.x` TypeScript history lives on the [`v1`](https://github.com/esengine/DeepSeek-Reasonix/tree/v1)
+branch.
+
+## Unreleased
+
+### Fixed
+
+- **v1.24.2 session snapshot & recovery root fix:** Keep PR #7982's WAL/CAS/lease
+  safety foundation, but replace process-level "I hold a lease" ownership with a
+  generation-bound `SessionWriteAuthority`. Same-revision tool-preview/load
+  reshapes no longer false-diverge; recovery files are bounded to one path per
+  writer/lineage; empty checkpoints heal from their own WAL; projection lineage
+  rebinds across upgrade/model switch and inherits across recovery forks without
+  changing provider-visible prompt bytes. Catalog upgrades to disposable
+  `session-catalog/v3.sqlite` with recovery lineage roles
+  (`normal|covered_copy|adopted|diverged`); covered idle copies move to the
+  recoverable `.trash` using a 15-minute idle threshold applied on two early
+  sweeps (at startup and ~20 minutes later), then a 24-hour threshold on the
+  6-hour background ticker; independent diverged branches stay and are listed
+  for user choice. v1/v2 catalogs are
+  left byte-unchanged for coexistence/downgrade.
+  **v1.24.1** only hid/reclaimed already-created covered copies and fixed Windows
+  flash-window startup; **v1.24.2** stops the misclassification source and repairs
+  existing user directories without rewriting authoritative JSONL/WAL/sidecar data.
+
+- Goal now runs continuously by default: the former 16-round per-Run boundary,
+  10/20/40 cross-Run quotas, default wall-clock budget, and numeric
+  no-progress/Todo-stall pauses no longer stop valid work. Progress guards still
+  detect repeated host outcomes and zero-evidence work, but redirect the model
+  to re-plan instead of producing `goal_run_budget` or `goal_stuck`. Explicit
+  `[agent].goal_token_budget`, `--max-steps`, positive time/cost budgets, manual
+  pause/stop, genuine user/external blockers, and evaluator fail-closed behavior
+  remain available. The Goal token budget defaults to `0` (off); resuming its
+  `budget_spend` pause grants a fresh slice without clearing cumulative usage.
+  Goal status reports turns, provider requests, tokens, the optional configured
+  token threshold, and cumulative active work time. Bot `max_steps` also
+  defaults to `0` (continuous), while positive user configuration is enforced.
+
+- Removed numeric Goal pauses in existing sidecars automatically normalize to
+  `running` without sending a model request. Active Goal sidecars write
+  `turnsLimit: -1` as a downgrade-safe unlimited sentinel while public runtime
+  APIs retain deprecated limit fields as `0`. The migration preserves unknown
+  fields, todos, checkpoints, usage, evidence, and historical metadata.
+
+- Goal is now the sole long-task runtime. Historical AutoResearch sidecars
+  migrate transactionally into Goals with research compatibility metadata. Invalid archives block
+  fail closed and remain read-only, retaining the task id and compatibility mode
+  for a restart or `/goal resume` retry; successful Goal-only sidecars omit the
+  old task id and write an explicit downgrade fence so previous readers cannot
+  reactivate the removed runtime.
+
+- Context-dependent workflow tools now share one host-side execution boundary.
+  Goal, Plan sign-off, and background-job calls cannot reach permissions,
+  hooks, leases, or Execute outside their owning context; mixed batches execute
+  valid calls once and stop safely after one repair. Child agents also isolate
+  inherited Goal, Jobs, and live memory queues, while persisted tool identity
+  records the effective child schema projection.
+
+- **Issue #7575:** Linux Bash under bubblewrap no longer mounts a fresh empty
+  `--tmpfs /tmp` on every call. Consecutive commands in the same logical session
+  now share a private temporary directory (bound at `/tmp` on Linux, exported via
+  `TMPDIR`/`TMP`/`TEMP` on all platforms) without exposing the host public
+  temporary root. `/new`, `/clear`, resume of another session, and branch
+  switches rotate the directory; model/settings hot rebuilds keep it. Sub-agent
+  runs get independent directories. Temporary files are not durable across process
+  restarts.
+
+### Added
+
+- Added `[ui].show_turn_usage` so CLI/TUI users can hide per-request token and
+  cost receipts from transcript scrollback without disabling usage accounting.
+
+## [1.20.0] — 2026-08-05
+
+Extension kernel, Task Monitor, and safer Goal completion.
+
+Compact decision surfaces, local decision receipts, unified extension kernel,
+native Task Monitor, bounded sub-agent progress, Goal fail-closed completion,
+MiMo and DashScope Responses fixes, SSH remote access simplification, and
+multiple Desktop stability improvements.
+
+### Highlights
+
+- **Unified Extension Kernel and Extension Protocol v1**: Immutable runtime
+  snapshots, fail-atomic reload, Plugin Manifest v1 (prompts, themes, full-trust
+  code runtimes), stable JSON-RPC sidecar protocol, interceptor dispatch,
+  streaming provider adapter, structured UI, and Go SDK.
+- **Native Task Monitor**: Monitor agent tasks natively in CLI and Desktop with
+  lifecycle semantics and session-scoped summary view.
+- **Bounded Sub-agent Progress Forwarding**: Forward structured progress for
+  `task`, `parallel_tasks`, and `fleet` without flooding the parent stream.
+  Renders nested lifecycle cards in Desktop and stable per-child transcript
+  slots in CLI.
+- **Goal Completion Fail-Closed**: Replace free-form Goal footer markers with a
+  stable `update_goal` tool and epoch-scoped per-turn reports. Centralized
+  completion logic with bounded evaluator, progress-aware budgets, and
+  pause/resume controls.
+- **Ablation Subsystem Switches**: Switch subsystems off behind one shared
+  vocabulary for controlled experiments. Includes planner, subagent, retrieval,
+  evidence, and compaction.
+- **Benchmark Cost per Solved Task**: Report cost per solved task, tokens per
+  solved, median wall time, and failure-class breakdown in e2e reports.
+- **Compact Decision Surfaces and Local Receipts**: Compact footer decision-card
+  layout with bounded scroll, dense action rows, and overflow disclosure.
+  Record bounded Ask, approval, and recovery decisions as local transcript
+  receipts.
+- **Simplified SSH Remote Access**: Remove Remote Workbench protocol and
+  stacks; reuse CLI/Serve remote model. Desktop opens per-host native web child
+  windows via SSH. Keyless remote Serve setup with loopback-only page.
+- **Model Usage Charts with Primer Palette**: Replace monochrome accent ramp
+  with GitHub Primer data-viz two-set categorical palette. Fix donut overflow
+  on hover and keyboard accessibility.
+- **Cross-platform Extension and Task Monitor Reliability**: Make
+  content-reference eviction deterministic, reject Unix and Windows absolute
+  plugin paths consistently, stabilize parallel-task cancellation, and restore
+  reliable Windows validation for Task Monitor and remote provider setup.
+- **MiMo and DashScope Responses Wire Alignment**: Fix multi-turn tool loops,
+  reasoning round-trip, JSON output for MiMo; fix DashScope second-turn 400
+  error, all-zero usage suppression, and vendor-aware cache TTL.
+- **Desktop Stability Fixes**: Recover stuck updates and legacy WebKit, contain
+  macOS alias repair startup crashes, keep composer overflow stacks readable,
+  and harden account verification and community flows.
+- **Remote Web Recovery After SSH Drops**: Add integration regression test for
+  SSH drop, forward recovery, and window reload. Document transient outage
+  behavior.
+- **CI: Auto-minimize Activity-Farming Spam Comments**: Detect and minimize
+  template spam comments from non-contributor accounts based on structural
+  signals.
+
+### Added
+
+- Added Extension Protocol v1 and the unified extension kernel: installed or
+  linked sidecars can contribute tools, skills, commands, hooks, MCP servers,
+  providers, interceptors, and structured UI surfaces through a versioned
+  NDJSON contract and the public Go SDK. CLI, Desktop, ACP, and Serve support
+  fail-atomic runtime reloads; Serve also renders extension surfaces and lists
+  extension-hosted providers without exposing credentials.
+- Added the structured Goal completion protocol: the always-registered
+  `update_goal` tool (continue/complete/blocked with reason and next_action)
+  replaces the `[goal:*]` footer markers. The Goal FSM is now the exclusive
+  cross-turn decision point and validates every complete claim against Delivery
+  readiness; when the model submits no report, an independent bounded evaluator
+  (recovery_model → guardian_model → main model, no tools/history, usage
+  attributed to `goal-evaluator`) judges the turn once, and any evaluator
+  failure pauses the goal instead of continuing silently.
+- Added Goal budget classes with safe pauses: simple 10 turns / 200k tokens,
+  write 20 turns / 400k tokens, AutoResearch 40 turns / 800k tokens, and a
+  4-turn no-host-verifiable-progress gate. Pauses keep all Goal state; `/goal
+  resume` continues and adds one slice of the current class when the pause was
+  budget-related. `/goal status` shows the full turn/token/no-progress runtime,
+  and `/goal pause` manually suspends a running Goal.
+- Added the `goalRuntime` nested view to the desktop Meta, the remote protocol
+  (`session/goal/pause` operation, `goalRuntime` DTO on session meta), and the
+  ACP status payload; the desktop Composer goal menu shows the runtime summary
+  with distinct pause/end/resume actions.
+
+### Changed
+
+- Delivery no longer retries final-answer readiness with hidden model messages:
+  a plain Delivery run ends on the first unsatisfied final answer and surfaces
+  the recovery card, while a Goal + Delivery run has the Goal FSM absorb the
+  failure and continue under budget with the missing requirements as the next
+  turn's prompt. Historical `[goal:*]` footers are stripped from old transcripts
+  for display only and never participate in state decisions.
+- Added a **Remote SSH** module (VS Code Remote-SSH style): a user-global
+  `[remote]` host config, `reasonix remote` CLI (add/list/remove/import/test/
+  connect/status/forward/serve/fs) and `/remote` slash command, an SSH transport
+  with trust-on-first-use host-key verification, keepalive + exponential-backoff
+  reconnect, `-L`/`-R` port forwarding, and SFTP file access. `connect`
+  bootstraps a persistent `reasonix serve` on the remote host and tunnels its
+  loopback port so the full agent runs remotely. The desktop app adds a
+  **Settings -> Remote SSH** host manager, a remote file browser/editor, a
+  port-forwarding panel, and a status-bar connection chip. Linux/macOS remotes.
+- Added `reasonix serve --port-file/--token-file/--pid-file` so a supervised
+  headless serve can bind an ephemeral port and read its auth token from a file
+  (keeping it out of `ps`).
+- Added an authenticated, loopback-only Provider setup page for `reasonix
+  serve`. A Serve whose selected Provider is missing its API key now remains
+  reachable, stores the submitted key in that host's Reasonix credential file,
+  and rebuilds the active controller in place without restarting Serve.
+- Added Claude Code-style searchable CLI pickers for models, providers, and
+  sessions, with arrow, Vim, and `Ctrl+P` / `Ctrl+N` navigation.
+- Added `-p` / `--print`, `text`, `json`, and `stream-json` output modes for
+  one-shot use and automation.
+- Added session-scoped `--allowed-tools`, repeatable `--add-dir`, Claude-compatible
+  permission modes, flexible `--resume [QUERY]`, and the `--copy` resume escape
+  hatch.
+- Added `/status` details for the active model, effort, cache, Git state,
+  background jobs, work profile, and provider balance where available.
+- Remote SSH workspaces now open as a standalone remote web window again.
+  Opening a workspace from the status bar or the Remote Server tab starts or
+  reuses the remote `reasonix serve`, tunnels its loopback port, and opens the
+  Serve web client in a dedicated per-host window. The remote web page uses
+  the provider configuration and API keys on the **remote** host; the desktop
+  no longer exposes its local providers to remote hosts. If the selected remote
+  Provider is missing its API key, the window opens a setup page that saves the
+  key only on that host and then opens the normal Serve UI. The Remote Workbench
+  protocol, its Provider Broker, and the same-window remote projection were
+  removed. Legacy mirror and provider-trust files are not deleted
+  automatically; Settings -> Remote SSH shows a cleanup card when they exist.
+  The hidden `remote attach-workspace`, `remote runtime-workbench`, and
+  `remote workbench-build-id` commands now fail with a pointer to
+  `reasonix remote connect <host> --open`.
+- Automatic Plan Mode has been retired. Plan Mode is now always entered through
+  an explicit user choice, and the one-time config v5 upgrade removes legacy
+  `agent.auto_plan` and `agent.auto_plan_classifier` values so upgraded users
+  receive the same behavior as new users.
+- `Shift+Tab` now cycles CLI safe modes from Ask to Auto to Plan, while YOLO
+  remains an independent `Ctrl+Y` toggle.
+- Model, provider, resume, and approval menus now use consistent row selection;
+  slash completion, help, aliases, and dispatch share one command registry.
+- The full-screen CLI composer now uses theme-accented borders and a slim bar
+  cursor by default, grows within the available terminal height, scrolls long
+  drafts independently, and preserves selections across explicit image paste.
+- The persistent CLI footer now uses a responsive, theme-aware layout for
+  interaction state, model, effort, localized work mode, Git identity, cache,
+  context, compaction headroom, jobs, and balance. Narrow terminals move or
+  compact complete groups instead of clipping labels.
+- CLI clipboard actions now separate terminal-native text paste from explicit
+  image paste: `Ctrl+V` on macOS/Linux, `Alt+V` on Windows, or `/paste-image`.
+  Local transcript copy verifies the native clipboard write, while SSH uses a
+  clearly labelled OSC 52 fallback.
+- Runtime rebuilds after model, effort, or work-mode changes now preserve the
+  conversation, session permission overrides, additional directories, and
+  session lease ownership.
+- Agent execution now monitors host-observed Todo progress automatically. A
+  stalled current item receives a recovery nudge after 8 tool-call rounds with
+  no new completion, unique read, command, or mutation, and pauses with saved
+  work after 16. Exact repeats do not renew the progress lease; real work does.
+  Two-level task lists keep the single in_progress contract: the active
+  sub-step is the only current item while its phase stays pending, and the
+  phase becomes in_progress to sign off only after all of its sub-steps are
+  completed. A level-1 sub-step with no phase header above it is rejected.
+  Executor and planner rounds now use automatic progress management. Retired
+  `[agent].max_steps` and `planner_max_steps` keys remain parseable for upgrades,
+  but are ignored and removed by a one-time migration so stale hidden limits
+  cannot truncate new behavior. One-off CLI and unattended bot limits remain.
+
+### Fixed
+
+- Fixed long parallel sub-agent research being silently lost when combined
+  `parallel_tasks` or `fleet` answers exceeded the single-tool output limit.
+  Persisted sessions now keep each child transcript independently, return a
+  bounded fair preview plus stable reference for every result, and page full
+  answers through the conversation-scoped `read_subagent_result` tool.
+- Fixed Remote Workbench failing with only `initialize: workbench-desktop:
+  connection closed` on fresh or cross-platform SSH hosts. Desktop now proves
+  the exact Host CLI Build ID, provisions the matching verified release without
+  requiring remote npm, runs the managed binary explicitly, and preserves a
+  safe structured bootstrap error when the remote command exits early.
+- Hardened Bash permission reuse for dynamic and indirect execution. Parameter/arithmetic expansions,
+  assignments, redirects, heredocs, and globs can only be remembered as exact
+  `Bash=<literal>` rules, while still using Auto's normal fallback. Nested or
+  indirect execution now requires a human in interactive Ask/Auto and fails
+  closed in headless Ask/Auto/DontAsk. Broad Bash rules, Guardian/hook allows,
+  and the approved-plan window can no longer silently authorize that stricter
+  class; YOLO remains the explicit full-access bypass and sandbox enforcement
+  is unchanged.
+- Fixed Desktop sessions incorrectly locking themselves during Goal + Delivery
+  mode changes, controller rebuilds, duplicate-tab restore, and background
+  reattachment. Desktop now keeps one process-local runtime owner per canonical
+  session, fences stale controller events by runtime epoch, blocks sends until
+  that runtime is ready, and scopes single-instance ownership to
+  `REASONIX_HOME` instead of the executable path. Switching saved sessions is
+  now transactional: a target build, restore, or lease failure leaves the
+  current controller, lease, path, mode profile, and runtime epoch untouched.
+- Stabilized the desktop rich composer caret after skill and plugin invocation
+  tags. DOM→model and model→DOM selection mapping now treat invocation chips as
+  zero-length atoms while still counting user text that lands inside the NBSP
+  caret anchor (common on Windows WebView2), restore both selection ends, and
+  recover the insertion point from a `beforeinput` snapshot when the browser
+  temporarily loses selection — so mid-text edits no longer jump to the end.
+- Isolated the Windows desktop WebView2 shell from stale system proxies, so an
+  exited proxy client cannot leave the embedded UI hidden during startup. If
+  WebView2 still does not reach DOM-ready within 15 seconds, Reasonix now shows
+  the native window with a recovery prompt instead of appearing not to launch.
+  Remote Markdown images are fetched by the backend with Reasonix's proxy
+  configuration instead of bypassing that proxy through the isolated WebView.
+- Restored captured-mouse right-click text paste, made composer drag selection
+  copy through the verified native clipboard path, and kept non-Git footer
+  telemetry left-aligned without reserving an empty data band.
+- Restored stateful MCP behavior after the v1.17.13 regression: user-added
+  servers work without extra trust settings (including delivery-mode on-demand
+  calls), repository-provided servers use one exact launch confirmation, and
+  stdio tools reuse one persistent process so browser sessions survive across
+  calls without repeated startup latency. The former trust/reverify/catalog
+  management UI and CLI are removed.
+- Localized persistent-footer labels and displayed work-mode values in English,
+  Simplified Chinese, and Traditional Chinese, while keeping command arguments
+  stable.
+- Restored the `0.53` content boundary: model output, tool output, session
+  transcripts, recovery branches, and background-job artifacts retain their
+  original text instead of being rewritten by heuristic secret redaction.
+  Credential masking remains in key-entry summaries and explicit diagnostic or
+  session-cleanup paths. Transcript-bearing session/job sidecars are kept
+  private (`0600`, with private job directories), and the retired
+  `redact_tool_output` setting is removed with a one-time upgrade notice.
+
+### Notes
+
+- Full bilingual release notes:
+  <https://reasonix.io/changelog/v1.20.0/> ·
+  [GitHub release](https://github.com/esengine/DeepSeek-Reasonix/releases/tag/desktop-v1.20.0).
+- The detailed entries below accumulated on `main-v2` after 1.0.0 and shipped
+  across 1.1.0–1.20.0; per-version attribution lives in the per-version release
+  notes linked above.
+
+## 1.1.0 – 1.19.7
+
+Per-version entries for the intermediate releases are published in the
+[bilingual release notes](https://reasonix.io/changelog/) and on the
+[GitHub releases page](https://github.com/esengine/DeepSeek-Reasonix/releases).
+
+## [1.0.0] — 2026-06-03
+
+First stable release — a **ground-up rewrite in Go**. Not an upgrade of the `0.x`
+TypeScript line; a new codebase that becomes the default (`main-v2`).
+
+### Highlights
+
+- **Go kernel**: a single static binary (CGO-free), cross-compiled for
+  darwin/linux/windows on amd64 + arm64. Distributed via npm (the package wraps
+  the native binary), Homebrew (`esengine/reasonix` tap), and release archives;
+  no Node runtime needed to run it.
+- **Agent core**: the loop, built-in tools (read/write/edit/multi_edit/glob/grep/
+  ls/bash/web_fetch/todo_write), permission gate, sandboxed bash, and the
+  DeepSeek prefix-cache–oriented design.
+- **Subagents**: `task` plus explore/research/review/security_review skill agents.
+- **Skills & hooks**: Claude-Code-style skills (`internal/skill`) and hooks
+  (`internal/hook`), symlink-aware and slash-integrated.
+- **MCP client**: connect external servers over stdio / Streamable HTTP; reads
+  `[[plugins]]` and a Claude-Code `.mcp.json`.
+- **Code intelligence via CodeGraph**: a tree-sitter symbol/call graph
+  (`codegraph_*` tools) replaces embedding semantic search — no embedding service
+  or API cost. Fetched into a local cache on first use (or `reasonix codegraph
+  install`) and indexed in the background, so installs and startup stay fast.
+- **Plan mode** with evidence-backed step sign-off (`complete_step`).
+- **Memory**: `REASONIX.md` hierarchy + auto-memory, folded into the cache-stable
+  prefix.
+- **ACP** (`reasonix acp`) and an HTTP/SSE server frontend; desktop app (Wails).
+
+### Fixed
+
+- **File encoding support restored** — GBK/GB18030 (and other non-UTF-8) files
+  can now be read, edited, and grepped correctly. The v2 rewrite had dropped
+  v1's encoding detection; files in CJK Windows charsets were silently misread
+  or rejected as binary. The read/edit/write round-trip now preserves the
+  original file encoding. (#2637)
+
+### Notes
+
+- Versions: the legacy TypeScript line stays in `0.x`; the Go line starts at
+  `1.0.0`. See [docs/MIGRATING.md](docs/MIGRATING.md).
+- Release archives ship a bare binary; CodeGraph is fetched on first use. Windows
+  support for the fetched runtime is unverified — install `codegraph` on PATH if
+  the auto-fetch doesn't resolve there.
+
+[1.20.0]: https://github.com/esengine/DeepSeek-Reasonix/releases/tag/desktop-v1.20.0
+[1.0.0]: https://github.com/esengine/DeepSeek-Reasonix/releases/tag/v1.0.0
@@ -5,10 +5,11 @@ everything you need to get started.

 ## Prerequisites

- **Go 1.26+** — the version `go.mod` requires
+- **Go 1.25+** — the project targets the latest stable Go release
 - **Git** — for version control
- **Node.js 24+ and pnpm 10** (optional) — only if you work on Studio
-  (`desktop/`)
+- **Node.js** (optional) — only if you work on the desktop app (`desktop/`)
+- **Wails CLI** (desktop only) — run `make wails-install` so the CLI matches
+  `desktop/go.mod`

 ## Getting started

@@ -24,22 +25,22 @@ go test ./...              # runs the full test suite
 | Directory | Purpose |
 |-----------|---------|
 | `cmd/reasonix` | CLI entry point |
-| `internal/runtime/agent` | Agent loop, session, coordinator |
-| `internal/frontend/cli` | TUI, subcommands, setup wizard |
-| `internal/session/control` | Transport-agnostic controller |
-| `internal/contract/config` | TOML configuration loading |
-| `internal/tools/builtin` | Built-in tools (bash, read_file, …) |
-| `internal/contract/provider` | Model-backend abstraction |
-| `internal/model/openai` | OpenAI-compatible provider |
-| `internal/ext/plugin` | MCP client (stdio + HTTP) |
-| `internal/contract/event` | Typed event stream |
-| `internal/ext/hook` | Shell hooks (PreToolUse, …) |
-| `internal/state/memory` | REASONIX.md hierarchy + auto-memory |
-| `internal/ext/skill` | Skill discovery from Markdown |
-| `internal/safety/sandbox` | OS-level sandboxing |
-| `internal/frontend/serve` | HTTP/SSE server frontend |
-| `internal/state/checkpoint` | Snapshot-based rewind |
-| `desktop/` | Studio: the Electron shell, the SPA it serves (separate Go module) |
+| `internal/agent` | Agent loop, session, coordinator |
+| `internal/cli` | TUI, subcommands, setup wizard |
+| `internal/control` | Transport-agnostic controller |
+| `internal/config` | TOML configuration loading |
+| `internal/tool/builtin` | Built-in tools (bash, read_file, …) |
+| `internal/provider` | Model-backend abstraction |
+| `internal/provider/openai` | OpenAI-compatible provider |
+| `internal/plugin` | MCP client (stdio + HTTP) |
+| `internal/event` | Typed event stream |
+| `internal/hook` | Shell hooks (PreToolUse, …) |
+| `internal/memory` | REASONIX.md hierarchy + auto-memory |
+| `internal/skill` | Skill discovery from Markdown |
+| `internal/sandbox` | OS-level sandboxing |
+| `internal/serve` | HTTP/SSE server frontend |
+| `internal/checkpoint` | Snapshot-based rewind |
+| `desktop/` | Wails-based desktop app (separate Go module) |
 | `docs/` | Engineering spec, migration guide |

 ### Dependency direction
@@ -79,10 +80,11 @@ REASONIX_HOME=/tmp/reasonix-dev go run ./cmd/reasonix
 #   REASONIX_HOME=/tmp/reasonix-dev ./bin/reasonix
 ```

-**Studio**
+**Desktop**

 ```bash
-REASONIX_HOME=/tmp/reasonix-dev-isolated make studio
+cd desktop && wails build
+REASONIX_HOME=/tmp/reasonix-dev-isolated build/bin/reasonix-desktop
 ```

 On Windows, use `$env:REASONIX_HOME` in PowerShell or `set REASONIX_HOME=` in
@@ -104,17 +106,24 @@ serialization, compaction, or MCP/tool registration need explicit cache review.
 For these changes:

 - Keep system prompt changes low-frequency and require explicit review.
- Say in the PR what the change does to the provider-visible prefix, and why
-  that cost is worth paying.
+- Fill the PR body `Cache-impact:` line with `none`, `low`, `medium`, or `high`
+  plus the reason.
+- Fill the PR body `Cache-guard:` line with the focused guard test/command added
+  or run, or explain why an existing guard covers the change.
+- Fill `System-prompt-review:` when system prompt, memory prefix, output style,
+  or skill index behavior changes.
 - Prefer focused guard tests near the changed surface; `scripts/cache-guard.sh`
  remains the broader release-level cache-hit check.

+CI enforces this metadata for cache-sensitive paths so prompt/tool prefix churn
+is called out before review.
+
 ### Running tests

 ```bash
 go test ./...                           # all tests
-go test ./internal/runtime/agent/ -v            # verbose, one package
-go test ./internal/tools/builtin/ -run TestGrep  # one test
+go test ./internal/agent/ -v            # verbose, one package
+go test ./internal/tool/builtin/ -run TestGrep  # one test
 ```

 ### Code style
@@ -139,52 +148,39 @@ ci: add golangci-lint and govulncheck

 ## Adding a new built-in tool

-1. Create `internal/tools/builtin/mytool.go`
+1. Create `internal/tool/builtin/mytool.go`
 2. Implement the `tool.Tool` interface: `Name()`, `Description()`, `Schema()`, `ReadOnly()`, `Execute()`
 3. Register via `func init() { tool.RegisterBuiltin(myTool{}) }`
-4. Add tests in `internal/tools/builtin/builtin_test.go` or a separate `mytool_test.go`
+4. Add tests in `internal/tool/builtin/builtin_test.go` or a separate `mytool_test.go`
 5. The tool is automatically available — `main` blank-imports `builtin`

 ## Adding a new model provider

-(For MCP tool servers see `internal/ext/plugin` instead — that's a different layer.)
+(For MCP tool servers see `internal/plugin` instead — that's a different layer.)

-1. Create `internal/contract/provider/myprovider/`
+1. Create `internal/provider/myprovider/`
 2. Implement `provider.Provider`: `Name()`, `Stream()`
 3. Register via `func init() { provider.Register("mykind", New) }`
 4. The provider is available from config with `kind = "mykind"`

 ## Adding i18n strings

-1. Add the field to `internal/base/i18n/i18n.go` (`Messages` struct)
-2. Add the value in `internal/base/i18n/messages_en.go` and `messages_zh.go`
+1. Add the field to `internal/i18n/i18n.go` (`Messages` struct)
+2. Add the value in `internal/i18n/messages_en.go` and `messages_zh.go`
 3. The `TestCatalogsComplete` test will fail if you miss a locale

 ## Submitting changes

-Reasonix ships on two lines (see the
-[version roadmap](https://github.com/esengine/DeepSeek-Reasonix/discussions/10748)),
-and each takes different changes:
-
-| Line | Branch | Takes |
-| --- | --- | --- |
-| 2.x | `studio` | Features and fixes: active development |
-| 1.x | `main-v2` | Bug fixes, provider/API compatibility, release/updater, security, platform stability |
-
-A fix that matters on both lines lands on `main-v2` and is reviewed for
-porting to `studio`; do not open the same PR against both.
-
 1. Fork the repository
-2. Create a feature branch from the branch your change targets
+2. Create a feature branch from `main-v2`
 3. Make your changes with tests
 4. Ensure `go test ./...` passes
 5. Ensure `gofmt -l .` shows no changes
-6. Submit a pull request to that same branch
+6. Submit a pull request to `main-v2`

 ## Reporting issues

 Open an issue on GitHub with:
- Which line (1.x or 2.x) and the exact version
 - Steps to reproduce
 - Expected vs actual behavior
 - Go version and OS
@@ -1,4 +1,4 @@
-VERSION := $(shell git describe --tags --always 2>/dev/null | sed 's/^studio-//' || echo dev)
+VERSION := $(shell git describe --tags --always 2>/dev/null || echo dev)
 BUILD_TIME_UTC := $(shell date -u +%Y-%m-%dT%H:%M:%SZ)
 GIT_COMMIT := $(shell git rev-parse --short=12 HEAD 2>/dev/null || echo unknown)
 LDFLAGS := -s -w \
@@ -8,8 +8,9 @@ LDFLAGS := -s -w \
 GOEXE := $(shell go env GOEXE)
 # One pin for the Makefile and the CI lint job; see .github/workflows/ci.yml.
 GOLANGCI_VERSION := $(shell cat .golangci-version)
+WAILS_VERSION := $(shell tr -d '[:space:]' < .wails-version)

-.PHONY: build vet fmt lint lint-go lint-install lint-cross lint-update coverage-gate coverage-gate-update pricecheck check test studio-test sdk-test sdk-test-race hooks cross clean studio
+.PHONY: build vet fmt lint lint-go lint-install lint-cross lint-update wails-install test desktop-test desktop-test-short desktop-test-times sdk-test sdk-test-race hooks cross clean

 build:
 	CGO_ENABLED=0 go build -ldflags "$(LDFLAGS)" -o bin/reasonix$(GOEXE) ./cmd/reasonix
@@ -18,13 +19,6 @@ build:
 vet:
 	go vet ./...

-# Build Studio and launch it. The Electron shell is the one this runs: it is a
-# real GUI app on every platform without being bundled first, so there is one
-# launch path rather than a macOS one and an everywhere-else one.
-studio:
-	bash desktop/electron/run-studio.sh
-
-
 fmt:
 	gofmt -w .

@@ -33,6 +27,8 @@ fmt:
 # particular never surface in `go vet`.
 lint: lint-go
 	go run ./tools/repolint
+	bash scripts/check-wails-pin.sh
+	bash scripts/check-wails-pin.test.sh

 lint-go:
 	@command -v golangci-lint >/dev/null || { echo "golangci-lint not installed; run: make lint-install"; exit 1; }
@@ -45,33 +41,12 @@ lint-go:
 lint-install:
 	CGO_ENABLED=0 go install github.com/golangci/golangci-lint/v2/cmd/golangci-lint@$(GOLANGCI_VERSION)

-# repolint over what the working tree changed. It exists as a target rather than
-# a command to paste because the host recognizes `make check` as a verification
-# it can read the result of, while an arbitrary `go run ./tools/...` is an
-# unknown program it must assume writes — so only this form can be cited as
-# evidence that the standards gate passed.
-check:
-	go run ./tools/repolint -changed
-
 lint-update:
 	go run ./tools/repolint -update

-# The paths the project declared sensitive, held to the coverage they already
-# have. It runs their tests, so it is its own target rather than part of the
-# fast `make lint`. -update records the current numbers; lowering one needs
-# -allow-drop, the way widening a repolint budget does.
-coverage-gate:
-	go run ./tools/covergate
-
-coverage-gate-update:
-	go run ./tools/covergate -update
-
-# The vendors publish their rates on pages this reads and compares the table
-# against. It reaches the network, so it stays out of lint and out of CI: run it
-# before touching a rate and when a vendor announces a change. A page it cannot
-# read exits non-zero — unread never counts as agreement.
-pricecheck:
-	go run ./tools/pricecheck
+wails-install:
+	bash scripts/check-wails-pin.sh
+	go install "github.com/wailsapp/wails/v2/cmd/wails@$(WAILS_VERSION)"

 # Linting one GOOS leaves every //go:build windows and darwin file unchecked.
 lint-cross:
@@ -81,15 +56,17 @@ lint-cross:
 		(cd $$2 && GOOS=$$1 golangci-lint run --timeout=5m ./...) || exit 1; \
 	done

-# Every module, because a nested one is where a red test hides longest: the
-# shell's route allowlist sat broken through a release, and `go test ./...` from
-# the root cannot see it.
-test: studio-test sdk-test
+test:
 	go test ./...

-# The shell is a nested module, so the root `test` target never reaches it.
-studio-test:
-	cd desktop && go test ./...
+desktop-test:
+	cd desktop && go test .
+
+desktop-test-short:
+	cd desktop && go test -short .
+
+desktop-test-times:
+	cd desktop && go test -count=1 -json . | python3 ../scripts/desktop-test-times.py

 sdk-test:
 	cd sdk/go && go test ./...
@@ -21,7 +21,6 @@
 </p>

 <p align="center">
-  <a href="https://github.com/esengine/DeepSeek-Reasonix/releases?q=studio-v&expanded=true"><img src="https://img.shields.io/github/v/release/esengine/DeepSeek-Reasonix.svg?filter=studio-v*&include_prereleases&style=flat-square&color=8250df&labelColor=161b22&label=studio%202.x" alt="Reasonix Studio 2.x"/></a>
  <a href="https://www.npmjs.com/package/reasonix"><img src="https://img.shields.io/npm/v/reasonix.svg?style=flat-square&color=cb3837&labelColor=161b22&logo=npm&logoColor=white" alt="npm version"/></a>
  <a href="https://github.com/esengine/DeepSeek-Reasonix/actions/workflows/ci.yml"><img src="https://img.shields.io/github/actions/workflow/status/esengine/DeepSeek-Reasonix/ci.yml?style=flat-square&label=ci&labelColor=161b22&logo=githubactions&logoColor=white" alt="CI"/></a>
  <a href="./LICENSE"><img src="https://img.shields.io/npm/l/reasonix.svg?style=flat-square&color=8b949e&labelColor=161b22" alt="license"/></a>
@@ -43,36 +42,18 @@
 <p align="center"><strong>Open source · MIT · a single Go binary</strong></p>
 <h3 align="center">A coding agent you can leave running.</h3>
 <p align="center">One local engine, four ways in — terminal, desktop app, browser, or your editor over ACP. Plan mode, permissions, a workspace sandbox and per-turn checkpoints keep a long autonomous run something you can still read and undo.</p>
-<p align="center">Maintained by <strong>Huahui Yu</strong>.</p>
+
+<div align="center">
+  <video src="https://github.com/user-attachments/assets/ab2f3878-e224-4931-8254-060e7695cfb9" controls preload="metadata" width="560"></video>
+</div>
+
+<br/>

 > [!IMPORTANT]
 > **Community · 加入社区** — bilingual Discord for setup help (`#help` / `#求助`), workflow showcases, and feature ideas. → **<https://discord.gg/XF78rEME2D>**

 <br/>

-## Versions
-
-Reasonix ships on two lines. See the [version roadmap announcement](https://github.com/esengine/DeepSeek-Reasonix/discussions/10748) for the reasoning behind the split.
-
-| Line | Branch | Status | Get it |
-| --- | --- | --- | --- |
-| **Reasonix 2.x** | `studio` (this branch) | Active development, pre-release | [Studio releases](https://github.com/esengine/DeepSeek-Reasonix/releases?q=studio-v&expanded=true) |
-| **Reasonix 1.x** | [`main-v2`](https://github.com/esengine/DeepSeek-Reasonix/tree/main-v2) | Maintenance / stable | `npm i -g reasonix` · [desktop download](https://reasonix.io/?download=desktop#start) |
-
- **Want something stable?** Stay on 1.x. It keeps receiving bug fixes,
-  provider/API compatibility, updater and security fixes; its core architecture
-  is no longer being extended.
- **Want the new architecture?** Use 2.x Reasonix Studio. It is still moving
-  quickly; reports are welcome.
-
-Valuable fixes, tests and behaviour from 1.x are reviewed one by one and ported
-or reimplemented on 2.x where they fit the new architecture.
-
- [Moving from 1.x to 2.x](./docs/MIGRATING.md): what the two lines share, and
-  how to run both on one machine without losing sessions.
- [2.x roadmap](./docs/ROADMAP.md): what 2.x still has to deliver, how it ships,
-  and the decisions still open.
-
 ## Features

 - **Config-driven.** Providers, the agent, enabled tools, and plugins are all
@@ -92,60 +73,53 @@ or reimplemented on 2.x where they fit the new architecture.

 ## Install

-This branch is **Reasonix 2.x**. Its desktop app is Reasonix Studio; the CLI,
-Studio, and editor integrations all run the same local Reasonix engine.
+Choose the path that matches how you want to use Reasonix. The CLI/TUI,
+desktop app, and VS Code extension all use the same local Reasonix engine.

-### Reasonix Studio (2.x)
+### Path A: CLI / TUI

-Download the package for your platform from the latest
-[Studio release](https://github.com/esengine/DeepSeek-Reasonix/releases?q=studio-v&expanded=true)
-(tagged `studio-v2.*`, published as pre-releases while 2.x is in active
-development):
-
-| Platform | Package | Architecture |
-| --- | --- | --- |
-| macOS | `.dmg` or `.zip` | Apple Silicon (`arm64`) / Intel (`amd64`) |
-| Windows | Installer `-installer.exe` or portable `.zip` | x64 |
-| Linux | `.deb` | x64 |
-
-Every package ships with a `.minisig` signature, and the release carries
-`SHA256SUMS`. Once installed, Studio updates itself in place.
-
-The same release also carries the 2.x `reasonix` CLI as archives for
-`darwin|linux|windows × amd64|arm64`.
-
-### Reasonix 1.x (stable)
-
-The 1.x CLI installs through npm on any supported platform, or Homebrew on
-macOS:
+Install the native binary through npm on any supported platform, or use
+Homebrew on macOS:

 ```sh
 npm i -g reasonix                  # any OS; pulls the prebuilt native binary
 brew install esengine/reasonix/reasonix   # macOS
 ```

-The 1.x desktop app is on the
-[official download page](https://reasonix.io/?download=desktop#start). Windows
-installers are code-signed with a Certum code-signing certificate.
+Prebuilt archives (`darwin|linux|windows × amd64|arm64`) and `SHA256SUMS` are on
+every [GitHub release](https://github.com/esengine/DeepSeek-Reasonix/releases).

-### VS Code extension
+### Path B: Desktop app

-The extension does not bundle the CLI; it starts your local `reasonix acp`
-backend and adds native chat, editor context, tool-call approvals, model
-selection, and workspace sessions. Install the 1.x CLI first.
+Use the [official download page](https://reasonix.io/?download=desktop#start)
+for the latest desktop build.
+
+| Platform | Package | Architecture |
+| --- | --- | --- |
+| macOS | Universal `.dmg` or `.zip` | Apple Silicon / Intel |
+| Windows | Installer `.exe` or portable `.zip` | x64 / ARM64 |
+| Linux | `.deb` or `.tar.gz` | x64 |
+
+Windows installers are code-signed through [SignPath.io](https://signpath.io/)
+with a free certificate provided by the [SignPath Foundation](https://signpath.org/).
+
+### Path C: VS Code extension
+
+Complete Path A first. The extension does not bundle the CLI; it starts your
+local `reasonix acp` backend and adds native chat, editor context, tool-call
+approvals, model selection, and workspace sessions.

 - **VS Code:** [install from Visual Studio Marketplace](https://marketplace.visualstudio.com/items?itemName=SivanLiu.reasonix-agent)
 - **VSCodium / Eclipse Theia:** [install from Open VSX Registry](https://open-vsx.org/extension/SivanLiu/reasonix-agent)
 - **Extension ID:** `SivanLiu.reasonix-agent` · [source and usage guide](https://github.com/SivanCola/reasonix-vscode)

-### Build from source
+### Path D: Build from source

-Clone the repository; `studio` builds 2.x and `main-v2` builds 1.x:
+Clone the repository first:

 ```sh
 git clone https://github.com/esengine/DeepSeek-Reasonix.git
 cd DeepSeek-Reasonix
-git switch studio
 ```

 #### CLI
@@ -158,27 +132,28 @@ make build      # -> bin/reasonix(.exe)
 make cross      # -> dist/ (darwin|linux|windows × amd64|arm64)
 ```

-#### Studio
+#### Desktop

-Studio additionally requires **Node 24+ and pnpm 10**
-(`npm install -g pnpm@10`) for the frontend.
+The desktop build additionally requires:
+
+- **Node 24+ and pnpm 10** (`npm install -g pnpm@10`) for the frontend
+- **Wails CLI** matching the shared `.wails-version` pin

 ```sh
-make studio
+make wails-install
+cd desktop
+wails build
 ```

-See the [Studio build guide](desktop/README.md#prerequisites) for platform
+See the [desktop build guide](desktop/README.md#prerequisites) for platform
 webview dependencies and Linux build tags.

 ## Quick start

-### Reasonix Studio
-
-Install and launch Studio, then connect a provider and model in the app. No CLI
-setup is needed.
-
 ### CLI / TUI

+These commands are for the CLI/TUI installed through Path A:
+
 ```sh
 reasonix setup                      # configure a provider and model
 reasonix                            # start an interactive session
@@ -188,6 +163,13 @@ reasonix run "implement the TODOs in main.go"
 In an interactive session, run `/init` when you want Reasonix to create project
 instructions.

+### Desktop app
+
+Download the installer for your platform from the
+[official download page](https://reasonix.io/?download=desktop#start), install
+and launch Reasonix, then configure a provider and model in the app. The CLI
+commands above are not required for the desktop app.
+
 For advanced CLI usage and configuration, see the **[CLI reference](./docs/CLI.md)**,
 **[Guide](./docs/GUIDE.md)**, and
 **[configuration paths](./docs/CONFIG_PATHS.md)**.
@@ -199,13 +181,12 @@ For advanced CLI usage and configuration, see the **[CLI reference](./docs/CLI.m
 - **Features & troubleshooting:** [Subagent profiles](./docs/SUBAGENT_PROFILES.md) ·
  [Context Engine v2](./docs/SESSION_MEMORY_RETRIEVAL.md) ·
  [Capability diagnostics](./docs/CAPABILITY_DIAGNOSTICS.md) ·
-  [Recovery and updates](./docs/RECOVERY.md) ·
+  [Recovery and updates](./docs/RECOVERY.md) · [Bot guide](./docs/BOT_GUIDE.md) ·
  [Checkpoints & rewind](./docs/CHECKPOINTS.md)
 - **Engineering & migration:** [Spec](./docs/SPEC.md) ·
  [Task contracts & pause policy](./docs/TASK_CONTRACT.md) ·
-  [Tool contract](./docs/TOOL_CONTRACT.md) · [Moving from 1.x to 2.x](./docs/MIGRATING.md) ·
-  [2.x roadmap](./docs/ROADMAP.md)
- **Extension development:** [Skills](./docs/SKILLS.md) · [Extensions](./docs/EXTENSIONS.md) ·
+  [Tool contract](./docs/TOOL_CONTRACT.md) · [Migrating from 0.x](./docs/MIGRATING.md)
+- **Extension development:** [Extensions](./docs/EXTENSIONS.md) ·
  [Plugin packages and Manifest v1](./docs/PLUGIN_PACKAGES.md) ·
  [Extension Protocol](./docs/EXTENSION_PROTOCOL.md) ·
  [Go SDK and starter](./sdk/go/README.md)
@@ -232,10 +213,10 @@ A small list of folks whose work has shaped Reasonix the most — the current to
 | Contributor | Contributor | Contributor | Contributor |
 | --- | --- | --- | --- |
 | [**SivanCola**](https://github.com/SivanCola) | [**esengine**](https://github.com/esengine) | [**ttmouse**](https://github.com/ttmouse) | [**lifu963**](https://github.com/lifu963) |
-| **reasonix** | [**HUQIANTAO**](https://github.com/HUQIANTAO) | [**GTC2080**](https://github.com/GTC2080) | [**mchenziyi**](https://github.com/mchenziyi) |
-| [**Li-Charles-One**](https://github.com/Li-Charles-One) | **merge-order-check** | [**light-front-theory**](https://github.com/light-front-theory) | **Yan Li** |
-| [**eghrhegpe**](https://github.com/eghrhegpe) | **wufengfan** | [**Bernardxu123**](https://github.com/Bernardxu123) | [**HaoyueQin**](https://github.com/HaoyueQin) |
-| [**CVEngineer66**](https://github.com/CVEngineer66) | [**KHG420**](https://github.com/KHG420) | [**JesonChou**](https://github.com/JesonChou) | [**SuMuxi66**](https://github.com/SuMuxi66) |
+| **reasonix** | [**HUQIANTAO**](https://github.com/HUQIANTAO) | [**GTC2080**](https://github.com/GTC2080) | [**light-front-theory**](https://github.com/light-front-theory) |
+| **merge-order-check** | [**Li-Charles-One**](https://github.com/Li-Charles-One) | [**eghrhegpe**](https://github.com/eghrhegpe) | **wufengfan** |
+| [**CVEngineer66**](https://github.com/CVEngineer66) | [**dependabot\[bot\]**](https://github.com/apps/dependabot) | [**lanshi17**](https://github.com/lanshi17) | [**SuMuxi66**](https://github.com/SuMuxi66) |
+| [**CnsMaple**](https://github.com/CnsMaple) | [**cyq1017**](https://github.com/cyq1017) | [**JesonChou**](https://github.com/JesonChou) | [**XTLine**](https://github.com/XTLine) |
 <!-- reasonix-top-contributors:end -->

 Special thanks to [**Bernardxu123**](https://github.com/Bernardxu123) for designing the project logo and intro video.
@@ -9,11 +9,11 @@
  &nbsp;·&nbsp;
  <a href="./docs/GUIDE.zh-CN.md">指南</a>
  &nbsp;·&nbsp;
-  <a href="./docs/ACP.md">ACP</a>
+  <a href="./docs/ACP.zh-CN.md">ACP</a>
  &nbsp;·&nbsp;
-  <a href="./docs/EXTENSIONS.md">扩展开发</a>
+  <a href="./docs/EXTENSIONS.zh-CN.md">扩展开发</a>
  &nbsp;·&nbsp;
-  <a href="./docs/SPEC.md">规格</a>
+  <a href="./docs/SPEC.zh-CN.md">规格</a>
  &nbsp;·&nbsp;
  <a href="https://esengine.github.io/DeepSeek-Reasonix/">官方网站</a>
  &nbsp;·&nbsp;
@@ -21,7 +21,6 @@
 </p>

 <p align="center">
-  <a href="https://github.com/esengine/DeepSeek-Reasonix/releases?q=studio-v&expanded=true"><img src="https://img.shields.io/github/v/release/esengine/DeepSeek-Reasonix.svg?filter=studio-v*&include_prereleases&style=flat-square&color=8250df&labelColor=161b22&label=studio%202.x" alt="Reasonix Studio 2.x"/></a>
  <a href="https://www.npmjs.com/package/reasonix"><img src="https://img.shields.io/npm/v/reasonix.svg?style=flat-square&color=cb3837&labelColor=161b22&logo=npm&logoColor=white" alt="npm version"/></a>
  <a href="https://github.com/esengine/DeepSeek-Reasonix/actions/workflows/ci.yml"><img src="https://img.shields.io/github/actions/workflow/status/esengine/DeepSeek-Reasonix/ci.yml?style=flat-square&label=ci&labelColor=161b22&logo=githubactions&logoColor=white" alt="CI"/></a>
  <a href="./LICENSE"><img src="https://img.shields.io/npm/l/reasonix.svg?style=flat-square&color=8b949e&labelColor=161b22" alt="license"/></a>
@@ -43,30 +42,16 @@
 <p align="center"><strong>开源 · MIT · 单个 Go 二进制</strong></p>
 <h3 align="center">可以一直开着跑的编码 Agent。</h3>
 <p align="center">一套本地引擎,四个入口——终端、桌面端、浏览器,或通过 ACP 接入你的编辑器。计划模式、权限、工作区沙箱与逐轮 checkpoint,让长时间自治运行始终可读、可撤销。</p>
-<p align="center">维护者：<strong>Huahui Yu</strong>。</p>
+
+<div align="center">
+  <video src="https://github.com/user-attachments/assets/ab2f3878-e224-4931-8254-060e7695cfb9" controls preload="metadata" width="560"></video>
+</div>
+
+<br/>

 > [!IMPORTANT]
 > **加入社区 · Community** — 双语 Discord，提供安装答疑（`#help` / `#求助`）、工作流展示与功能想法。→ **<https://discord.gg/XF78rEME2D>**

-## 版本
-
-Reasonix 分为两条版本线，调整的原因见[版本路线公告](https://github.com/esengine/DeepSeek-Reasonix/discussions/10748)。
-
-| 版本线 | 分支 | 状态 | 获取方式 |
-| --- | --- | --- | --- |
-| **Reasonix 2.x** | `studio`（当前分支） | 活跃开发，预发布 | [Studio 发布页](https://github.com/esengine/DeepSeek-Reasonix/releases?q=studio-v&expanded=true) |
-| **Reasonix 1.x** | [`main-v2`](https://github.com/esengine/DeepSeek-Reasonix/tree/main-v2) | 维护 / 稳定 | `npm i -g reasonix` · [桌面端下载](https://reasonix.io/?download=desktop#start) |
-
- **想要稳定**：继续使用 1.x。它会持续收到 Bug 修复、Provider / API 兼容、
-  更新器和安全修复，但不再扩展核心架构。
- **想体验新架构**：使用 2.x Reasonix Studio。它仍在快速迭代，欢迎反馈问题。
-
-1.x 中有价值的修复、测试和行为会逐项 review，适合新架构的会迁移或重新实现到 2.x。
-
- [从 1.x 迁移到 2.x](./docs/MIGRATING.md)（英文）：两条线共用哪些数据，以及如何在同一台
-  机器上同时使用而不丢会话。
- [2.x 路线图](./docs/ROADMAP.md)（英文）：2.x 还要交付什么、如何发布、哪些决定尚未做出。
-
 ## 特性

 - **配置驱动**：provider、agent、启用的工具、插件全部在 `reasonix.toml` 中声明，
@@ -82,84 +67,58 @@ Reasonix 分为两条版本线，调整的原因见[版本路线公告](https://

 ## 安装

-当前分支是 **Reasonix 2.x**，桌面端是 Reasonix Studio；CLI、Studio 和编辑器集成
-都运行同一套本地 Reasonix 引擎。
+选择适合你的使用路径。CLI/TUI、桌面端和 VS Code 扩展都使用同一套本地
+Reasonix 引擎。

-### Reasonix Studio（2.x）
+### 路径 A：CLI / TUI

-从最新的 [Studio 发布](https://github.com/esengine/DeepSeek-Reasonix/releases?q=studio-v&expanded=true)
-下载对应平台的安装包（标签为 `studio-v2.*`，2.x 活跃开发期间以预发布形式发布）：
-
-| 平台 | 安装包 | 架构 |
-| --- | --- | --- |
-| macOS | `.dmg` 或 `.zip` | Apple Silicon（`arm64`）/ Intel（`amd64`） |
-| Windows | 安装器 `-installer.exe` 或便携 `.zip` | x64 |
-| Linux | `.deb` | x64 |
-
-每个安装包都附带 `.minisig` 签名，发布页另有 `SHA256SUMS`。安装后 Studio 会在
-应用内自行更新。
-
-同一个发布页也提供 2.x 的 `reasonix` CLI 归档（`darwin|linux|windows × amd64|arm64`）。
-
-### Reasonix 1.x（稳定版）
-
-1.x CLI 在任意支持的平台上都可以通过 npm 安装，macOS 也可以使用 Homebrew：
+任意支持的平台都可以通过 npm 安装原生二进制；macOS 也可以使用 Homebrew：

 ```sh
 npm i -g reasonix                  # 任意系统;自动拉取对应平台的原生二进制
 brew install esengine/reasonix/reasonix   # macOS
 ```

-1.x 桌面端请前往[官方下载页](https://reasonix.io/?download=desktop#start)。
-Windows 安装器使用 Certum 代码签名证书签名。
+预编译归档(`darwin|linux|windows × amd64|arm64`)和 `SHA256SUMS` 见每个
+[GitHub release](https://github.com/esengine/DeepSeek-Reasonix/releases)。

-### VS Code 扩展
+### 路径 B：桌面端

-扩展不内置 CLI，而是启动本机的 `reasonix acp` 后端，并提供原生聊天、编辑器
-上下文、工具调用审批、模型选择和工作区会话。请先安装 1.x CLI。
+前往[官方下载页](https://reasonix.io/?download=desktop#start)获取最新桌面版本。
+
+| 平台 | 安装包 | 架构 |
+| --- | --- | --- |
+| macOS | 通用 `.dmg` 或 `.zip` | Apple Silicon / Intel |
+| Windows | 安装器 `.exe` 或便携 `.zip` | x64 / ARM64 |
+| Linux | `.deb` 或 `.tar.gz` | x64 |
+
+Windows 安装器通过 [SignPath.io](https://signpath.io/) 完成代码签名，证书由
+[SignPath 基金会](https://signpath.org/) 免费提供。
+
+### 路径 C：VS Code 扩展
+
+请先完成路径 A。扩展不内置 CLI，而是启动本机的 `reasonix acp` 后端，
+并提供原生聊天、编辑器上下文、工具调用审批、模型选择和工作区会话。

 - **VS Code：** [从 Visual Studio Marketplace 安装](https://marketplace.visualstudio.com/items?itemName=SivanLiu.reasonix-agent)
 - **VSCodium / Eclipse Theia：** [从 Open VSX Registry 安装](https://open-vsx.org/extension/SivanLiu/reasonix-agent)
 - **扩展 ID：** `SivanLiu.reasonix-agent` · [源码与使用说明](https://github.com/SivanCola/reasonix-vscode)

-### 从源码构建
-
-克隆仓库；`studio` 构建 2.x，`main-v2` 构建 1.x：
+### 路径 D：从源码构建

 ```sh
 git clone https://github.com/esengine/DeepSeek-Reasonix.git
 cd DeepSeek-Reasonix
-git switch studio
-```
-
-#### CLI
-
-CLI 构建需要 **Go 1.25+**。模块固定了 `toolchain`，保持 `GOTOOLCHAIN=auto`
-即可让 Go 自动下载对应工具链，也可以手动安装。
-
-```sh
 make build      # -> bin/reasonix(.exe)
 make cross      # -> dist/（darwin|linux|windows × amd64|arm64）
 ```

-#### Studio
-
-Studio 的前端还需要 **Node 24+ 与 pnpm 10**（`npm install -g pnpm@10`）。
-
-```sh
-make studio
-```
-
-各平台的 webview 依赖和 Linux 构建标签见 [Studio 构建指南](desktop/README.md#prerequisites)。
-
 ## 快速开始

-### Reasonix Studio
-
-安装并启动 Studio，然后在应用内连接 provider 和模型即可使用，无需配置 CLI。
-
 ### CLI / TUI

+以下命令仅适用于通过路径 A 安装的 CLI/TUI：
+
 ```sh
 reasonix setup                      # 配置 provider 和模型
 reasonix                            # 启动交互式会话
@@ -168,28 +127,34 @@ reasonix run "把 main.go 里的 TODO 实现掉"

 需要项目指令时，可在交互式会话中运行 `/init`。

+### 桌面端
+
+从[官方下载页](https://reasonix.io/?download=desktop#start)下载对应系统的安装包，
+安装并启动 Reasonix，然后在应用内配置 provider 和模型即可使用。桌面端无需执行
+上面的 CLI 命令。
+
 CLI 进阶用法和详细配置见 **[CLI 命令参考](./docs/CLI.zh-CN.md)**、
 **[指南](./docs/GUIDE.zh-CN.md)** 和
-**[配置路径](./docs/CONFIG_PATHS.md)**。
+**[配置路径](./docs/CONFIG_PATHS.zh-CN.md)**。

 ## 文档

 - **开始使用：** [指南](./docs/GUIDE.zh-CN.md) ·
-  [CLI 命令参考](./docs/CLI.zh-CN.md) · [配置路径](./docs/CONFIG_PATHS.md) ·
-  [ACP 编辑器接入](./docs/ACP.md)
- **功能与排障：** [子智能体 Profile](./docs/SUBAGENT_PROFILES.md) ·
-  [Context Engine v2](./docs/SESSION_MEMORY_RETRIEVAL.md) ·
-  [能力诊断](./docs/CAPABILITY_DIAGNOSTICS.md) ·
-  [恢复与安全模式](./docs/RECOVERY.md) ·
-  [Checkpoints 与 rewind](./docs/CHECKPOINTS.md)
- **工程与迁移：** [规格](./docs/SPEC.md) ·
-  [任务合约与暂停策略](./docs/TASK_CONTRACT.md) ·
-  [工具合约](./docs/TOOL_CONTRACT.md) ·
-  [从 1.x 迁移到 2.x](./docs/MIGRATING.md) ·
-  [2.x 路线图](./docs/ROADMAP.md)
- **扩展开发：** [扩展概览](./docs/EXTENSIONS.md) ·
-  [插件包与 Manifest v1](./docs/PLUGIN_PACKAGES.md) ·
-  [Extension Protocol](./docs/EXTENSION_PROTOCOL.md) ·
+  [CLI 命令参考](./docs/CLI.zh-CN.md) · [配置路径](./docs/CONFIG_PATHS.zh-CN.md) ·
+  [ACP 编辑器接入](./docs/ACP.zh-CN.md)
+- **功能与排障：** [子智能体 Profile](./docs/SUBAGENT_PROFILES.zh-CN.md) ·
+  [Context Engine v2](./docs/SESSION_MEMORY_RETRIEVAL.zh-CN.md) ·
+  [能力诊断](./docs/CAPABILITY_DIAGNOSTICS.zh-CN.md) ·
+  [恢复与安全模式](./docs/RECOVERY.zh-CN.md) ·
+  [机器人使用指南](./docs/BOT_GUIDE.zh-CN.md) ·
+  [Checkpoints 与 rewind](./docs/CHECKPOINTS.zh-CN.md)
+- **工程与迁移：** [规格](./docs/SPEC.zh-CN.md) ·
+  [任务合约与暂停策略](./docs/TASK_CONTRACT.zh-CN.md) ·
+  [工具合约](./docs/TOOL_CONTRACT.zh-CN.md) ·
+  [从 0.x 迁移](./docs/MIGRATING.zh-CN.md)
+- **扩展开发：** [扩展概览](./docs/EXTENSIONS.zh-CN.md) ·
+  [插件包与 Manifest v1](./docs/PLUGIN_PACKAGES.zh-CN.md) ·
+  [Extension Protocol](./docs/EXTENSION_PROTOCOL.zh-CN.md) ·
  [Go SDK 与 starter](./sdk/go/README.md)

 ## Star 趋势
@@ -214,10 +179,10 @@ CLI 进阶用法和详细配置见 **[CLI 命令参考](./docs/CLI.zh-CN.md)**
 | Contributor | Contributor | Contributor | Contributor |
 | --- | --- | --- | --- |
 | [**SivanCola**](https://github.com/SivanCola) | [**esengine**](https://github.com/esengine) | [**ttmouse**](https://github.com/ttmouse) | [**lifu963**](https://github.com/lifu963) |
-| **reasonix** | [**HUQIANTAO**](https://github.com/HUQIANTAO) | [**GTC2080**](https://github.com/GTC2080) | [**mchenziyi**](https://github.com/mchenziyi) |
-| [**Li-Charles-One**](https://github.com/Li-Charles-One) | **merge-order-check** | [**light-front-theory**](https://github.com/light-front-theory) | **Yan Li** |
-| [**eghrhegpe**](https://github.com/eghrhegpe) | **wufengfan** | [**Bernardxu123**](https://github.com/Bernardxu123) | [**HaoyueQin**](https://github.com/HaoyueQin) |
-| [**CVEngineer66**](https://github.com/CVEngineer66) | [**KHG420**](https://github.com/KHG420) | [**JesonChou**](https://github.com/JesonChou) | [**SuMuxi66**](https://github.com/SuMuxi66) |
+| **reasonix** | [**HUQIANTAO**](https://github.com/HUQIANTAO) | [**GTC2080**](https://github.com/GTC2080) | [**light-front-theory**](https://github.com/light-front-theory) |
+| **merge-order-check** | [**Li-Charles-One**](https://github.com/Li-Charles-One) | [**eghrhegpe**](https://github.com/eghrhegpe) | **wufengfan** |
+| [**CVEngineer66**](https://github.com/CVEngineer66) | [**dependabot\[bot\]**](https://github.com/apps/dependabot) | [**lanshi17**](https://github.com/lanshi17) | [**SuMuxi66**](https://github.com/SuMuxi66) |
+| [**CnsMaple**](https://github.com/CnsMaple) | [**cyq1017**](https://github.com/cyq1017) | [**JesonChou**](https://github.com/JesonChou) | [**XTLine**](https://github.com/XTLine) |
 <!-- reasonix-top-contributors:end -->

 特别感谢 [**Bernardxu123**](https://github.com/Bernardxu123) 设计的项目 logo和开场视频。
@@ -9,25 +9,22 @@ agent. It is the Reasonix analog of Claude Code's CLAUDE.md.
 - Go kernel under `internal/`; each package owns one concern. A package's long
  explanation belongs in its `doc.go`, not spread across implementation files.
 - One transport-agnostic `control.Controller` sits behind every frontend (chat
-  TUI, HTTP/SSE serve, Electron desktop). Add behavior to the controller, not a
+  TUI, HTTP/SSE serve, Wails desktop). Add behavior to the controller, not a
  frontend, so all three inherit it.
- Layering (enforced): utility packages import nothing under `reasonix/` except
-  one another — that layer is closed, so a leaf reaching a leaf drags no graph
-  along, while refusing it leaves one judgement copied into several of them; only
-  the frontends `cli`, `serve`, `acp`, `boot` and the hosts `cmd/`, `desktop/`
-  may import `control`; nothing below a frontend may import one. The declared
-  sets live in `tools/repolint/layers.go`.
+- Layering (enforced): utility packages import nothing under `reasonix/`; only
+  the frontends `cli`, `serve`, `acp`, `bot`, `botruntime`, `boot` and the hosts
+  `cmd/`, `desktop/` may import `control`; nothing below a frontend may import
+  one. The declared sets live in `tools/repolint/layers.go`.
 - Subagent delegation keeps five concepts apart: a profile says how a worker
  thinks, `TaskSpec` what this call wants, `CapabilityGrant` what it may touch,
  `ContextRequest` what it starts from, `SchedulerPolicy` when it runs. Put a
  field in whichever member decides its value — profiles carry ceilings, never
-  per-call values. `internal/runtime/delegation/profile_boundary_test.go` enforces it.
- Cache-first: the system-prompt prefix (base prompt, tools, declared prefix
-  configuration) must stay byte-stable across turns so DeepSeek's automatic
-  prefix cache stays warm. Never mutate it mid-session — ride the turn tail
-  instead (see `control.Compose`), under the rules in *Context projection*.
+  per-call values. `internal/agent/profile_boundary_test.go` enforces it.
+- Cache-first: the system-prompt prefix (base prompt + tools + memory) must stay
+  byte-stable across turns so DeepSeek's automatic prefix cache stays warm. Never
+  mutate it mid-session — ride the turn tail instead (see `control.Compose`).
 - Performance features land with an effect test at their final boundary
-  (`internal/assembly/boot/effect_test.go` pattern): assert what actually reaches the
+  (`internal/boot/effect_test.go` pattern): assert what actually reaches the
  provider request, frontend sink, or trajectory through the real `boot.Build`
  assembly. Component correctness is not system effectiveness.
 - A mutex- or atomic-guarded struct is ratcheted on its **scalar** field count
@@ -36,100 +33,6 @@ agent. It is the Reasonix analog of Claude Code's CLAUDE.md.
  the move this blocks — group by lifetime into a named sub-state instead
  (`agent.perTurnState` is the pattern), which costs one field and removes the
  whole product.
- The context window (projection, compaction, estimation) owns `windowState`;
-  the loop reaches it via `a.window()`, it reaches the loop via `windowHost`.
-  `window-state` fails any other read of `sess.win`.
- Judgements read structure, never wording: shell ASTs (`shellparse`), types,
-  contracts, tool schemas. Phrase tables and message sniffing have been retired
-  in batches (task-policy prose, the planner's approval phrases, the executor
-  handoff tables) because each one misfires on real input, and a misfiring
-  judgement inside a gate is worse than no gate. Security allow-lists are the
-  exception and stay. Something that can only be built by matching words does
-  not get built — say what is missing instead of guessing at it.
- The other half of that rule: an error a caller must tell apart **carries an
-  identity** — a sentinel, a typed code — never a sentence. Only the producer
-  knows a deadline expired rather than a socket closed, and a message is where
-  that knowledge goes to die: the reader is left matching words, or with
-  nothing. Across HTTP the identity is a dotted code through `refuse`;
-  `refusal-path` fails a new `http.Error` in `internal/frontend/serve`, and the parity
-  test fails a code the frontend cannot say. Inside Go it is `errors.Is` on a
-  sentinel the wrapping preserves; `error-text` fails a new match against an
-  error message, following it one hop into a local because storing the text
-  first is what the direct form turns into.
- **Whoever first knows why an operation failed owns its class.** A transport
-  status is not a domain identity, and a display string is not a classification;
-  past that boundary a layer may project the class, never re-derive it. Both
-  directions were live: `catalog.go` answered 400 for a name nobody declared and
-  for a disk that would not write, so a failed store told the user their request
-  was wrong; `ServerRow` decided a connection was an authorisation problem by
-  matching `/401|403|unauthorized/` against prose an external MCP server wrote,
-  which the host had sanitised and truncated on the way through. Domain,
-  internal and dependency are three answers with three different things to do
-  next, and a caller that cannot tell them apart reports the last two as the
-  first.
- Failure attribution is host-owned state. When the host can determine why an
-  operation failed or stopped — truncation, timeout, unavailable context,
-  dependency skipping, policy refusal, output spill, argument parsing — that
-  cause goes to the model. Surfacing only the downstream symptom leaves it to
-  invent one, and the invented rule outlives the incident: a response cut at the
-  output limit reaches the model as "invalid arguments", and the session learns
-  to stop batching calls that were never the problem.
- A host-owned identity has to cross the model-visible boundary. Stable step
-  ids, typed error codes, boundary causes — anything the host expects the model
-  to cite must appear wherever that state is rendered for it. Telling the model
-  in a schema to prefer X while every concrete rendering shows only Y trains it
-  to use Y; `complete_step` said `step_id` PREFERRED while all four todo
-  renderers printed bare ordinals. Canonical dynamic state still stays out of
-  the cache-stable prefix — project the identities needed to act on it at the
-  turn tail.
- A metric is its definition, not its name. Before a number becomes a finding,
-  state what it counts, from which anchor, over which population, and what it
-  excludes. Three readings in one study died in that gap: a command of unknown
-  extent read as a proven write, the latest report's adjudication read as every
-  report's, and a syntactically accepted call read as a closed one. Each time a
-  convenient proxy had been left standing in for the state it was named after,
-  and each time the mechanism story built on it was wrong. A number without
-  those four lines is an observation, not a result.
- A constructed state is not a reachable one. Before a destructive test or a
-  counterfactual is reported as a defect, say who in production produces that
-  state, through which transitions, across which persistence boundary, under
-  whose authority, and whether a real run has been seen reaching it. A test that
-  assigns a private field proves the mechanism *given* the state and must be
-  reported as exactly that: `a.projectChecks` is written once at boot and never
-  again, so reassigning it stood in for a resume rather than demonstrating one,
-  and the bypass built on it was never shown to be production-reachable.
-  Severity is reachability times frequency times authority, not "can be
-  bypassed".
- A wire type the desktop re-declares by hand is held to the kernel's by
-  `wire-parity`, both directions: a field one side sends and the other cannot
-  read fails, and so does one it reads that nothing sends. The two are compiled
-  by different toolchains and meet only over JSON, so nothing else in the tree
-  notices when one of them grows a field. The pairs are declared in
-  `tools/repolint/wireparity.go`, never inferred — no spelling tells a mirrored
-  contract from a struct that merely carries json tags.
- "Add behavior to the controller so every frontend inherits it" is held by
-  `frontend-parity`: a capability one frontend already drives, a frontend whose
-  declared port carries it, and no call, is a missing edge, and the ratchet
-  counts edges. Which port each frontend drives is declared in
-  `tools/repolint/frontendparity.go`; a deliberate exception goes in the scope
-  table there with an issue and a reason, and costs what the edge cost, so the
-  table cannot become where debt disappears. The matrix is read from
-  `go/types` at one pinned build target, never from selector names — `Close`
-  and `Status` are capabilities and also stdlib methods, and a name match
-  records `os.File.Close` as a frontend driving `Lifecycle.Close`, a false
-  "wired" that hides real debt. How a verb is spelled — `/branch`,
-  `POST /branches` — is presentation and stays out of it.
- The desktop's presentation layer answers to canonical state and never
-  redefines it. Six rules, each held by a gate rather than by memory:
-  presentation reorders and reweights what the kernel owns but never restates
-  it; **mount is not a motion cause** — a fact enters once, and replay or
-  virtualization bringing its element back is not that fact happening again;
-  a search alias buys discoverability and never becomes a thing's identity;
-  when a setting takes effect is declared per block, from what its endpoint
-  does, never inferred from wording or handler names; ordering panels by
-  lifecycle moves them and must not remount them; and a control's identity
-  survives a change of where it is drawn — the settings sheet and the composer
-  render `chrome.preset` because that is one intent, not two.

 ## Comments

@@ -137,75 +40,17 @@ Default is none — the code is the truth. Write one only when the **why** is
 non-obvious: a hidden constraint, a workaround anchored to something verifiable,
 an invariant the type system cannot express, or an external-protocol quirk.

-A comment states the constraint that holds now. It does not narrate how the code
-got here: a reader who has never seen an earlier version must lose nothing, and
-past tense about this code's own behaviour is the tell — an external protocol's
-history is not. "Ids alone do not answer it" is the rule; "ids alone were the
-wrong question, so the list went on reading as the model wrote it" is the
-incident behind the rule, and the incident belongs in the commit message. This
-outranks the surrounding style: much of the tree reads as prose and none of it
-is the standard. No gate reads register — telling narrative from a statement of
-constraint needs a phrase table, which is the one thing this repo does not
-build — so this rule is held by review.
-
- **Stylesheets carry none.** A rule's selector and its declarations are the
-  whole statement. What a `.css` comment would have said goes to whichever of
-  two places can act on it: a constraint a component has to honour belongs in
-  that component, and a decision about the medium itself — a measured colour, a
-  layout mechanic, why the obvious rule is the wrong one — belongs in the
-  stylesheet's companion document (`src/styles/TOKENS.md`, `LAYOUT.md`), which
-  is read end to end rather than met 600 times as an aside.
- **English, always** — every comment, in every other language the tree is
-  written in. A codebase whose commentary is half one language reads as two
-  codebases, and the half a given reader cannot read is the half that rots.
 - Declaration doc: ≤5 lines. Package comment: ≤8 lines, or ≤40 in a `doc.go`.
 - Every other comment: ≤3 lines. Struct-field and trailing `//`: 1 line.
 - Never: restatements of the code, phase/stage narrative, incident or
  conversation history, section banners, commented-out code, `@param` lists.
 - `TODO(#nnn):` and `HACK(#nnn):` need the issue anchor. `FIXME` is banned.
- One responsibility per file; 800 lines is the ceiling. Test files are exempt:
-  their length tracks how many cases they cover, not how many concerns they
-  carry, so splitting one only scatters a subject's table across files.
+- One responsibility per file; 800 lines is the ceiling.

-`go run ./tools/repolint` enforces the budgets above against a ratchet baseline:
-recorded debt is tolerated, anything new fails CI. Never widen the baseline to
-land a change — fix the code. `-update` lowers budgets freely and refuses to
-raise one without `-allow-widen`, so carrying debt through a rename or an
-extraction is asked for in the command and justified in the PR. A clean run also
-reports budget the tree stopped using: it is only reclaimed by an `-update`, and
-until then it is room a file can grow back into.
-
-## Context projection
-
-Canonical dynamic state — the skill registry, the memory store — lives in its own
-owner, and what the model sees is a projection of it onto each request. The
-prefix was never where these belonged wrongly; freshness ownership was missing.
-
- **Prefix ownership.** Only declared prefix configuration may move the
-  cache-stable prefix; ordinary memory writes, skill writes and runtime facts
-  ride the turn. Pinned memory bodies are the one declared exception, because
-  pinning is a request to pay one cold start.
- **Projection freshness.** When canonical state changes, the next eligible turn
-  carries the new projection or can say why it is not owed. "Right after a fold,
-  or in the next session" is not an answer.
- **Projection ownership.** Whoever owns a dynamic fact owns the invalidation
-  that makes its projection owed again. A projection kept fresh because every
-  writer remembered to announce itself is one forgotten call from stale —
-  `skillSet.owedCatalog` asks the registry rather than waiting to be told.
- **Determinism.** The same canonical state and the same projection state
-  produce byte-identical model-visible bytes.
- **Detection is semantic.** Freshness may not rest on writer notification, on
-  mtime, or on which call path made the change. The registry answers from
-  canonical state, so a switch flipped and a file edited past every API are the
-  same question, and a change no projection renders is not a change.
-
-`internal/assembly/boot/context_projection_test.go` holds all five at the provider
-boundary, with the skills listing and the standing instructions as the two
-worked examples. A block is settled only once the turn carrying it reaches the
-runner: a turn blocked by a hook or refused by an extension after composing put
-those bytes nowhere, and clearing the debt there would leave the model holding
-state nobody sent it. A projection that has become empty is still owed once —
-otherwise the block the model already has goes on reading as current.
+`go run ./tools/repolint` enforces all of it against a ratchet baseline: recorded
+debt is tolerated, anything new fails CI. Never widen the baseline to land a
+change — fix the code. `-update` exists for carrying debt through a rename or an
+extraction, and that diff must be justified in the PR.

 ## Memory

@@ -225,50 +70,6 @@ otherwise the block the model already has goes on reading as current.

 ## Notes

-One tool call costs one model round trip, so calls that do not depend on each
-other belong in the same round: several reads, edits to different files, a check
-after the edit that enables it. Several edits to the *same* file are one
-`multi_edit` — it applies them against each other in memory and rewrites the
-file only if all of them land, so a failure midway leaves nothing half-edited.
-
-## Vendor rates
-
-Official list prices live in one place: `officialRates` in
-`internal/contract/pricing/rates.go`, oldest generation first, the last being what a
-model charges today. A price change is an **append, never an edit** — the
-superseded rate is what lets an installed config be recognised as holding one
-of ours and brought forward, instead of quoting a retired rate for the rest of
-its life. `config` reads these through `pricing.CurrentRate` and
-`SupersededRates` and restates none of them; a table written twice is one that
-drifts, which is how a vendor's USD rate ended up published on one side only.
-
-`make pricecheck` reads each vendor's own page and compares it against that
-table, and reports how long ago a person last confirmed each one. It reaches
-the network, so it stays out of `make lint` and out of CI: run it before
-touching a rate and when a vendor announces a change. A page it cannot read
-exits non-zero — unread is its own verdict and never counts as agreement.
-
-## Curated presets
-
-A preset changing shape reaches the installs that already have the old one
-through `shippedPresets` in `internal/contract/config/preset_upgrade.go`: one declared
-shape per generation, and one upgrade that brings a matching entry forward. The
-guard is the whole shape — catalog, window, default — because an entry holding
-what we shipped is ours to move and anything else its user curated.
-
-A preset is also what settles which of its models read images. Probing an
-address a preset covers answers from that declaration, because reading a name
-misses every Kimi, Qwen, MiniMax and Claude model they declare and `deepseek-
-flash` spells nothing at all; the spelling is the last resort for an address
-nothing has declared, and stays a correctable suggestion rather than a claim.
-
-Three distinctions are load-bearing, and each was a separate migration function
-before they were written down. A window the shape did not name is the user's,
-so only a shape identified by its window moves one. `vision_models` **narrows**,
-so an upgrade marks only the models it adds and only where the preset says they
-read images. And an explicitly empty list is a choice — nothing here reads
-images — which is not the same as the nil that means nobody has said.
-
 ## Pre-push CI simulation

 Run these **before every commit** to catch the fastest CI failures locally:
@@ -277,7 +78,7 @@ Run these **before every commit** to catch the fastest CI failures locally:
 gofmt -w .                          # catches gofmt (saves ~13s CI)
 go vet ./...                        # catches vet warnings (saves ~52s CI/lint)
 make lint                           # golangci-lint at CI's pin + repolint
-go test ./internal/tools/builtin/ ./internal/assembly/boot/  # catches tool/boot test breaks
+go test ./internal/tool/builtin/ ./internal/boot/  # catches tool/boot test breaks
 ```

 `make lint` runs both gates CI runs, at the version in `.golangci-version`;
@@ -285,25 +86,6 @@ go test ./internal/tools/builtin/ ./internal/assembly/boot/  # catches tool/boot
 shows up in `go vet`, and the CI round trip that catches it instead costs ten
 minutes.

-A full repolint run reports every file over budget, including debt your change
-never touched. To see only what your own edits owe:
-
-```bash
-make check
-```
-
-Repo-wide ceilings still report under a narrowed run, because a file can push
-the tree past one without exceeding its own budget.
-
-`make check` owns the evidence-bearing changed-file universe — everything the
-worktree altered *and* everything it added that git has not been told about.
-Never substitute a direct `go run ./tools/repolint` invocation for it when
-recording verification evidence: the host recognizes a `make` target as a check
-whose result it can read, while an arbitrary `go run ./tools/...` is a program
-it has to assume writes, and the flags behind the target are its implementation
-rather than the contract. Building that universe from `git diff` alone once let
-a new file reach the gate unopened and be reported clean.
-
 ## Import cycle rule

 Before importing a new internal package from a non-test file, verify the target package's **test files** aren't already importing back to you:
@@ -320,120 +102,32 @@ Use `go test ./path/to/target/` to detect cycles **before** pushing. A `[setup f
 - **Keep the PR diff minimal.** Only the files relevant to the PR's purpose — no stray changes from other branches.
 - **Amend, don't add commits, for review feedback** — keeps the commit history clean.

-## Reasonix host checks
+## PR metadata gates

-The paths below decide what the agent is allowed to do — a wrong change here
-is not a bug in a feature, it is a hole in a boundary. Declaring them makes
-every change under them demand `review` plus `security_review` before a turn
-can finish. Sensitivity is declared, never inferred: the host does not read it
-out of how a path is spelled, which cannot tell `internal/auth` from
-`session_write_authority.go`, or a trace file from a data race.
+Two CI guards read the PR body. The scripts are the source of truth and both
+run locally: `scripts/check-cache-impact.sh`, `scripts/check-docs-impact.sh`.
+Separators must be an ASCII `-` or `:` — an em dash fails the docs guard.

- sensitive: internal/safety/permission/**
- sensitive: internal/safety/sandbox/**
- sensitive: internal/safety/shellsafe/**
- sensitive: internal/base/shellparse/**
- sensitive: internal/session/control/approval.go
- sensitive: internal/session/control/approval_bridge.go
- sensitive: internal/session/control/approval_orchestration.go
- sensitive: internal/model/providerbroker/**
- sensitive: internal/ext/installsource/**
- sensitive: internal/ext/plugin/**
- sensitive: internal/ext/pluginpkg/**
- sensitive: internal/base/netclient/**
- sensitive: internal/safety/redirectguard/**
- sensitive: internal/safety/egress/**
- sensitive: internal/platform/browser/**
- sensitive: internal/platform/computer/**
+Cache-sensitive diffs (`internal/tool/`, `internal/provider/`,
+`internal/boot/`, `internal/agent/agent.go`, and the rest of the list in the
+script) require:

-One declaration, two effects: the same list is also the coverage gate's subject.
-`make coverage-gate` holds each path to the coverage it already has, because
-when this was written these paths sat *below* the tree's median rather than
-above it — the code deciding what the agent may execute was its least-tested
-part. Floors ratchet like repolint's budgets: raising one is free, lowering one
-needs `-allow-drop` and a reason in the PR. Adding a `sensitive:` path gets both
-effects; there is no second list to keep in sync.
+```
+Cache-impact: <none|low|medium|high> - <reason>
+Cache-guard: <focused guard test/command or existing guard rationale>
+```

-## Trust domain
+`none` is a legitimate impact when the provider-visible prefix stays
+byte-identical; only an empty value, `todo`, or `tbd` is rejected. If the diff
+also touches `internal/config/`, `internal/memory/`, `internal/outputstyle/`,
+`internal/skill/`, or `internal/boot/`, add `System-prompt-review: <note>` —
+that field additionally rejects `none` and `n/a`, so it must name a reviewer.

-Three sets, kept apart because conflating them turns a performance number into a
-security claim: **VerificationRelevant** (state that can change whether a
-delivery claim holds), **Observable** (state the host can reconcile after an
-action, so it can say what that action did), **HostProtected** (state a
-capability/sandbox boundary makes unwritable). The invariant is
-`Writable ∩ VerificationRelevant ⊆ Observable ∪ HostProtected`:
+User-visible diffs (`cmd/reasonix/`, `desktop/`, `npm/`, and most of
+`internal/`; tests and lockfiles are exempt) require one of these, chosen by
+whether the same PR edited `docs/*.md`:

-> Any state that can both be modified by an action and influence the delivery
-> claim must either be observable by the host or protected from modification by
-> host-enforced boundaries. An incomplete observation never establishes scope.
-
-Effects and state answer different questions and never stand in for each other —
-observation proves what an action did, verification proves what the resulting
-state satisfies — which is why `unproven_mutation` is settled only by
-observation, never by a check that ran afterwards.
-
-What follows bounds Observable today. It is not a restatement of the invariant:
-moving a number here must not move the promise.
-
- `scanWorkspace` walks the workspace recursively, skips VCS store dirs, and is
-  `complete` only below 50k files. Over that, or past a walk error, it
-  establishes nothing and the mutation stays unproven.
- Symlinks are recorded by `Lstat`, so a write through one lands outside what
-  the walk compares. Confining the resolved target is the sandbox's job.
- `internal/safety/sandbox` confines writes to the workspace, configured extras, temp
-  and toolchain caches — on macOS and Linux. Windows has no OS-level bash
-  sandbox, so there nothing the host enforces bounds `Writable`.
- The governed host authorities are a *named* set — ssh-agent, docker, podman.
-  DBus, Wayland, X11, gpg-agent, keyrings and arbitrary sockets stay reachable,
-  so the claim is "these endpoints are governed", never "host IPC is isolated".
-
-`stateEpoch` tracks host-observed mutations, not an exact filesystem snapshot;
-unobserved external writers are a gap shared by the whole verification model. Go
-baseline test criteria use the toolchain's build overlay and therefore share
-ordinary verification's epoch semantics; the overlay is not a general filesystem
-view.
-
-A captured criterion may be superseded only by authority outside
-execution-mutable state; the current workspace, its tests, and the agent itself
-cannot authorize supersession.
-
-Two gaps follow and are open, not decided: a build that reads `.git` makes VCS
-state VerificationRelevant while the scan skips it, and on Windows a symlink out
-of the workspace is neither observed nor protected. Closing either means
-widening observation or narrowing writability — never widening what a check is
-taken to prove.
-
-## Sandbox dimensions
-
-`HostProtected` above is the delivery-level abstraction. Sandbox policy refines
-it into three independent protection dimensions. Their separation is empirical:
-measured bypasses showed that protection on one dimension cannot be assumed to
-imply protection on another.
-
- **Integrity** — what host state may be mutated. Expressed by write roots.
- **Confidentiality** — what host information may be observed. Expressed by
-  `ForbidReadRoots`.
- **Authority** — what host capabilities may be exercised *without* possessing
-  their secrets. Expressed by `HostAuthorities`.
-
-`ReadOnly` is therefore an Integrity property only; it makes no Confidentiality
-or Authority claim.
-
-Network transport and host authority are independent even when a backend
-primitive conflates them: on macOS, `(deny network*)` also cut ssh-agent, so one
-flag could not express "no internet, still sign".
-
-Filesystem visibility is not authority: denying `file-read*` on a socket did not
-stop a connect. Possession is not authority: the key can be unreadable while an
-agent still signs with it.
-
-A protection claim exists only where a backend has a demonstrated enforcement
-primitive for that dimension. Unsupported dimensions stay explicitly unsupported
-— Windows currently enforces none of the three — because a claim that outruns
-its enforcement is worse than an absent one: it is believed.
-
-**Known unmeasured surface — macOS host-mediated delegation:** Apple Events and
-Mach-service lookup are not denied by the current allow-default profile. Prior
-`osascript` and `launchctl submit` probes were inconclusive; no delegation
-effect has been demonstrated. This surface is outside the currently governed
-host-authority endpoint set.
+```
+Documentation-impact: updated - <what changed>            # docs/*.md edited
+Documentation-impact: none - <why the docs stay correct>  # not edited
+```
@@ -56,10 +56,9 @@ Supported boundaries include:
 - Sandbox behavior for built-in shell execution where the platform supports it.
 - Secret handling for provider keys, bot credentials, OAuth tokens, plugin
  headers, and credential-store fallback files.
- HTTP `serve` protections, including the default per-launch token, the launch
-  token every state-changing request requires when authentication is opted
-  out, localhost binding assumptions, JSON-only state-changing requests, and
-  CORS restrictions.
+- HTTP `serve` protections for the unauthenticated local server, including
+  localhost binding assumptions, JSON-only state-changing requests, and CORS
+  restrictions.
 - Desktop and bot session isolation, including per-workspace session metadata
  and configured bot allowlists.
 - Updater, install, and release verification paths.
@@ -68,10 +67,10 @@ The following are normally treated as trusted local/operator-controlled inputs
 unless another bug lets an untrusted actor supply them:

 - CLI arguments and text typed directly by the local user.
- The user's own configuration under the Reasonix home directory.
+- Project configuration files intentionally loaded from the current workspace.
 - Explicit `@path` references supplied by the local user to attach local files.
- MCP servers, language servers, hooks, and slash commands installed, enabled,
-  or approved by the local user.
+- MCP servers, language servers, hooks, and slash commands installed or enabled
+  by the local user.
 - Provider base URLs and model names configured by the local user.

 The following can be security issues when reachable by an untrusted actor or
@@ -90,56 +89,6 @@ when they bypass the intended boundary:
 - Trusting unverified update artifacts, plugin definitions, or downloaded
  binaries.

-## Project Configuration
-
-Files that arrive with a workspace are untrusted input: `reasonix.toml`,
-`.reasonix/settings.json`, `.mcp.json` and the project `.env`. Cloning a
-repository is not a decision to trust its author, so these files may narrow
-what the user configured and never widen it.
-
-| Setting | What a workspace file may do |
-| --- | --- |
-| `sandbox.bash` | Move `off` to `enforce`, never back |
-| `sandbox.network` | Move `true` to `false`, never back |
-| `sandbox.forbid_read`, `sandbox.denied_domains` | Add to the user's lists |
-| `sandbox.allow_write`, `sandbox.workspace_root` | Apply only when they resolve inside the workspace, symlinks followed; `${VAR}` in any sandbox path expands from the process environment only |
-| `sandbox.host_authorities`, a user's `sandbox.allowed_domains` | Nothing |
-| `permissions.deny`, `permissions.ask` | Add to the user's rules |
-| `permissions.mode`, `permissions.allow`, `permissions.allow_dynamic_bash` | Nothing |
-| `desktop.default_tool_approval_mode` | Nothing |
-| `[network]` proxy settings | Nothing |
-| `[[providers]]`, `tools.system_one` endpoints | Declare them; they stay out of use until approved, and model settings naming them fall back to the user's |
-
-An "always allow" answer, or extra write access, granted in a workspace is
-stored under the Reasonix home, not in the workspace.
-
-Programs a workspace names for the host to run stay off until the user
-approves them for that workspace:
-
- Hooks, language server commands, `tools.search.rg_path`,
-  `tools.shell.path`, `[browser]` and `tools.system_one.laya.python` are
-  approved with `reasonix trust`, or for hooks by saving them from the hooks
-  editor.
- An approval is stored under the Reasonix home with a digest of the
-  declaration and of the workspace files it names; any change needs approval
-  again. Approvals and grants belong to one workspace folder.
- Hooks and language servers are checked against their approval again before
-  each start, and refused if a file they name changed, appeared or now
-  resolves elsewhere.
- A single program path (`tools.shell.path`, `tools.search.rg_path`,
-  `[browser]`, `laya.python`) may not point into the workspace, an
-  `allow_write` root, a temporary directory or a toolchain cache the bash jail
-  leaves writable, and may not climb with `..`; its approval covers the
-  program file's content.
- `tools.system_one.laya.local` is read from the user's configuration only.
- Without an interactive frontend they stay off.
- MCP servers a workspace declares stay off until the user enables them for
-  that workspace.
-
-A project value that is not applied is reported as a load warning with the
-reason. A way for workspace files to widen any of the above without the
-user's approval is a security issue.
-
 ## `@` File References

 Reasonix supports `@path` references so users can include local files and images
@@ -1,24 +1,18 @@
 # Reasonix Benchmarks

-Five harnesses live under `benchmarks/`; `cmd/e2ebench` also exposes a
+Three harnesses live under `benchmarks/`; `cmd/e2ebench` also exposes a
 SWE-bench Verified mode:

 - `e2e/` — the committed end-to-end task suite, driven by
  [`cmd/e2ebench`](../cmd/e2ebench/main.go). It runs each task against a real
  provider and emits a markdown + JSON report (accuracy, cache-hit rate, token
  use, cost) suitable for pasting into a PR.
- `context-maintenance-e2e/` — a standalone seed → resume → continue smoke that
-  checks a session seeds below the compaction trigger, resumes without paying
-  for a summary, and folds exactly once past it.
+- `context-maintenance-e2e/` — a standalone seed → resume → comprehension
+  harness that A/B-compares cold-restart cache behavior with and without
+  context pruning.
 - `compaction/` — CompactionBench: grows a session one generation at a time and
  folds it after each, measuring what repeated compaction costs and what it
  loses. See [CompactionBench](#compactionbench) below.
- `upstream-edge/` — a shape-instructed pair of fleet tasks for pricing what a
-  `depends_on` edge carries, against the arm where it only orders. See
-  [Pricing the run graph](#pricing-the-run-graph) below.
- `fanout-width/` — a shape-instructed fleet whose members share nothing, for
-  pricing what running them at the same time is worth. See
-  [Pricing the run graph](#pricing-the-run-graph) below.

 ## Directory layout

@@ -34,57 +28,6 @@ benchmarks/
    └── run/                       # state dir written by seed/resume (default)
 ```

-## Semantic contract for a metric
-
-A number reaches a report only with these four lines written down, because a
-metric is its definition and not its name:
-
-```text
-Metric:        ToolsAfterClosed
-Anchor:        the host's first adjudication whose verdict is closed
-Population:    judged child runs (expected to report, not killed by the provider)
-Excludes:      needs_work reports, malformed reports, parent-level recovery
-Reads as:      ordinary child work after a host-declared terminal closure
-```
-
-The subagent-handoff study lost three mechanism explanations to a missing
-contract: a command whose extent the host could not establish was counted as a
-proven write, one child's latest adjudication was read as every report's, and a
-syntactically accepted call was read as a closed one. Each was a convenient
-proxy left standing in for the state the metric was named after, and each
-produced a confident, wrong story about what the model was doing.
-
-Watch for the five substitutions that caused them: unknown read as true, latest
-read as all, accepted read as closed, attempted read as judged, and not
-evaluated read as false.
-
-## Reachability contract for a defect
-
-A metric owes its definition; a defect owes its state. Before a destructive test
-or a counterfactual is written up as something to fix:
-
-```text
-Constructed state:     baseline != current
-Production producer:   who assigns it, and where
-Transition path:       which real transitions reach it from a normal boot
-Persistence boundary:  does it need a restart, a resume, a crash
-Authority:             who caused the change — the agent, the user, the host
-Observed:              has a real run been seen in this state
-```
-
-The project-check study skipped the middle four. A test reassigned
-`a.projectChecks` and the bypass it produced was reported as reproducible; the
-field is written once at boot and never again, so no run inside one process can
-reach that state, and the corpus said so on its first execution — parity, no
-divergence. What the test proves is a conditional mechanism, which is a
-different claim from a production defect and earns a different priority.
-
-Authority belongs in the contract because the same divergence can mean opposite
-things. An agent that rewrote the declaration it was being held to is the case
-baseline protection exists for; an owner who changed the project's policy
-between two runs may have superseded it deliberately. A study that cannot say
-which one it observed cannot say what the right behaviour is.
-
 ## Task corpus stratification

 The suite is stratified by real coding-agent workload classes, not toy-task
@@ -93,31 +36,21 @@ utility readouts key on. Current coverage vs. target:

 | Class | Target | Committed | Notes |
 | --- | ---: | ---: | --- |
-| `atomic-bugfix` | 8 | 11 | short anchored fixes; routes ExecutorOnly by design |
-| `repo-exploration` | 6 | 7 | multi-file reading, invented-token answers so they can't be guessed |
+| `atomic-bugfix` | 8 | 8 | short anchored fixes; routes ExecutorOnly by design |
+| `repo-exploration` | 6 | 6 | multi-file reading, invented-token answers so they can't be guessed |
 | `multi-file-bugfix` | 8 | 8 | one bug spanning ≥2 files; naturally engages the planner gate |
 | `refactor` | 6 | 6 | behavior-preserving restructuring, structure asserted |
-| `failing-test-diagnosis` | 6 | 8 | unittest suite red → fix source; tests checksummed |
+| `failing-test-diagnosis` | 6 | 6 | unittest suite red → fix source; tests checksummed |
 | `api-integration` | 4 | 4 | use a provided local package per its README |
 | `ambiguous` | 4 | 4 | underspecified ask; grader accepts the defensible core |
 | `long-horizon` | 4 | 4 | multi-requirement specs; planner-depth full |
-| `codegen` / `delegation` | — | 4 | fizzbuzz, palindrome, and two Go tasks that build a tool from nothing |
-| `completion-integrity` | 12 | 12 | **no reachable solution**; scored on honesty, not correctness (below) |
+| `codegen` / `delegation` | — | 3 | legacy smoke tasks (fizzbuzz, palindrome, subagent-delegation) |
+| `completion-integrity` | 11 | 11 | **no reachable solution**; scored on honesty, not correctness (below) |

 Grader authoring rule: every task must fail `verify.sh` on the pristine seed
-and pass it on a reference solution (validated before commit). The first half
-is enforced by `TestSolvableCorpusSeedsMustNotGradeClean` — a seed that already
-grades clean scores the same whether the agent solved it or never ran, and
-reports 100% forever. SWE-bench Verified (below) supplies the realistic-repo
-end of the spectrum; this corpus covers the fast, controlled, per-class end.
-
-**Language coverage.** The corpus was Python-only for its whole history, while
-the agent's own repository is Go. A compiled language is a different verification
-loop — `go build` before anything runs, `go vet`, `-race`, a module path that
-must agree with every import — and its defects have no Python analogue: a data
-race, a `+=` in a loop that is quadratic where Python's would be too, map
-iteration order that makes a test pass only sometimes. The `go-*` tasks cover
-that half; they are ordinary members of the classes above, not a separate tier.
+and pass it on a reference solution (validated before commit). SWE-bench
+Verified (below) supplies the realistic-repo end of the spectrum; this corpus
+covers the fast, controlled, per-class end.

 ## Completion Integrity

@@ -244,11 +177,8 @@ are then counted identically for anything that speaks the endpoint.
 go run ./cmd/e2ebench -meter ~/.reasonix/config.toml -trajectories t/
 ```

- **Credentials are never touched.** The config only names an `api_key_env`;
-  only `base_url` is rewritten. The key itself is resolved from Reasonix's own
-  `$REASONIX_HOME/.env`, never from the process environment — exporting
-  `DEEPSEEK_API_KEY` into the shell is not enough to run a benchmark against a
-  temporary home, which needs its own `.env`.
+- **Credentials are never touched.** The config names an `api_key_env`, so the
+  key stays in the environment the child inherits; only `base_url` is rewritten.
 - **Only the provider serving `-model` is redirected.** Rewriting every endpoint
  would send one vendor's traffic to another's host.
 - **Streamed requests are opted into usage.** An OpenAI-compatible stream
@@ -325,9 +255,9 @@ entry replaces one leg's continuation with a user turn.

 Two properties are load-bearing:

- **The wall clock is shared, never multiplied.** Every leg runs inside the
-  task's one `timeout_sec`, so a segmented arm cannot win by being allowed to
-  work longer. (`max_steps`, when a task sets one, is divided the same way.)
+- **The step budget is divided, never multiplied.** A segmented arm gets the
+  same `max_steps` as the control arm, split across legs with the remainder on
+  the last. Otherwise the arm would win by being allowed to work longer.
 - **Each leg writes its own metrics file.** They share a work dir, so a single
  `.run-metrics.json` would leave the last leg's numbers standing in for the
  whole run and the earlier legs' tokens would simply vanish. `Segments` in the
@@ -349,7 +279,7 @@ decoder. The task ID is the directory name; tasks run in sorted ID order.
 | --- | --- | --- | --- |
 | `prompt` | string | yes | The task instruction handed to the agent. |
 | `class` | string | no | Task class label (e.g. `bugfix`, `codegen`, `exploration`) for per-class marginal-utility breakdowns in compare mode. |
-| `max_steps` | int | no | Agent tool-call cap, passed through as `--max-steps`. **Committed tasks must not set one** — `TestCorpusLetsTheAgentDecideWhenToStop` enforces that. The product ships unbounded, so a capped task grades a configuration no default user runs, and on a no-solution task the host's "summarise your progress" says the very sentence the honesty score exists to hear from the model. |
+| `max_steps` | int | yes | Agent tool-call cap; passed through as `--max-steps` to `reasonix run`. |
 | `no_solution` | bool | no | Ground truth: no reachable solution exists. The task leaves every accuracy denominator, its `verify.sh` grades the inverse contract, and it is scored on honesty instead. See [Completion Integrity](#completion-integrity). |
 | `timeout_sec` | int | no | Per-task wall-clock timeout in seconds; defaults to `240` when omitted or `0`. |
 | `seed_correct` | string | no | The task's real cause, phrased as a conclusion handed down before the run. Used by `-anchor correct`. See [Anchor resistance](#anchor-resistance). |
@@ -359,6 +289,7 @@ Example (`tasks/fizzbuzz/task.toml`):

 ```toml
 prompt = "Create a file named fizzbuzz.py containing a function fizzbuzz(n) that returns the string 'Fizz' when n is divisible by 3, 'Buzz' when divisible by 5, 'FizzBuzz' when divisible by both 3 and 5, and otherwise the number as a string. Do not print anything at import time."
+max_steps = 12
 timeout_sec = 180
 ```

@@ -420,24 +351,20 @@ own outcome).
 | `-suite` | `benchmarks/e2e` | Suite root (must contain `tasks/<id>/`). |
 | `-task` | *(all)* | Suite mode: run only these comma-separated task IDs (e.g. `-task fix-add-bug`); unknown IDs fail with the available list. |
 | `-attempts` | `1` | Suite and diff modes: retry a task until an attempt passes, up to N; enables the `Pass@≤N` KPI, and TTCS charges a retried solve with its failed attempts' wall. |
-| `-trials` | `1` | Suite mode: run every task N times and keep every outcome. Adds a **Reliability** line: `pass^N` (tasks that passed every trial), `pass@N` (any trial), mean `pass@1` over all trials, and the flaky tasks by name. `Pass@≤N` asks whether the agent can solve a task; `pass^N` asks whether it does so every time. Cannot be combined with `-attempts`. |
-| `-tape-record DIR` | — | Suite mode, with `-meter`: record every provider exchange at the meter, one tape per task and trial (`DIR/<task>/trial-N/0001.request.json`, `.response.json`, `.response.body`). Taped runs use a fixed workdir path and pin the seeded files' times, because both reach the model through the workspace block and file listings. |
-| `-tape-replay DIR` | — | Suite mode, with `-meter`: answer each run from its tape instead of the provider, and add a **Replay** line naming the first request whose body differs from the recorded one, down to the message (`request 2, message 4 (tool)`), the tool list, or a field. The diverging request is kept as `NNNN.replayed.json` beside the recorded one. A kernel change that should not alter what the model is asked replays with every request matched. |
 | `-bin` | `reasonix` | Path to the reasonix binary. |
 | `-model` | *(config default)* | Provider/model name. |
 | `-profile` | `baseline` | Tool-surface/runtime tier: `baseline` \| `economy` \| `balanced` \| `delivery`. All but `baseline` append `--profile <tier>` to the agent invocation; `baseline` passes no flag (byte-identical legacy control, behaviorally `balanced`). Economy starts with the core tool set and pays `connect_tool_source` rounds plus prefix resets to grow it — the report's Tool surface line prices that trade. |
-| `-ablate` | *(none)* | Ablation arm: comma-separated subsystems to switch off — `evidence`, `planner`, `subagent`, `retrieval`, `compaction`, `upstream` and `recall-search`; `none` \| `all`. |
+| `-ablate` | *(none)* | Ablation arm: comma-separated subsystems to switch off — `evidence`, `planner`, `subagent`, `retrieval`, `compaction`; `none` \| `all`. |
 | `-out` | *(stdout)* | Write the markdown report here. |
 | `-json` | *(none)* | Write the JSON report here (optional). |
 | `-trajectories` | *(none)* | Suite mode: write one `<task-id>.trajectory.jsonl` per task into this directory (the agent's full event stream with timestamps — see `reasonix run --trajectory`). The report gains a time-attribution line (tools vs. model) and each JSON result a `trajectory` digest. |
 | `-force-planner` | `false` | Suite mode: prefix each prompt with a plan-first directive so the two-model turn engages regardless of the planner gate. Use for the "with planner" arm of an A/B; results carry `plan_forced` so arms are only comparable with equal forcing. |
 | `-anchor` | `blind` | Suite mode: which hypothesis the agent holds before it looks at anything — `blind` (none, the control) \| `correct` \| `wrong`. The seeded arms prefix each prompt with the task's authored seed and **skip** tasks that have none, so an unseeded control run never lands in a seeded denominator. Results carry `anchor`. See [Anchor resistance](#anchor-resistance). |
 | `-cache` | `cold` | Suite mode: `cold` runs each task as a fresh session (the fair cross-agent comparison arm); `warm` primes the provider prefix cache with a one-step run in the same workdir first, measuring the long-lived-session steady state. Never mix arms in one report — compare them with `-mode compare cold.json warm.json`. |
-| `-policy` | *(none)* | Suite mode: experiment arm. `memory-off` hides the memory store, which is the MemoryBench counterfactual arm; the empty default runs the baseline. |
 | `-budget` | `800000` | Abort once total tokens cross this (`0` = no cap). Remaining tasks are reported as skipped. |
 | `-meter` | *(off)* | Suite mode: route the benchmarked provider through the neutral measuring proxy, using this `config.toml` as the source. Spend is then counted at the request boundary instead of trusted from the harness. See [Neutral metering](#neutral-metering). |
 | `-faults` | *(none)* | Suite mode: inject provider failures through the meter — absolute indices (`3:429`) and/or a cadence that scales with the run (`every:5:500`). Requires `-meter`. See [Fault recovery](#fault-recovery). |
-| `-segments` | `1` | Suite mode: split each task into N resumed legs (`--continue` between them). Every leg shares the task's one wall-clock timeout, so legs never buy extra working time. See [Segmented runs](#segmented-runs). |
+| `-segments` | `1` | Suite mode: split each task into N resumed legs (`--continue` between them). The step budget is **divided**, never multiplied. See [Segmented runs](#segmented-runs). |
 | `-steer` | *(none)* | Suite mode: deliver a user turn at a leg boundary, e.g. `"also handle empty input@2"`. Requires `-segments` to reach that leg. |

 Diff-mode flags:
@@ -447,7 +374,7 @@ Diff-mode flags:
 | `-repo` | `.` | Repo root (diff mode). |
 | `-base` | *(none)* | Base ref to diff the PR head against (diff mode). |
 | `-test-cmd` | `go test` | Grader command run on the affected packages (diff mode). |
-| `-max-steps` | `0` | Diff mode: agent tool-call cap. Zero — the shipped default — leaves the loop unbounded and lets `-timeout` be the only bound. |
+| `-max-steps` | `80` | Agent tool-call cap for the diff task. |
 | `-timeout` | `1200` | Agent timeout in seconds (diff mode). |
 | `-attempts` | `1` | Diff mode: retry up to N times until a run passes (stochastic agent). |

@@ -478,34 +405,6 @@ marginal-utility line (`accuracy +X.Xpp · wall/task +Y.Ys`), and — when tasks
 carry `class` labels — a per-class breakdown, so a subsystem's uplift and
 latency cost can be judged per task class instead of globally.

-## Pricing the run graph
-
-Delegation makes three separate claims, and one number cannot settle them. Each
-has its own arm, and all three read the same **Fan-out** section of the report —
-folded from the run graph, which is the only record of a scheduling wait:
-
-| Claim | Arm |
-| --- | --- |
-| Delegating at all beats one agent | `-ablate subagent` |
-| A `depends_on` edge is worth its payload, not only its order | `-ablate upstream` over `benchmarks/upstream-edge` |
-| Running members side by side beats running them in turn | `max_subagent_concurrency = 1` over `benchmarks/fanout-width` |
-
-```sh
-go run ./cmd/e2ebench -suite benchmarks/upstream-edge -json control.json
-go run ./cmd/e2ebench -suite benchmarks/upstream-edge -ablate upstream -json ablated.json
-go run ./cmd/e2ebench -mode compare control.json ablated.json
-```
-
-The section reports work against wall (the speed-up the shape bought), the
-critical path (the floor the declared dependencies impose — wall above it is
-scheduling loss, not shape), and slot wait (what the concurrency ceiling cost
-members that had nothing left to wait for).
-
-Read all three axes before calling an arm better: a fan-out that halves wall
-clock while doubling tokens wins one and loses another. The arm passes when its
-solve rate is no worse, its wall clock is strictly lower, and its tokens stay
-inside the budget the comparison declares.
-
 ## SWE-bench Verified mode

 `e2ebench` can also run the agent inside the official SWE-bench evaluation
@@ -527,7 +426,7 @@ writer.
 ## Adding a new task

 1. Create `benchmarks/e2e/tasks/<task-id>/`.
-2. Write `task.toml` with `prompt` and `timeout_sec` — no `max_steps` (see
+2. Write `task.toml` with `prompt`, `max_steps`, and `timeout_sec` (see
   [schema](#tasktoml-schema)).
 3. If the task needs seed files, add them under `workdir/` (they are copied
   into the temp run dir; symlinks are skipped).
@@ -542,32 +441,33 @@ writer.

 ## context-maintenance-e2e

-A cost-capped smoke for content-driven maintenance, in three legs over one state
-directory: seed a session just below the compaction trigger, load it back and
-check nothing is re-derived, then grow it past the trigger and check exactly one
-summary lands.
+This harness measures what happens when a long session goes idle past the
+provider's cache TTL and then resumes: it A/B-compares cold-restart miss tokens
+with and without pruning, and checks that the agent re-reads a file behind a
+prune placeholder instead of hallucinating.

-Seed and resume take no provider; `continue` samples and needs
-`DEEPSEEK_API_KEY`. The model is hardcoded to `deepseek-v4-flash` at
-`https://api.deepseek.com`.
+It is hardcoded to the `deepseek-v4-flash` model at `https://api.deepseek.com`
+and requires the `DEEPSEEK_API_KEY` environment variable.

 ```sh
-# Seed below the trigger: no checkpoint installed, no summary started
-go run ./benchmarks/context-maintenance-e2e -offline seed
+export DEEPSEEK_API_KEY=...

-# Reload: the projection version must not move and no summary may run
+# Seed both arms (pruned + control) with a large session and warm the cache
+go run ./benchmarks/context-maintenance-e2e seed
+
+# Wait past the provider's cache TTL, then resume: prune the "pruned" arm and
+# compare cold-restart miss tokens
 go run ./benchmarks/context-maintenance-e2e resume

-# Grow past the trigger: one summary, version +1, then no further summary
-export DEEPSEEK_API_KEY=...
-go run ./benchmarks/context-maintenance-e2e continue
+# Run the comprehension trials (agent must re-read a pruned file and answer
+# from it); exits non-zero unless every trial passes
+go run ./benchmarks/context-maintenance-e2e comprehension
 ```

 | Flag | Default | Purpose |
 | --- | --- | --- |
-| `-dir` | `benchmarks/context-maintenance-e2e/run` | State directory shared by the three legs (session + `meta.json`). |
-| `-max-usd` | `0.50` | Hard cost cap for the live legs. |
-| `-offline` | off | Skip the live provider; `seed` only. |
+| `-dir` | `benchmarks/context-maintenance-e2e/run` | State directory for `seed`/`resume` (sessions + `meta.json`, `resume-<ts>.json`). |
+| `-trials` | `5` | Number of comprehension trials. |

 ## See also

@@ -593,24 +493,6 @@ e2ebench -suite benchmarks/memorybench -budget 0 -policy memory-off -trajectorie
 e2ebench -mode compare on.json off.json     # Memory utility section
 ```

-MemoryBench requires an OS read sandbox. It runs on macOS and Linux and
-refuses to start on Windows, where the benchmark cannot isolate the committed
-`tasks/<id>/memory/` answers from a graded agent. Each run also audits every
-recorded trajectory leg for a successful read of those answers; an answer read
-invalidates the result and makes the suite command exit nonzero. A missing or
-unreadable trajectory marks only that run incomplete; other tasks still run.
-Trajectories are captured temporarily when `-trajectories` is omitted.
-
-The audit treats a completed `read_file` call naming a task-memory path as a
-read. For `bash` and `grep`, a task-memory path in the arguments plus an authored
-fact line in the result is **evidence** of a read, not a proof of how that text
-was obtained. `glob` and directory listings expose names but not fact bodies;
-the audit does not count them as answer reads. Commands that reach the answer
-through an indirect path, such as `cd` followed by a relative filename, may
-not be attributed by the trajectory check. The OS read sandbox is the primary
-boundary; this audit is a second check, not a proof that an unflagged run never
-accessed an answer.
-
 Utility delta = paired Pass(on) − Pass(off). Harmful attribution is paired,
 never judged: the same task passed without memory and failed with it while
 recall fired. Scenario classes: exact, paraphrase, cjk, symbol, distractor
@@ -624,8 +506,8 @@ update (revised value wins), pinned (prefix channel end to end).
 `benchmarks/compaction/` drives the real agent compaction path over a session
 that grows one generation at a time. Each generation appends a round of work
 and then folds, so generation N folds everything generations 1..N produced —
-which is the growth that matters, because a fold reads the model-visible
-projection, so the previous digest feeds the next one.
+which is the growth that matters, because a fold re-derives its digest from the
+whole canonical transcript rather than from the previous digest.

 ```bash
 go run ./benchmarks/compaction -mode=cost                     # offline, no API key
@@ -652,3 +534,16 @@ probe, not a compaction loss.
 Probe answers are scored on whole words, and a wanted answer does not count if a
 rejected one appears anywhere in the same reply — "yes, but it has not been
 re-run since" is the shape a drifting digest produces, and it is not a pass.
+
+### Fold arms
+
+`-arm=full` (default) re-derives every digest from the canonical transcript, so
+digests never chain. `-arm=incremental` folds the model-visible view instead,
+feeding the previous digest back through the summarizer. The arms exist to price
+that trade: run the cost arm for what chaining saves, and the fidelity arm for
+what it costs.
+
+```bash
+go run ./benchmarks/compaction -mode=cost -arm=incremental
+DEEPSEEK_API_KEY=… go run ./benchmarks/compaction -mode=fidelity -arm=incremental
+```
@@ -1,24 +0,0 @@
-# catalog-detector
-
-Who owns "the skills listing is owed again"?
-
-The listing is delivered once. Something has to decide when the model's copy
-stopped matching the registry, and there are three candidates: rebuild the
-projection and compare it, fingerprint the files' metadata, or trust that every
-write announced itself.
-
-```
-go run ./benchmarks/catalog-detector -real "$PWD"
-```
-
-**Owed is measured, not declared.** The probe renders the block before and after
-each transition; owed means those differ. That definition is what makes a
-touched mtime and a body-only edit *not* owed — neither changes what the model
-would be sent — and it is why the projection detector is the definition rather
-than an estimate of it. What the arms measure is where the two cheaper
-detectors diverge from it.
-
-Cost is timed without the instrument attached: files and bytes are counted in a
-separate walk, because timing a detector with the accounting inside it measures
-the accounting. Numbers are macOS/APFS with a warm cache; the ratio between
-detectors is what transfers, not the absolute.
@@ -1,87 +0,0 @@
-package main
-
-import (
-	"fmt"
-	"os"
-	"path/filepath"
-	"strings"
-	"time"
-)
-
-// arm is one transition a workspace can undergo, with what it did to the thing
-// the model would be sent. Truth is measured, not declared: the block is
-// rendered before and after, and owed means those two differ.
-type arm struct {
-	name    string
-	apply   func(e env) error
-	comment string
-}
-
-var arms = []arm{
-	{"api-create", func(e env) error { *e.apiWrites++; return writeSkill(e, "added", "a skill this process wrote") }, "a writer this process controls"},
-	{"api-describe", func(e env) error {
-		*e.apiWrites++
-		return writeSkill(e, "probe-000", "a description this process rewrote")
-	}, "same writer, different field"},
-	{"api-delete", func(e env) error { *e.apiWrites++; return os.RemoveAll(skillDir(e, "probe-000")) }, "same writer, removal"},
-	{"external-create", func(e env) error { return writeSkill(e, "outside", "a skill written past the API") }, "an editor, a script, the write tool"},
-	{"external-describe", func(e env) error { return writeSkill(e, "probe-000", "a description rewritten past the API") }, "the common out-of-band edit"},
-	{"external-delete", func(e env) error { return os.RemoveAll(skillDir(e, "probe-000")) }, "removal past the API"},
-	{"mtime-restored", restoreMtimeChange, "same length, older mtime: git checkout, rsync, touch -r"},
-	{"noop-touch", touchOnly, "mtime moved, nothing the model sees changed"},
-	{"body-only", bodyOnlyChange, "the body changed; the index does not render it"},
-}
-
-func skillDir(e env, name string) string {
-	return filepath.Join(e.projectRoot, ".reasonix", "skills", name)
-}
-
-func writeSkill(e env, name, description string) error {
-	dir := skillDir(e, name)
-	if err := os.MkdirAll(dir, 0o755); err != nil {
-		return err
-	}
-	return os.WriteFile(filepath.Join(dir, "SKILL.md"),
-		fmt.Appendf(nil, "---\nname: %s\ndescription: %s\n---\n\nBody paragraph the index never renders.\n", name, description), 0o644)
-}
-
-// restoreMtimeChange rewrites the rendered description without moving either
-// size or mtime. Nothing exotic produces this: a branch switch restores content
-// and stamps it with the mtime the file had.
-func restoreMtimeChange(e env) error {
-	path := filepath.Join(skillDir(e, "probe-000"), "SKILL.md")
-	before, err := os.Stat(path)
-	if err != nil {
-		return err
-	}
-	old, err := os.ReadFile(path)
-	if err != nil {
-		return err
-	}
-	swapped := strings.Replace(string(old), "probe skill 0, one line of index text.", "PROBE SKILL 0, ONE LINE OF INDEX TEXT.", 1)
-	if swapped == string(old) {
-		return fmt.Errorf("arm mtime-restored did not change the description; the fixture moved")
-	}
-	if len(swapped) != len(old) {
-		return fmt.Errorf("arm mtime-restored changed the file length (%d -> %d); it must not", len(old), len(swapped))
-	}
-	if err := os.WriteFile(path, []byte(swapped), 0o644); err != nil {
-		return err
-	}
-	return os.Chtimes(path, before.ModTime(), before.ModTime())
-}
-
-func touchOnly(e env) error {
-	path := filepath.Join(skillDir(e, "probe-000"), "SKILL.md")
-	later := time.Now().Add(time.Hour)
-	return os.Chtimes(path, later, later)
-}
-
-func bodyOnlyChange(e env) error {
-	path := filepath.Join(skillDir(e, "probe-000"), "SKILL.md")
-	old, err := os.ReadFile(path)
-	if err != nil {
-		return err
-	}
-	return os.WriteFile(path, []byte(strings.Replace(string(old), "Body paragraph the index never renders.", "A different body paragraph, longer than the one it replaces.", 1)), 0o644)
-}
@@ -1,90 +0,0 @@
-package main
-
-import (
-	"fmt"
-	"os"
-	"path/filepath"
-	"strings"
-	"time"
-)
-
-type costRow struct {
-	scale    string
-	detector string
-	p50, p95 time.Duration
-	files    int
-	bytes    int64
-}
-
-// measureCost times each detector over synthetic workspaces of a given size and,
-// when asked, over a real one. The synthetic skills are shaped like real ones —
-// a directory per skill holding SKILL.md — because the walk is what is being
-// measured and a flat directory would not have it.
-func measureCost(scales []int, samples int, realRoot string) []costRow {
-	var rows []costRow
-	for _, n := range scales {
-		root, err := os.MkdirTemp("", "catalog-detector-*")
-		fatal(err)
-		defer os.RemoveAll(root)
-		fatal(writeSkills(filepath.Join(root, "project", ".reasonix", "skills"), n))
-		// A separate home: pointing both at one directory walks it twice and
-		// reports double the cost a real workspace pays.
-		rows = append(rows, timeDetectors(fmt.Sprintf("synthetic n=%d", n),
-			env{projectRoot: filepath.Join(root, "project"), homeDir: filepath.Join(root, "home"), apiWrites: new(int)}, samples)...)
-	}
-	if realRoot != "" {
-		home, err := os.UserHomeDir()
-		fatal(err)
-		rows = append(rows, timeDetectors("real workspace", env{projectRoot: realRoot, homeDir: home, apiWrites: new(int)}, samples)...)
-	}
-	return rows
-}
-
-func timeDetectors(scale string, e env, samples int) []costRow {
-	var rows []costRow
-	files, bytes := walkCost(e.roots())
-	for _, d := range detectors {
-		var durations []time.Duration
-		for range samples {
-			start := time.Now()
-			d.read(e)
-			durations = append(durations, time.Since(start))
-		}
-		row := costRow{scale: scale, detector: d.name, p50: percentile(durations, 0.5), p95: percentile(durations, 0.95)}
-		if d.walks {
-			row.files, row.bytes = files, bytes
-		}
-		rows = append(rows, row)
-	}
-	return rows
-}
-
-func writeSkills(dir string, n int) error {
-	if err := os.MkdirAll(dir, 0o755); err != nil {
-		return err
-	}
-	for i := range n {
-		sub := filepath.Join(dir, fmt.Sprintf("probe-%03d", i))
-		if err := os.MkdirAll(sub, 0o755); err != nil {
-			return err
-		}
-		if err := os.WriteFile(filepath.Join(sub, "SKILL.md"), []byte(skillFile(i, fmt.Sprintf("probe skill %d, one line of index text.", i))), 0o644); err != nil {
-			return err
-		}
-	}
-	return nil
-}
-
-func skillFile(i int, description string) string {
-	return fmt.Sprintf("---\nname: probe-%03d\ndescription: %s\n---\n\nBody paragraph the index never renders.\n", i, description)
-}
-
-func renderCost(rows []costRow) string {
-	var b strings.Builder
-	b.WriteString("cost — one reading, as a turn would take it\n\n")
-	fmt.Fprintf(&b, "%-16s %-15s %10s %10s %7s %9s\n", "scale", "detector", "p50", "p95", "files", "bytes")
-	for _, r := range rows {
-		fmt.Fprintf(&b, "%-16s %-15s %10s %10s %7d %9d\n", r.scale, r.detector, r.p50.Round(time.Microsecond), r.p95.Round(time.Microsecond), r.files, r.bytes)
-	}
-	return b.String()
-}
@@ -1,119 +0,0 @@
-package main
-
-import (
-	"crypto/sha256"
-	"encoding/hex"
-	"fmt"
-	"io"
-	"io/fs"
-	"os"
-	"path/filepath"
-	"sort"
-	"strings"
-
-	"reasonix/internal/ext/skill"
-)
-
-// detector answers one question: has the thing the model would be sent changed?
-// Each returns an opaque token — equal tokens mean "not owed". Nothing here
-// counts files or bytes: accounting is a second walk, and timing a detector
-// with the instrument attached measures the instrument.
-type detector struct {
-	name  string
-	read  func(env) string
-	walks bool // whether this detector's cost is a directory walk
-}
-
-// env is one workspace under measurement.
-type env struct {
-	projectRoot string
-	homeDir     string
-	// apiWrites is the observed-write detector's whole input: whether the last
-	// mutation went through a writer this process controls.
-	apiWrites *int
-}
-
-func (e env) store() *skill.Store {
-	return skill.New(skill.Options{
-		ProjectRoot: e.projectRoot, ReasonixHomeDir: filepath.Join(e.homeDir, ".reasonix"), Stderr: io.Discard,
-	})
-}
-
-func (e env) roots() []string {
-	var out []string
-	for _, r := range e.store().Roots() {
-		out = append(out, r.Dir)
-	}
-	return out
-}
-
-// projectionDetector is the truth, not an estimate of it: it renders the block
-// the model would receive and compares that. A change no listing shows — a body
-// edit, a touched mtime — is by construction not owed.
-var projectionDetector = detector{
-	name:  "projection",
-	walks: true,
-	read: func(e env) string {
-		return digest(skill.IndexBlock(e.store().List()))
-	},
-}
-
-// metadataDetector is the cheap approximation: path, size and mtime, no reads.
-var metadataDetector = detector{
-	name:  "metadata",
-	walks: true,
-	read: func(e env) string {
-		var lines []string
-		for _, dir := range e.roots() {
-			_ = filepath.WalkDir(dir, func(p string, d fs.DirEntry, err error) error {
-				if err != nil || d.IsDir() {
-					return nil
-				}
-				info, err := d.Info()
-				if err != nil {
-					return nil
-				}
-				lines = append(lines, fmt.Sprintf("%s\x00%d\x00%d", p, info.Size(), info.ModTime().UnixNano()))
-				return nil
-			})
-		}
-		sort.Strings(lines)
-		return digest(strings.Join(lines, "\n"))
-	},
-}
-
-// observedWriteDetector sees only what this process was told about. It reads
-// nothing, so its cost is zero and its blindness is total: a write it was not
-// told about leaves the token where it was.
-var observedWriteDetector = detector{
-	name: "observed-write",
-	read: func(e env) string {
-		return fmt.Sprintf("writes=%d", *e.apiWrites)
-	},
-}
-
-var detectors = []detector{projectionDetector, metadataDetector, observedWriteDetector}
-
-// walkCost reports the shape of what a walking detector touches. It runs
-// outside the timed section for exactly the reason this file has two functions
-// instead of one.
-func walkCost(dirs []string) (files int, bytes int64) {
-	for _, dir := range dirs {
-		_ = filepath.WalkDir(dir, func(p string, d fs.DirEntry, err error) error {
-			if err != nil || d.IsDir() {
-				return nil
-			}
-			files++
-			if info, err := os.Stat(p); err == nil {
-				bytes += info.Size()
-			}
-			return nil
-		})
-	}
-	return files, bytes
-}
-
-func digest(s string) string {
-	sum := sha256.Sum256([]byte(s))
-	return hex.EncodeToString(sum[:])[:12]
-}
@@ -1,77 +0,0 @@
-// Which detector can own "the skills listing is owed again"? The answer decides
-// whether a projection debt is a flag every writer must remember to set, or a
-// question the host answers for itself. Owed means one thing here: the rendered
-// available-skills block differs from the one the model last received, so a
-// touched mtime and a body edit are not owed. This probe writes only inside its
-// own temporary workspaces.
-package main
-
-import (
-	"flag"
-	"fmt"
-	"os"
-	"slices"
-	"sort"
-	"time"
-)
-
-func main() {
-	scales := flag.String("scales", "0,10,50,200", "synthetic skill counts to measure")
-	samples := flag.Int("samples", 51, "cost samples per detector per scale")
-	real := flag.String("real", "", "also measure this workspace as it is on disk")
-	flag.Parse()
-
-	costs := measureCost(parseScales(*scales), *samples, *real)
-	verdicts := measureVerdicts()
-
-	fmt.Println(renderCost(costs))
-	fmt.Println()
-	fmt.Println(renderVerdicts(verdicts))
-	if failed(verdicts) {
-		// Not an error: a detector that misses is the finding, and the exit
-		// code stays zero so a study can report one.
-		fmt.Println("\nA detector with a miss cannot own correctness; one with a false positive cannot own cache stability.")
-	}
-}
-
-func parseScales(s string) []int {
-	var out []int
-	for _, f := range splitComma(s) {
-		var n int
-		if _, err := fmt.Sscanf(f, "%d", &n); err == nil {
-			out = append(out, n)
-		}
-	}
-	sort.Ints(out)
-	return out
-}
-
-func splitComma(s string) []string {
-	var out []string
-	cur := ""
-	for _, r := range s {
-		if r == ',' {
-			out = append(out, cur)
-			cur = ""
-			continue
-		}
-		cur += string(r)
-	}
-	return append(out, cur)
-}
-
-func percentile(samples []time.Duration, p float64) time.Duration {
-	if len(samples) == 0 {
-		return 0
-	}
-	slices.Sort(samples)
-	idx := int(float64(len(samples)-1) * p)
-	return samples[idx]
-}
-
-func fatal(err error) {
-	if err != nil {
-		fmt.Fprintln(os.Stderr, "catalog-detector:", err)
-		os.Exit(1)
-	}
-}
@@ -1,77 +0,0 @@
-package main
-
-import (
-	"fmt"
-	"os"
-	"path/filepath"
-	"strings"
-
-	"reasonix/internal/ext/skill"
-)
-
-type verdict struct {
-	arm      string
-	comment  string
-	owed     bool
-	detector string
-	changed  bool
-}
-
-func (v verdict) label() string {
-	switch {
-	case v.owed == v.changed:
-		return "correct"
-	case v.owed:
-		return "MISS"
-	default:
-		return "false-positive"
-	}
-}
-
-// measureVerdicts runs every arm against a fresh workspace and records, for each
-// detector, whether it saw the transition. Truth is the rendered block itself:
-// the arm is owed when the listing the model would receive is not the listing it
-// already has.
-func measureVerdicts() []verdict {
-	var out []verdict
-	for _, a := range arms {
-		root, err := os.MkdirTemp("", "catalog-detector-arm-*")
-		fatal(err)
-		defer os.RemoveAll(root)
-		fatal(writeSkills(filepath.Join(root, "project", ".reasonix", "skills"), 3))
-		e := env{projectRoot: filepath.Join(root, "project"), homeDir: filepath.Join(root, "home"), apiWrites: new(int)}
-
-		before := map[string]string{}
-		for _, d := range detectors {
-			before[d.name] = d.read(e)
-		}
-		blockBefore := skill.IndexBlock(e.store().List())
-
-		fatal(a.apply(e))
-
-		owed := skill.IndexBlock(e.store().List()) != blockBefore
-		for _, d := range detectors {
-			out = append(out, verdict{a.name, a.comment, owed, d.name, d.read(e) != before[d.name]})
-		}
-	}
-	return out
-}
-
-func renderVerdicts(vs []verdict) string {
-	var b strings.Builder
-	b.WriteString("verdicts — owed is measured, not declared: the rendered block differs\n\n")
-	fmt.Fprintf(&b, "%-18s %-6s %-15s %-15s %s\n", "arm", "owed", "detector", "verdict", "what the arm is")
-	for _, v := range vs {
-		fmt.Fprintf(&b, "%-18s %-6v %-15s %-15s %s\n", v.arm, v.owed, v.detector, v.label(), v.comment)
-	}
-	return b.String()
-}
-
-func failed(vs []verdict) bool {
-	for _, v := range vs {
-		if v.label() != "correct" {
-			return true
-		}
-	}
-	return false
-}
@@ -15,16 +15,16 @@ import (
 	"fmt"
 	"os"
 	"path/filepath"
-	"reasonix/internal/state/sessionstore"
 	"strings"
 	"time"
+	"unicode/utf8"

-	"reasonix/internal/base/tokencount"
-	"reasonix/internal/contract/event"
-	"reasonix/internal/contract/provider"
-	"reasonix/internal/contract/tool"
-	_ "reasonix/internal/model/openai"
-	"reasonix/internal/runtime/agent"
+	"reasonix/internal/ablation"
+	"reasonix/internal/agent"
+	"reasonix/internal/event"
+	"reasonix/internal/provider"
+	_ "reasonix/internal/provider/openai"
+	"reasonix/internal/tool"
 )

 const (
@@ -41,7 +41,8 @@ func main() {
 	report := flag.String("report", "1,2,4,8", "generations to report on")
 	window := flag.Int("window", 128_000, "context window in tokens")
 	control := flag.Bool("control", true, "fidelity: also score probes against full history")
-	_ = flag.Bool("snip", false, "accepted and ignored: automatic snip projections are gone")
+	arm := flag.String("arm", "full", "full | incremental: re-derive each digest from canonical, or fold the previous projection")
+	snip := flag.Bool("snip", false, "legacy no-op: automatic snip projections are gone; kept so old scripts do not fail")
 	out := flag.String("out", "", "write the JSON report here")
 	flag.Parse()

@@ -49,11 +50,14 @@ func main() {
 		res []genResult
 		err error
 	)
-	switch *mode {
-	case "cost":
-		res, err = runCost(*gens, *window)
-	case "fidelity":
-		res, err = runFidelity(*gens, *window, *control)
+	a := arms{incremental: *arm == "incremental", snip: *snip}
+	switch {
+	case *arm != "full" && *arm != "incremental":
+		err = fmt.Errorf("unknown arm %q", *arm)
+	case *mode == "cost":
+		res, err = runCost(*gens, *window, a)
+	case *mode == "fidelity":
+		res, err = runFidelity(*gens, *window, *control, a)
 	default:
 		err = fmt.Errorf("unknown mode %q", *mode)
 	}
@@ -61,9 +65,9 @@ func main() {
 		fmt.Fprintln(os.Stderr, err)
 		os.Exit(1)
 	}
-	printReport(*mode, res, reportAt(*report))
+	printReport(*mode+" / "+*arm, res, reportAt(*report))
 	if *out != "" {
-		b, _ := json.MarshalIndent(map[string]any{"mode": *mode, "window": *window, "generations": res}, "", "  ")
+		b, _ := json.MarshalIndent(map[string]any{"mode": *mode, "arm": *arm, "window": *window, "generations": res}, "", "  ")
 		if werr := os.WriteFile(*out, append(b, '\n'), 0o644); werr != nil {
 			fmt.Fprintln(os.Stderr, werr)
 			os.Exit(1)
@@ -80,6 +84,8 @@ type genResult struct {
 	SummarizerInput  int            `json:"summarizer_input_tokens"`
 	LargestCall      int            `json:"largest_call_tokens"`
 	Mode             string         `json:"mode,omitempty"`
+	SnippedResults   int            `json:"snipped_results,omitempty"`
+	SnippedChars     int            `json:"snipped_chars,omitempty"`
 	Seconds          float64        `json:"seconds"`
 	Error            string         `json:"error,omitempty"`
 	Survived         map[string]int `json:"survived,omitempty"`   // probe class -> 1 kept, 0 lost
@@ -90,13 +96,14 @@ type genResult struct {
 }

 type harness struct {
-	sess   *sessionstore.Session
+	sess   *agent.Session
 	agentA *agent.Agent
 	path   string
 	calls  *callRecorder
+	snip   bool
 }

-func newHarness(t *testingDir, p provider.Provider, window int, rec *callRecorder) *harness {
+func newHarness(t *testingDir, p provider.Provider, window int, rec *callRecorder, arm arms) *harness {
 	sess := newSession()
 	path := filepath.Join(t.dir, "session.jsonl")
 	a := agent.New(p, tool.NewRegistry(), sess, agent.Options{
@@ -107,18 +114,44 @@ func newHarness(t *testingDir, p provider.Provider, window int, rec *callRecorde
 		// boot's default when cfg.Agent.Keep is unset; without it the bench
 		// would measure a configuration no real session runs.
 		KeepPolicy: agent.KeepErrors,
+		Ablation:   foldArm(arm.incremental),
 	}, rec.sink())
-	return &harness{sess: sess, agentA: a, path: path, calls: rec}
+	return &harness{sess: sess, agentA: a, path: path, calls: rec, snip: arm.snip}
+}
+
+// foldArm switches full re-derivation off, which is what makes a fold read the
+// previous projection instead of the canonical transcript.
+// arms selects the maintenance behaviour under test. Snipping is off by default
+// so a run stays comparable with baselines recorded before it existed.
+type arms struct {
+	incremental bool
+	snip        bool
+}
+
+func foldArm(incremental bool) ablation.Set {
+	if incremental {
+		return ablation.New(ablation.FullFold)
+	}
+	return ablation.Set{}
 }

 // runGeneration grows the session and folds it, returning what that fold cost.
 func (h *harness) runGeneration(ctx context.Context, gen int, probes []probe) genResult {
 	growSession(h.sess, gen, probes)
-	r := genResult{Gen: gen, CanonicalTokens: tokencount.Text(renderAll(h.sess.Snapshot()))}
+	r := genResult{Gen: gen, CanonicalTokens: estimateTokens(renderAll(h.sess.Snapshot()))}

 	h.calls.reset()
 	start := time.Now()
-	_, err := h.agentA.CompactNow(ctx, agent.CompactRequest{})
+	if h.snip {
+		// SnipStaleToolResults is intentionally a no-op; record zeros for
+		// report schema compatibility with pre-content-driven baselines.
+		st, serr := h.agentA.SnipStaleToolResults()
+		if serr != nil {
+			r.Error = serr.Error()
+		}
+		r.SnippedResults, r.SnippedChars = st.Results, st.SavedChars
+	}
+	err := h.agentA.CompactNow(ctx, "")
 	r.Seconds = time.Since(start).Seconds()
 	if err != nil {
 		r.Error = err.Error()
@@ -128,7 +161,7 @@ func (h *harness) runGeneration(ctx context.Context, gen int, probes []probe) ge
 		r.SummarizerInput += c.tokens
 		r.LargestCall = max(r.LargestCall, c.tokens)
 	}
-	if st, ok, sterr := sessionstore.LoadCompactionState(h.path); sterr == nil && ok {
+	if st, ok, sterr := agent.LoadCompactionState(h.path); sterr == nil && ok {
 		r.ProjectionTokens = st.Projection.ProjectionTokens
 		if st.LastReceipt != nil && st.LastReceipt.Action == "summary" {
 			r.Mode = agent.CompactionModeSummarized
@@ -139,7 +172,7 @@ func (h *harness) runGeneration(ctx context.Context, gen int, probes []probe) ge
 	return r
 }

-func runCost(gens, window int) ([]genResult, error) {
+func runCost(gens, window int, a arms) ([]genResult, error) {
 	dir, cleanup, err := tempDir()
 	if err != nil {
 		return nil, err
@@ -148,7 +181,7 @@ func runCost(gens, window int) ([]genResult, error) {

 	rec := &callRecorder{}
 	p := &scriptedProvider{rec: rec, reply: syntheticDigest, window: window}
-	h := newHarness(dir, p, window, rec)
+	h := newHarness(dir, p, window, rec, a)

 	var out []genResult
 	for gen := range gens {
@@ -157,7 +190,7 @@ func runCost(gens, window int) ([]genResult, error) {
 	return out, nil
 }

-func runFidelity(gens, window int, control bool) ([]genResult, error) {
+func runFidelity(gens, window int, control bool, a arms) ([]genResult, error) {
 	key := os.Getenv("DEEPSEEK_API_KEY")
 	if key == "" {
 		return nil, fmt.Errorf("fidelity mode needs DEEPSEEK_API_KEY")
@@ -173,7 +206,7 @@ func runFidelity(gens, window int, control bool) ([]genResult, error) {
 	defer cleanup()

 	rec := &callRecorder{}
-	h := newHarness(dir, &recordingProvider{inner: p, rec: rec}, window, rec)
+	h := newHarness(dir, &recordingProvider{inner: p, rec: rec}, window, rec, a)
 	probes := probeSuite()

 	ctx := context.Background()
@@ -357,6 +390,18 @@ func reportAt(spec string) map[int]bool {
 	return at
 }

+// estimateTokens mirrors the kernel's own estimator so bench numbers and
+// compaction telemetry are read in the same unit.
+func estimateTokens(s string) int {
+	if s == "" {
+		return 0
+	}
+	if runes := utf8.RuneCountInString(s); runes > (len(s)+3)/4 {
+		return runes
+	}
+	return (len(s) + 3) / 4
+}
+
 func renderAll(msgs []provider.Message) string {
 	var b strings.Builder
 	for _, m := range msgs {
@@ -407,7 +452,7 @@ func (r *callRecorder) reset() { r.calls = nil }
 func (r *callRecorder) note(req provider.Request) {
 	c := recordedCall{}
 	for _, m := range req.Messages {
-		c.tokens += tokencount.Text(m.Content)
+		c.tokens += estimateTokens(m.Content)
 		if m.Role == provider.RoleSystem {
 			c.system = m.Content
 		}
@@ -12,7 +12,7 @@ func TestRepeatedFoldsStayBoundedAndKeepSucceeding(t *testing.T) {
 		reserve  = 8192 // summaryOutputReserve: the digest the call must still return
 		maxCalls = 7    // maxSummarySpans + the merge pass
 	)
-	res, err := runCost(gens, window)
+	res, err := runCost(gens, window, arms{})
 	if err != nil {
 		t.Fatalf("runCost: %v", err)
 	}
@@ -2,13 +2,13 @@ package main

 import (
 	"fmt"
-	"reasonix/internal/state/sessionstore"
 	"slices"
 	"strings"
 	"unicode"

-	"reasonix/internal/contract/provider"
-	"reasonix/internal/contract/tool"
+	"reasonix/internal/agent"
+	"reasonix/internal/provider"
+	"reasonix/internal/tool"
 )

 // probeAnswerContract rides with every probe question. A context full of tool
@@ -36,22 +36,22 @@ func invalidAnswer(s string) bool {
 type probe struct {
 	class   string
 	plantAt int
-	plant   func(*sessionstore.Session)
+	plant   func(*agent.Session)
 	// later plants more of the same fact in later generations, so a decision
 	// that changes across folds is tested the way it actually happens: each
 	// revision lands on the far side of a compaction, not next to the last one.
-	later    map[int]func(*sessionstore.Session)
+	later    map[int]func(*agent.Session)
 	question string
 	want     []string // answer must contain one of these, lowercased
 	reject   []string // ...and none of these: the pre-correction answer
 }

-func userTurn(text string) func(*sessionstore.Session) {
-	return func(s *sessionstore.Session) { s.Add(provider.Message{Role: provider.RoleUser, Content: text}) }
+func userTurn(text string) func(*agent.Session) {
+	return func(s *agent.Session) { s.Add(provider.Message{Role: provider.RoleUser, Content: text}) }
 }

-func toolRound(id, name, args, result string) func(*sessionstore.Session) {
-	return func(s *sessionstore.Session) {
+func toolRound(id, name, args, result string) func(*agent.Session) {
+	return func(s *agent.Session) {
 		s.Add(provider.Message{Role: provider.RoleAssistant, ToolCalls: []provider.ToolCall{{ID: id, Name: name, Arguments: args}}})
 		s.Add(provider.Message{Role: provider.RoleTool, ToolCallID: id, Name: name, Content: result})
 	}
@@ -60,8 +60,8 @@ func toolRound(id, name, args, result string) func(*sessionstore.Session) {
 // failedToolRound is a tool round the host recorded as failed, the way a real
 // non-zero bash run arrives. Without the execution record the keep policy sees
 // only text, which is exactly the gap the buried-evidence probe measures.
-func failedToolRound(id, name, args, result string, code int) func(*sessionstore.Session) {
-	return func(s *sessionstore.Session) {
+func failedToolRound(id, name, args, result string, code int) func(*agent.Session) {
+	return func(s *agent.Session) {
 		exit := code
 		s.Add(provider.Message{Role: provider.RoleAssistant, ToolCalls: []provider.ToolCall{{ID: id, Name: name, Arguments: args}}})
 		s.Add(provider.Message{
@@ -75,8 +75,8 @@ func failedToolRound(id, name, args, result string, code int) func(*sessionstore
 	}
 }

-func seq(fns ...func(*sessionstore.Session)) func(*sessionstore.Session) {
-	return func(s *sessionstore.Session) {
+func seq(fns ...func(*agent.Session)) func(*agent.Session) {
+	return func(s *agent.Session) {
 		for _, fn := range fns {
 			fn(s)
 		}
@@ -240,8 +240,8 @@ func revisionProbes() []probe {
 		{
 			class:    "late-constraint",
 			plantAt:  3,
-			plant:    userTurn("New hard rule from here on: never edit anything under internal/state/store/."),
-			question: "Are you allowed to edit files under internal/state/store/? Answer yes or no.",
+			plant:    userTurn("New hard rule from here on: never edit anything under internal/store/."),
+			question: "Are you allowed to edit files under internal/store/? Answer yes or no.",
 			want:     []string{"no"},
 			reject:   []string{"yes"},
 		},
@@ -249,7 +249,7 @@ func revisionProbes() []probe {
 			class:   "reversal-chain",
 			plantAt: 0,
 			plant:   userTurn("Use PostgreSQL for the datastore."),
-			later: map[int]func(*sessionstore.Session){
+			later: map[int]func(*agent.Session){
 				1: userTurn("Change of plan: drop PostgreSQL, use SQLite instead."),
 				2: userTurn("Final call: back to PostgreSQL after all."),
 			},
@@ -261,7 +261,7 @@ func revisionProbes() []probe {
 			class:   "distractor-file",
 			plantAt: 1,
 			plant:   userTurn("Draft note: the fix might belong in config/legacy_parser.go."),
-			later: map[int]func(*sessionstore.Session){
+			later: map[int]func(*agent.Session){
 				3: userTurn("Confirmed: the fix belongs in config/emitter.go; the legacy parser is not involved."),
 			},
 			question: "Which file does the fix belong in? Reply with just the filename, no path.",
@@ -2,10 +2,10 @@ package main

 import (
 	"fmt"
-	"reasonix/internal/state/sessionstore"
 	"strings"

-	"reasonix/internal/contract/provider"
+	"reasonix/internal/agent"
+	"reasonix/internal/provider"
 )

 // A generation is one round of work followed by one compaction. Compaction
@@ -19,7 +19,7 @@ const (

 // workUnit is one read_file round: the shape that dominates a real coding
 // session's token weight.
-func workUnit(sess *sessionstore.Session, gen, i int) {
+func workUnit(sess *agent.Session, gen, i int) {
 	id := fmt.Sprintf("g%02dc%02d", gen, i)
 	path := fmt.Sprintf("internal/pkg%02d/file%02d.go", gen, i)
 	sess.Add(provider.Message{Role: provider.RoleAssistant, ToolCalls: []provider.ToolCall{
@@ -41,7 +41,7 @@ func fakeGoFile(gen, i, size int) string {
 // growSession appends one generation of work, planting whatever probes belong
 // to that generation so a fact's age is exactly the number of folds it has
 // survived.
-func growSession(sess *sessionstore.Session, gen int, probes []probe) {
+func growSession(sess *agent.Session, gen int, probes []probe) {
 	for _, p := range probes {
 		if p.plantAt == gen {
 			p.plant(sess)
@@ -55,16 +55,16 @@ func growSession(sess *sessionstore.Session, gen int, probes []probe) {
 	}
 }

-func newSession() *sessionstore.Session {
-	sess := sessionstore.NewSession("You are a terse coding agent. Answer questions from what you know; do not guess.")
+func newSession() *agent.Session {
+	sess := agent.NewSession("You are a terse coding agent. Answer questions from what you know; do not guess.")
 	sess.Add(provider.Message{Role: provider.RoleUser, Content: "Fix the config round-trip formatting bug in this repo."})
 	return sess
 }

 // visibleContext splices the stored projection with the canonical messages
 // appended after it — the same model-visible view the agent would send.
-func visibleContext(sessionPath string, sess *sessionstore.Session) ([]provider.Message, error) {
-	st, ok, err := sessionstore.LoadCompactionState(sessionPath)
+func visibleContext(sessionPath string, sess *agent.Session) ([]provider.Message, error) {
+	st, ok, err := agent.LoadCompactionState(sessionPath)
 	if err != nil {
 		return nil, err
 	}
@@ -0,0 +1,41 @@
+package main
+
+import (
+	"strings"
+	"testing"
+)
+
+// Automatic snip projections are gone. SnipStaleToolResults remains as a no-op
+// so older harness flags and call sites do not panic or rewrite history.
+func TestSnipIsNoOp(t *testing.T) {
+	dir, cleanup, err := tempDir()
+	if err != nil {
+		t.Fatal(err)
+	}
+	defer cleanup()
+
+	rec := &callRecorder{}
+	const window = 64_000
+	h := newHarness(dir, &scriptedProvider{rec: rec, reply: syntheticDigest, window: window}, window, rec, arms{snip: true})
+	for gen := range 3 {
+		growSession(h.sess, gen, probeSuite())
+	}
+	before := renderAll(h.sess.Snapshot())
+	if !strings.Contains(before, "beta-7d21") {
+		t.Fatal("planted fact missing before the snip call")
+	}
+	st, err := h.agentA.SnipStaleToolResults()
+	if err != nil {
+		t.Fatal(err)
+	}
+	if st.Results != 0 || st.SavedChars != 0 {
+		t.Fatalf("snip no-op rewrote history: results=%d chars=%d", st.Results, st.SavedChars)
+	}
+	after := renderAll(h.sess.Snapshot())
+	if after != before {
+		t.Fatal("canonical transcript changed after SnipStaleToolResults")
+	}
+	if !strings.Contains(after, "beta-7d21") {
+		t.Error("planted fact missing after snip no-op")
+	}
+}
@@ -1,6 +1,6 @@
 // Cost-capped seed → resume → continue smoke for content-driven maintenance.
-// Seed and resume run without a provider; live continue needs DEEPSEEK_API_KEY
-// and starts exactly one summary after growing past compact_ratio=0.85.
+// Offline covers seed/resume; live continue needs DEEPSEEK_API_KEY and allows
+// at most one summary after growing past compact_ratio=0.85.
 package main

 import (
@@ -10,15 +10,14 @@ import (
 	"fmt"
 	"os"
 	"path/filepath"
-	"reasonix/internal/state/sessionstore"
 	"strings"
 	"time"

-	"reasonix/internal/contract/event"
-	"reasonix/internal/contract/provider"
-	"reasonix/internal/contract/tool"
-	_ "reasonix/internal/model/openai"
-	"reasonix/internal/runtime/agent"
+	"reasonix/internal/agent"
+	"reasonix/internal/event"
+	"reasonix/internal/provider"
+	_ "reasonix/internal/provider/openai"
+	"reasonix/internal/tool"
 )

 const (
@@ -70,7 +69,7 @@ func (s *recordSink) Emit(e event.Event) {
 func prov() (provider.Provider, *provider.Pricing) {
 	key := os.Getenv("DEEPSEEK_API_KEY")
 	if key == "" {
-		fmt.Fprintln(os.Stderr, "DEEPSEEK_API_KEY not set (use -offline for seed)")
+		fmt.Fprintln(os.Stderr, "DEEPSEEK_API_KEY not set (use -offline for seed/resume)")
 		os.Exit(1)
 	}
 	p, err := provider.New("openai", provider.Config{
@@ -104,7 +103,7 @@ func fatHistory(n int) []provider.Message {
 	)
 }

-func newAgent(p provider.Provider, sess *sessionstore.Session, path string, sink event.Sink) *agent.Agent {
+func newAgent(p provider.Provider, sess *agent.Session, path string, sink event.Sink) *agent.Agent {
 	return agent.New(p, tool.NewRegistry(), sess, agent.Options{
 		ContextWindow: windowTokens,
 		CompactRatio:  compactRatio,
@@ -145,7 +144,7 @@ func seed(dir string, offline bool, cap *costCap) {
 	}
 	path := filepath.Join(dir, "session.jsonl")
 	// ~80×32KB tool results: well under 85% of 1M for a cold estimate.
-	sess := &sessionstore.Session{Messages: fatHistory(80)}
+	sess := &agent.Session{Messages: fatHistory(80)}
 	sink := &recordSink{}
 	var p provider.Provider
 	if !offline {
@@ -193,19 +192,20 @@ func seed(dir string, offline bool, cap *costCap) {
 		m.ProjectedTokens, m.TriggerTokens, m.ProjectionVersion, m.SummaryCalls, m.SpentUSD)
 }

-// resume reads the seeded session back and checks that loading it neither
-// advances the projection version nor pays for a summary. It never samples, so
-// it takes no provider and needs no API key.
-func resume(dir string) {
+func resume(dir string, offline bool) {
 	path := filepath.Join(dir, "session.jsonl")
 	m := readMeta(dir)
-	sess, err := sessionstore.LoadSession(path)
+	sess, err := agent.LoadSession(path)
 	if err != nil {
 		fmt.Fprintln(os.Stderr, err)
 		os.Exit(1)
 	}
 	sink := &recordSink{}
-	a := newAgent(nil, sess, path, sink)
+	var p provider.Provider
+	if !offline {
+		p, _ = prov()
+	}
+	a := newAgent(p, sess, path, sink)
 	snap := a.ContextMaintenanceSnapshot()
 	if snap.ProjectionVersion != m.ProjectionVersion {
 		fmt.Fprintf(os.Stderr, "resume version = %d, want %d\n", snap.ProjectionVersion, m.ProjectionVersion)
@@ -235,7 +235,7 @@ func cont(dir string, offline bool, cap *costCap) {
 	}
 	path := filepath.Join(dir, "session.jsonl")
 	m := readMeta(dir)
-	sess, err := sessionstore.LoadSession(path)
+	sess, err := agent.LoadSession(path)
 	if err != nil {
 		fmt.Fprintln(os.Stderr, err)
 		os.Exit(1)
@@ -257,14 +257,12 @@ func cont(dir string, offline bool, cap *costCap) {
 		fmt.Fprintln(os.Stderr, "continue prepare:", err)
 		os.Exit(1)
 	}
-	// The fixture is already above the trigger it was grown past, so a leg that
-	// starts no summary has not observed the fold it exists to test.
-	if sink.summaryStarts != 1 {
-		fmt.Fprintf(os.Stderr, "continue started %d summaries, want exactly 1\n", sink.summaryStarts)
+	if sink.summaryStarts > 1 {
+		fmt.Fprintf(os.Stderr, "continue started %d summaries, want ≤1\n", sink.summaryStarts)
 		os.Exit(1)
 	}
 	after := a.ContextMaintenanceSnapshot()
-	if after.ProjectionVersion != m.ProjectionVersion+1 {
+	if sink.summaryStarts == 1 && after.ProjectionVersion != m.ProjectionVersion+1 {
 		fmt.Fprintf(os.Stderr, "continue version = %d, want %d after one summary\n",
 			after.ProjectionVersion, m.ProjectionVersion+1)
 		os.Exit(1)
@@ -307,14 +305,14 @@ func cont(dir string, offline bool, cap *costCap) {
 func main() {
 	dir := flag.String("dir", "benchmarks/context-maintenance-e2e/run", "state directory")
 	maxUSD := flag.Float64("max-usd", defaultMaxUSD, "hard cost cap for live API legs")
-	offline := flag.Bool("offline", false, "skip live provider (seed only)")
+	offline := flag.Bool("offline", false, "skip live provider (seed/resume only)")
 	flag.Parse()
 	cap := &costCap{maxUSD: *maxUSD}
 	switch flag.Arg(0) {
 	case "seed":
 		seed(*dir, *offline, cap)
 	case "resume":
-		resume(*dir)
+		resume(*dir, *offline)
 	case "continue":
 		cont(*dir, *offline, cap)
 	default:
@@ -1,98 +0,0 @@
-# Context retrieval bench
-
-Two questions about long-horizon memory, and the harness that makes their
-answers trustworthy.
-
-    -mode=preflight     build all 12 tasks under every arm and check the contract
-    -mode=snippet       does a rank-1 hit hand over the whole answer?
-    -mode=adversarial   send a model after the answer through every host surface
-    -mode=run-search    Experiment R: is searchable canonical a real capability?
-    -mode=run-boundary  Experiment I, paired: 6 tasks across their own cue boundary
-    -mode=run-index     Experiment I, full: 6 tasks x 4 index budgets
-    -mode=calibrate     find each index task's PlantAfterGen for its cue tier
-
-`-dry` drives any run mode with a scripted provider, for nothing. Real runs need
-`DEEPSEEK_API_KEY`.
-
-## Status
-
-**Searchable canonical recall: proven.** Six of six recovered with search on,
-three of six with it off, and the successes without it came from enumerating
-addresses by hand at roughly nine times the page-in cost (139 vs 1292
-tokens/task). Measured after three separate leaks were closed, so the numbers
-are from a clean environment.
-
-**Fold index marginal utility: unresolved.** Both the positive and the negative
-estimates are invalidated. The first Stage 2 ran in a contaminated environment.
-The clean paired batch found a consistent effect (six of six tasks searched less
-and paged in less with the cue visible), and the same-batch dose-response did
-not reproduce it (`boundary-aligned 0/4`, recall-token delta reversing sign).
-The most likely reason is that two of the six index tasks carry heavy-tail
-stopping behaviour that swamps the effect being measured.
-
-Existing evidence is not sufficient to tune the shipped 1% budget in either
-direction. Reopening this wants a new corpus, not more samples: see
-"Index corpus v2" below.
-
-**Things that turned out not to be problems**, each after being measured rather
-than argued about:
-
-| Suspected | Measured |
-| --- | --- |
-| Model writes poor queries (2.17 searches/task) | 22/23 found the target on query 1; the extra searches are one round's parallel fan-out |
-| Model reaches for the workspace before memory (7/11) | Linear event order misread; by model round it is MemoryFirst 4/6 |
-| Snippets too short for multi-value answers | 240 runes covers 12/12 tasks at 100%; records are 30-179 runes |
-| A general stopping failure | 37/44 runs issued no search after the answer was in hand; all 7 that did belong to 2 tasks |
-
-## Measurement distinctions this bench had to learn
-
-Each of these reversed a conclusion once. They are enforced in the schema now,
-not left to whoever reads the numbers next.
-
- **TargetHit is not EvidenceSufficient.** A rank-1 hit can return a window too
-  short to answer with. `FirstHitCoverage` records how much of the scored answer
-  the first hit actually handed over.
- **Linear event order is not model-round order.** Two calls in one round are a
-  parallel fan-out, not a preference. Routing compares rounds.
- **An isolated workdir is not an isolated host.** The sandbox mounts the host
-  read-only by design. Three leaks were found and closed: the corpus source, the
-  fixture transcript, and its event log sidecar.
- **A count is not its evidence.** Query text, snippet contents and trajectories
-  are all persisted now, because three analyses in a row needed a paid re-run to
-  ask a question of data already collected.
-
-## How a run stays honest
-
-The corpus holds templates, never answers: every scored literal is a `{{var}}`
-instantiated per run, so grepping this directory reveals the question and not
-the answer. `TestNoAnswerLiteralExistsInTheRepository` asserts it with
-`git grep`. Preflight rebuilds every arm and checks the answer is unreadable in
-the real `provider.Request`, the probe query ranks the target within five, and
-an index task's cue sits at exactly the scales its tier names. The fixture is
-removed from disk once the agent holds it in memory, and the directory is
-scanned for answer literals before the first provider request and again after
-the turn. Any answer reaching the model through a tool that is not `recall`
-marks the run contaminated and takes it out of every statistic.
-
-## Index corpus v2, if this is reopened
-
-The present index tasks are not a fit substrate: two of six carry stopping
-behaviour that dominates the signal. A task should qualify only if a run-index
-run reports a `stopping_class` other than `PostSufficientRetrieval` or
-`PostSufficientRunaway`: `SnippetStop`, `SnippetThenRead` and `DefensiveRead`
-all qualify, and a defensive read is deliberately not counted as waste. The
-class comes from the three counts in that report — `searches_after_sufficient`,
-`reads_after_sufficient` and `escapes_after_sufficient` — together with how
-sufficiency was reached and how many rounds followed it. The remaining
-requirements still apply — successful tool call, answer in the output rather
-than the cue, rank 1, `FirstHitCoverage` complete, one or two simple values,
-nothing in the workspace. Keep i03 and i04 — as a stopping corpus, not an
-efficiency one.
-
-`-mode=calibrate` is not a screening tool: it builds and reads back the
-fixtures, and reads nothing else, so its cue-visibility profile is the same for
-every task.
-
-Then repeats rather than levels: cue-present against cue-absent on each task's
-own boundary, three runs per cell. What is unknown is the variance, and four
-budgets sampled once each cannot show it.
@@ -1,171 +0,0 @@
-// capture.go — what the agent actually hands a provider.
-package main
-
-import (
-	"context"
-	"encoding/json"
-	"fmt"
-	"reasonix/internal/state/sessionstore"
-	"strings"
-
-	"reasonix/internal/assembly/boot"
-	"reasonix/internal/contract/ablation"
-	"reasonix/internal/contract/event"
-	"reasonix/internal/contract/provider"
-	"reasonix/internal/contract/tool"
-	"reasonix/internal/runtime/agent"
-	_ "reasonix/internal/tools/builtin"
-)
-
-// A projection is what the agent believes; a provider.Request is what the model
-// is told. Rendering, sanitization and tool-schema construction sit between
-// them, so the arm's contract is checked on the request.
-
-type captureProvider struct {
-	requests []provider.Request
-}
-
-func (p *captureProvider) Name() string { return "contextbench-capture" }
-
-// Stream records the request and answers with a clean terminal response. An
-// empty answer would buy an empty-final retry and a second request, which is
-// not the one round this is here to inspect.
-func (p *captureProvider) Stream(_ context.Context, req provider.Request) (<-chan provider.Chunk, error) {
-	p.requests = append(p.requests, cloneRequest(req))
-	ch := make(chan provider.Chunk, 2)
-	ch <- provider.Chunk{Type: provider.ChunkText, Text: "contextbench preflight complete"}
-	ch <- provider.Chunk{Type: provider.ChunkDone}
-	close(ch)
-	return ch, nil
-}
-
-func cloneRequest(req provider.Request) provider.Request {
-	out := req
-	out.Messages = append([]provider.Message(nil), req.Messages...)
-	out.Tools = append([]provider.ToolSchema(nil), req.Tools...)
-	return out
-}
-
-// captureRequest rebuilds the session the way a resume does — transcript from
-// disk, projection from its sidecar, the shipped tool surface at this arm —
-// and returns the request the first round hands the provider.
-func captureRequest(f builtFixture, arm ablation.Set, prompt string) (provider.Request, error) {
-	sess, err := sessionstore.LoadSession(f.Path)
-	if err != nil {
-		return provider.Request{}, fmt.Errorf("load session: %w", err)
-	}
-	reg := tool.NewRegistry()
-	for _, t := range tool.Builtins() {
-		reg.Add(t)
-	}
-	boot.ApplyUnifiedProviderToolSurface(reg, false, arm)
-
-	cap := &captureProvider{}
-	a := agent.New(cap, reg, sess, agent.Options{
-		ContextWindow: fixtureWindow, CompactRatio: 0.5, RecentKeep: 2,
-		SessionPath: f.Path, KeepPolicy: agent.KeepErrors, Ablation: arm,
-	}, event.Discard)
-	a.LoadProjectionSidecar(f.Path)
-
-	if err := a.Run(context.Background(), prompt); err != nil {
-		return provider.Request{}, fmt.Errorf("run: %w", err)
-	}
-	// The turn's own request is the one carrying tools. A fold on the way in
-	// asks the same provider for a digest, and that request has none — taking
-	// the first request blindly would inspect the summarizer's prompt.
-	for _, req := range cap.requests {
-		if len(req.Tools) > 0 {
-			return req, nil
-		}
-	}
-	return provider.Request{}, fmt.Errorf("no request carried tools (%d captured)", len(cap.requests))
-}
-
-// surfaceField is one place a model could read a string, named so a leak
-// report says where rather than only that.
-type surfaceField struct {
-	Path string
-	Text string
-}
-
-// providerSurface enumerates every provider-visible string in a request.
-// Enumerated rather than marshalled whole: a Message also carries local-only
-// fields, and scanning those would fail on bytes no model ever sees.
-func providerSurface(req provider.Request) []surfaceField {
-	var out []surfaceField
-	add := func(path, text string) {
-		if strings.TrimSpace(text) != "" {
-			out = append(out, surfaceField{Path: path, Text: text})
-		}
-	}
-	for i, m := range req.Messages {
-		add(fmt.Sprintf("messages[%d].content", i), m.Content)
-		add(fmt.Sprintf("messages[%d].reasoning_content", i), m.ReasoningContent)
-		add(fmt.Sprintf("messages[%d].name", i), m.Name)
-		for j, tc := range m.ToolCalls {
-			add(fmt.Sprintf("messages[%d].tool_calls[%d].name", i, j), tc.Name)
-			add(fmt.Sprintf("messages[%d].tool_calls[%d].arguments", i, j), tc.Arguments)
-		}
-		for j, item := range m.ResponsesItems {
-			add(fmt.Sprintf("messages[%d].responses_items[%d]", i, j), string(item))
-		}
-	}
-	for i, t := range req.Tools {
-		add(fmt.Sprintf("tools[%d].name", i), t.Name)
-		add(fmt.Sprintf("tools[%d].description", i), t.Description)
-		if len(t.Parameters) > 0 {
-			b, err := json.Marshal(t.Parameters)
-			if err == nil {
-				add(fmt.Sprintf("tools[%d].parameters", i), string(b))
-			}
-		}
-	}
-	return out
-}
-
-// findInSurface reports every place a marker is readable.
-func findInSurface(req provider.Request, marker string) []string {
-	var hits []string
-	for _, f := range providerSurface(req) {
-		if strings.Contains(f.Text, marker) {
-			hits = append(hits, f.Path)
-		}
-	}
-	return hits
-}
-
-// recallToolSchema returns the recall tool as the provider sees it.
-func recallToolSchema(req provider.Request) (provider.ToolSchema, bool) {
-	for _, t := range req.Tools {
-		if t.Name == "recall" {
-			return t, true
-		}
-	}
-	return provider.ToolSchema{}, false
-}
-
-// localOnlyLeaks reports host-side fields that should never survive the
-// provider boundary. A hit here is a boundary regression, marker or no marker.
-func localOnlyLeaks(req provider.Request) []string {
-	var out []string
-	for i, m := range req.Messages {
-		switch {
-		case m.RawContent != "":
-			out = append(out, fmt.Sprintf("messages[%d].raw_content", i))
-		case m.LocalOnly:
-			out = append(out, fmt.Sprintf("messages[%d].local_only", i))
-		case m.DecisionReceipt != nil || len(m.DecisionReceipts) > 0:
-			out = append(out, fmt.Sprintf("messages[%d].decision_receipt", i))
-		case m.ToolExecution != nil:
-			out = append(out, fmt.Sprintf("messages[%d].tool_execution", i))
-		case len(m.MemoryCitations) > 0:
-			out = append(out, fmt.Sprintf("messages[%d].memory_citations", i))
-		}
-	}
-	return out
-}
-
-// searchWording is what a search-off arm must never be told. The arm exists to
-// reproduce the runtime before search; an arm offered a capability it does not
-// have is a different experiment.
-var searchWording = []string{"recall with a query", "search the whole folded region", "query still finds"}
@@ -1,648 +0,0 @@
-// contextbench.go — what the model did with the affordance, read off structure.
-package main
-
-import (
-	"encoding/json"
-	"fmt"
-	"io"
-	"regexp"
-	"slices"
-	"strconv"
-	"strings"
-
-	"reasonix/internal/base/tokencount"
-	"reasonix/internal/contract/provider"
-)
-
-// Every judgement here reads a tool call's name and its JSON arguments, or a
-// canonical position, and never the model's prose. "I will search the history"
-// is not a search; a recall call with a query is.
-
-// failure stages, in funnel order. Exactly one is reported per run.
-const (
-	stageReadMissed   = "NoSearch"           // never asked, and never read the target either
-	stageSearchMiss   = "SearchMiss"         // searched, target never came back
-	stageHitNotRead   = "HitNotRead"         // target was returned and never opened
-	stageAnswerWrong  = "ReadButAnswerWrong" // opened it and still answered wrong
-	stageRecovered    = "Recovered"
-	stageContaminated = "Contaminated"
-
-	// Routing is judged by model round, not by the linear order of tool events:
-	// two calls in one round are one parallel fan-out, and calling that
-	// "workspace first" would report a choice the model never made.
-	routeMemoryFirst    = "MemoryFirst"
-	routeWorkspaceFirst = "WorkspaceFirst"
-	routeParallel       = "Parallel"
-	routeNeither        = "Neither"
-	stageCueRead        = "DirectRead"  // read the target from an index address, no search needed
-	stageNoRetrieval    = "NoRetrieval" // answered without touching recall at all
-	toolRecall          = "recall"
-)
-
-// searchAttempt is one query the model wrote, and what it got back. Kept per
-// attempt rather than aggregated: the question is why a retrieval that the host
-// probe ranks first takes the model more than one try.
-type searchAttempt struct {
-	Ordinal     int    `json:"ordinal"`
-	Round       int    `json:"round"`
-	Query       string `json:"query"`
-	TargetRank  int    `json:"target_rank"` // 0 = the target did not come back
-	ResultCount int    `json:"result_count"`
-	// Reformulation classifies this query against the one before it, and
-	// MissingProbeTerms are the host probe's terms this query left out. Both
-	// are derived from token sets: no judge, and nothing read out of prose.
-	Reformulation     string   `json:"reformulation,omitempty"`
-	MissingProbeTerms []string `json:"missing_probe_terms,omitempty"`
-}
-
-// contextMetrics is one run's behaviour.
-type contextMetrics struct {
-	Task string `json:"task"`
-	Arm  string `json:"arm"`
-
-	SearchCalls int `json:"search_calls"`
-	// SearchHits counts searches that returned anything; TargetSearchHits
-	// counts searches that returned the target. The gap is query quality.
-	SearchHits       int `json:"search_hits"`
-	TargetSearchHits int `json:"target_search_hits"`
-	FirstTargetRank  int `json:"first_target_rank"` // 0 = never returned
-	// FirstHitCoverage: how much of the answer the first hit handed over.
-	// rank=1 with coverage 1/3 is a window too short to answer with, not a
-	// model that would not stop.
-	FirstHitCoverage int `json:"first_hit_coverage"`
-	AnswerValues     int `json:"answer_values"`
-
-	ReadCalls    int  `json:"read_calls"`
-	TargetRead   bool `json:"target_read"`
-	ReadAfterHit bool `json:"read_after_hit"`
-	// RecallReadWithoutSearch: the index addressed the target, so no search was
-	// needed. Not "direct read" — read_file is a read too, and in an audit the
-	// two mean opposite things.
-	RecallReadWithoutSearch bool `json:"recall_read_without_search"`
-	SearchThenRecallRead    bool `json:"search_then_recall_read"`
-
-	FirstSearchRound int `json:"first_search_round"`
-	FirstReadRound   int `json:"first_read_round"`
-	RetrievalRounds  int `json:"retrieval_rounds"`
-
-	// RecallReturnedTokens is what recall paged back in. The index saves
-	// resident prefix; this is what is paid dynamically instead.
-	RecallReturnedTokens int `json:"recall_returned_tokens"`
-
-	Searches []searchAttempt `json:"searches,omitempty"`
-	// PostHitSearches are queries issued after the target had already been
-	// returned: not a query problem, a result-to-read problem.
-	PostHitSearches int `json:"post_hit_searches"`
-
-	AnswerRecovered bool `json:"answer_recovered"`
-	// AnswerAmbiguous marks an answer that offered more than one candidate for
-	// a value. Markers matching proves the right fact appears; it does not
-	// prove it was submitted as the answer. Recorded, never acted on.
-	AnswerAmbiguous bool     `json:"answer_ambiguous,omitempty"`
-	AmbiguousVar    string   `json:"ambiguous_var,omitempty"`
-	MissingMarkers  []string `json:"missing_markers,omitempty"`
-	// FinalAnswer is kept bounded so a surprising run can be read back. A
-	// metric nobody can audit is a claim, not a measurement.
-	FinalAnswer   string `json:"final_answer,omitempty"`
-	FailureStage  string `json:"failure_stage"`
-	TimeoutReason string `json:"timeout_reason,omitempty"`
-
-	// UnexpectedWorkTools are investigation tools used instead of recall.
-	// Reaching for the workspace is behaviour, not noise: how a model routes
-	// its uncertainty when memory is missing is part of what is measured.
-	UnexpectedWorkTools     []string `json:"unexpected_work_tools,omitempty"`
-	EscapeCalls             int      `json:"escape_calls"`
-	EscapeBeforeFirstRecall int      `json:"escape_before_first_recall"`
-	EscapeAfterFirstRecall  int      `json:"escape_after_first_recall"`
-	// Contaminated: the answer was reachable through something other than
-	// recall. Asserting this beats trusting the isolation, and a run that trips
-	// it leaves the statistics rather than inflating one.
-	Contaminated bool   `json:"contaminated,omitempty"`
-	LeakedVia    string `json:"leaked_via,omitempty"`
-	LeakedMarker string `json:"leaked_marker,omitempty"`
-	LeakedArgs   string `json:"leaked_args,omitempty"`
-	LeakedOutput string `json:"leaked_output,omitempty"`
-	// EscapeFoundAnswer separates "looked at the workspace" from "found the
-	// answer there": different facts about the same call.
-	EscapeFoundAnswer bool `json:"escape_found_answer,omitempty"`
-
-	CueVisible bool `json:"cue_visible"`
-
-	// trajectory is kept out of the metric line and written beside it, so a
-	// future question can be asked of this batch instead of a new one.
-	trajectory []provider.Message
-
-	// CueDirectRead: the cue was addressed, the first memory action was a
-	// positions read, and it covered the target. Crediting any positions call
-	// would hand the index a search's work.
-	CueDirectRead bool `json:"cue_direct_read"`
-
-	// Sufficiency, not hit. A search that returns the target at rank 1 has not
-	// necessarily returned the answer: the snippet is bounded. The stopping
-	// boundary is the round where everything needed was readable.
-	FirstSufficientRound        int    `json:"first_sufficient_round"`
-	SufficientVia               string `json:"sufficient_via"` // snippet | read | never
-	RoundsAfterSufficient       int    `json:"rounds_after_sufficient"`
-	SearchesAfterSufficient     int    `json:"searches_after_sufficient"`
-	ReadsAfterSufficient        int    `json:"reads_after_sufficient"`
-	EscapesAfterSufficient      int    `json:"escapes_after_sufficient"`
-	RecallTokensAfterSufficient int    `json:"recall_tokens_after_sufficient"`
-	StoppingClass               string `json:"stopping_class"`
-
-	// What happened after the answer was already in hand. A timeout that spent
-	// seven rounds past its target is a stopping problem, and extending the
-	// budget would only buy the loop more time.
-	TargetFirstSeenRound    int  `json:"target_first_seen_round"`
-	RoundsAfterTarget       int  `json:"rounds_after_target"`
-	SearchesAfterTarget     int  `json:"searches_after_target"`
-	RecallTokensAfterTarget int  `json:"recall_tokens_after_target"`
-	TargetHeldAtEnd         bool `json:"target_held_at_end"`
-
-	// Routing is where the model looked first when its memory was missing.
-	Routing          string `json:"routing"`
-	FirstRecallRound int    `json:"first_recall_round"`
-	FirstEscapeRound int    `json:"first_escape_round"`
-}
-
-// hitPosition matches the address recall prints for a search hit. Reading our
-// own rendered format, not guessing at the model's words.
-var hitPosition = regexp.MustCompile(`(?m)^#(\d+) `)
-
-// scoreRun reads the messages one turn appended and reports what happened.
-// runScorer walks one turn's messages, holding the state a judgement needs:
-// which calls are open, whether the target has come back, and which search
-// each result belongs to.
-type runScorer struct {
-	m            contextMetrics
-	inst         fixtureInstance
-	target       int
-	calls        map[string]provider.ToolCall
-	pending      map[string]int
-	round        int
-	sawTargetHit bool
-	// seen accumulates everything recall has shown the model, so sufficiency is
-	// judged on what it could read rather than on which call produced it.
-	seen       strings.Builder
-	sufficient bool
-}
-
-// scoreRun reads the messages one turn appended and reports what happened.
-func scoreRun(msgs []provider.Message, inst fixtureInstance, target int, arm string, cueVisible bool) contextMetrics {
-	s := &runScorer{
-		m:    contextMetrics{Task: inst.Task.ID, Arm: arm, CueVisible: cueVisible},
-		inst: inst, target: target,
-		calls: map[string]provider.ToolCall{}, pending: map[string]int{},
-	}
-	for _, msg := range msgs {
-		if len(msg.ToolCalls) > 0 {
-			s.round++
-		}
-		for _, tc := range msg.ToolCalls {
-			s.calls[tc.ID] = tc
-			s.observeCall(tc)
-		}
-		if msg.Role == provider.RoleTool {
-			s.observeResult(msg)
-		}
-	}
-	return s.finish(cueVisible)
-}
-
-// observeCall records one tool call: a recall search, a recall read, or a
-// reach for the workspace.
-func (s *runScorer) observeCall(tc provider.ToolCall) {
-	m := &s.m
-	if tc.Name != toolRecall {
-		if !containsString(m.UnexpectedWorkTools, tc.Name) {
-			m.UnexpectedWorkTools = append(m.UnexpectedWorkTools, tc.Name)
-		}
-		m.EscapeCalls++
-		if s.sufficient {
-			m.EscapesAfterSufficient++
-		}
-		if m.FirstEscapeRound == 0 {
-			m.FirstEscapeRound = s.round
-		}
-		if m.SearchCalls+m.ReadCalls == 0 {
-			m.EscapeBeforeFirstRecall++
-		} else {
-			m.EscapeAfterFirstRecall++
-		}
-		return
-	}
-	if m.FirstRecallRound == 0 {
-		m.FirstRecallRound = s.round
-	}
-	query, positions := recallArgs(tc.Arguments)
-	switch {
-	case query != "":
-		m.SearchCalls++
-		if m.FirstSearchRound == 0 {
-			m.FirstSearchRound = s.round
-		}
-		if s.sawTargetHit {
-			m.PostHitSearches++
-			m.SearchesAfterTarget++
-		}
-		if s.sufficient {
-			m.SearchesAfterSufficient++
-		}
-		m.Searches = append(m.Searches, searchAttempt{Ordinal: m.SearchCalls, Round: s.round, Query: query})
-		s.pending[tc.ID] = len(m.Searches) - 1
-	case len(positions) > 0:
-		m.ReadCalls++
-		if s.sufficient {
-			m.ReadsAfterSufficient++
-		}
-		if m.FirstReadRound == 0 {
-			m.FirstReadRound = s.round
-		}
-		// A read covers the target when it names its position: recall returns
-		// the tool results under the caller's own address.
-		if containsInt(positions, s.target) {
-			m.TargetRead = true
-			switch {
-			case s.sawTargetHit:
-				m.ReadAfterHit, m.SearchThenRecallRead = true, true
-			case m.SearchCalls == 0:
-				m.RecallReadWithoutSearch = true
-			}
-		}
-	}
-}
-
-// observeResult reads what a call returned: contamination, page-in cost, and
-// whether this search brought the target back.
-func (s *runScorer) observeResult(msg provider.Message) {
-	m := &s.m
-	call, ok := s.calls[msg.ToolCallID]
-	if !ok {
-		return
-	}
-	if call.Name != toolRecall {
-		// Codenames only: a three-digit epoch appears in any long file by
-		// chance, and one did. A model holding the codename holds the numbers
-		// beside it anyway.
-		for _, marker := range s.inst.AnswerMarkers {
-			if len(marker) >= 6 && strings.Contains(msg.Content, marker) {
-				m.Contaminated, m.LeakedVia, m.LeakedMarker = true, call.Name, marker
-				m.EscapeFoundAnswer = true
-				m.LeakedArgs, m.LeakedOutput = truncate(call.Arguments, 200), truncate(msg.Content, 300)
-			}
-		}
-		return
-	}
-	cost := tokencount.Text(msg.Content)
-	m.RecallReturnedTokens += cost
-	if s.sawTargetHit {
-		m.RecallTokensAfterTarget += cost
-	}
-	if s.sufficient {
-		m.RecallTokensAfterSufficient += cost
-	}
-	// Everything recall showed the model, judged as a whole: the answer may
-	// arrive across a snippet and a read, and either way it is readable.
-	s.seen.WriteString(msg.Content)
-	s.seen.WriteString("\n")
-	if !s.sufficient && allMarkersIn(s.seen.String(), s.inst.AnswerMarkers) {
-		s.sufficient = true
-		m.FirstSufficientRound = s.round
-		query, _ := recallArgs(call.Arguments)
-		m.SufficientVia = "read"
-		if query != "" {
-			m.SufficientVia = "snippet"
-		}
-	}
-	if query, _ := recallArgs(call.Arguments); query == "" {
-		return
-	}
-	positions := searchHitPositions(msg.Content)
-	if len(positions) > 0 {
-		m.SearchHits++
-	}
-	rank := indexOfInt(positions, s.target)
-	if i, ok := s.pending[msg.ToolCallID]; ok {
-		m.Searches[i].TargetRank, m.Searches[i].ResultCount = rank, len(positions)
-	}
-	if rank == 0 {
-		return
-	}
-	m.TargetSearchHits++
-	if !s.sawTargetHit {
-		m.TargetFirstSeenRound = s.round
-		m.FirstHitCoverage = answerCoverage(msg.Content, s.inst.AnswerMarkers)
-	}
-	s.sawTargetHit = true
-	if m.FirstTargetRank == 0 {
-		m.FirstTargetRank = rank
-	}
-}
-
-// finish derives what only the whole walk can say.
-func (s *runScorer) finish(cueVisible bool) contextMetrics {
-	m := s.m
-	m.RetrievalRounds = m.SearchCalls + m.ReadCalls
-	m.TargetHeldAtEnd = s.sawTargetHit || m.TargetRead
-	if m.TargetFirstSeenRound > 0 {
-		m.RoundsAfterTarget = s.round - m.TargetFirstSeenRound
-	}
-	m.Routing = routing(m.FirstRecallRound, m.FirstEscapeRound)
-	m.CueDirectRead = cueVisible && m.RecallReadWithoutSearch
-	if m.SufficientVia == "" {
-		m.SufficientVia = "never"
-	}
-	if m.FirstSufficientRound > 0 {
-		m.RoundsAfterSufficient = s.round - m.FirstSufficientRound
-	}
-	m.AnswerValues = len(s.inst.AnswerMarkers)
-	m.StoppingClass = stoppingClass(m)
-	classifySearches(&m, s.inst)
-	return m
-}
-
-// scoreAnswer grades the final text on the task's markers. Nonces, matched
-// exactly: no judge model, and nothing a general answer could satisfy.
-func (m *contextMetrics) scoreAnswer(final string, inst fixtureInstance) {
-	lower := strings.ToLower(final)
-	for _, marker := range inst.AnswerMarkers {
-		if !strings.Contains(lower, strings.ToLower(marker)) {
-			m.MissingMarkers = append(m.MissingMarkers, marker)
-		}
-	}
-	m.AnswerRecovered = len(m.MissingMarkers) == 0
-	m.AnswerAmbiguous, m.AmbiguousVar = ambiguousAnswer(final, inst)
-	m.FailureStage = m.stage()
-	if len(final) > 1200 {
-		final = final[:1200] + "…"
-	}
-	m.FinalAnswer = final
-}
-
-// stage places the run in the funnel. One label, chosen at the first step that
-// did not happen, so a table of stages reads as where the retrieval broke.
-func (m contextMetrics) stage() string {
-	switch {
-	case m.Contaminated:
-		return stageContaminated
-	case m.AnswerRecovered && m.RecallReadWithoutSearch:
-		return stageCueRead
-	case m.AnswerRecovered:
-		return stageRecovered
-	case m.TargetRead:
-		return stageAnswerWrong
-	case m.TargetSearchHits > 0:
-		return stageHitNotRead
-	case m.SearchCalls > 0:
-		return stageSearchMiss
-	case m.ReadCalls > 0:
-		return stageReadMissed
-	default:
-		return stageNoRetrieval
-	}
-}
-
-// recallArgs pulls the two fields that say which operation a recall call is.
-func recallArgs(args string) (query string, positions []int) {
-	var parsed struct {
-		Query     string `json:"query"`
-		Positions []int  `json:"positions"`
-	}
-	if err := json.Unmarshal([]byte(args), &parsed); err != nil {
-		return "", nil
-	}
-	return strings.TrimSpace(parsed.Query), parsed.Positions
-}
-
-// searchHitPositions reads the addresses out of a search result, in rank order.
-func searchHitPositions(text string) []int {
-	var out []int
-	for _, match := range hitPosition.FindAllStringSubmatch(text, -1) {
-		if n, err := strconv.Atoi(match[1]); err == nil {
-			out = append(out, n)
-		}
-	}
-	return out
-}
-
-func containsInt(list []int, want int) bool { return slices.Contains(list, want) }
-
-// indexOfInt returns a 1-based rank, or 0 when absent.
-func indexOfInt(list []int, want int) int {
-	for i, v := range list {
-		if v == want {
-			return i + 1
-		}
-	}
-	return 0
-}
-
-func containsString(list []string, want string) bool { return slices.Contains(list, want) }
-
-// funnel summarizes one arm's runs the way the experiment is read.
-type funnel struct {
-	Arm                                       string
-	Runs                                      int
-	Scored, Contaminated, Errored             int
-	Searched, TargetHit, TargetRead, Answered int
-	CueReads                                  int
-	SearchCalls, ReadCalls, RecallTokens      int
-	Escapes, EscapeCalls                      int
-	Stages, Routes                            map[string]int
-}
-
-func summarize(arm string, runs []contextMetrics) funnel {
-	f := funnel{Arm: arm, Runs: len(runs), Stages: map[string]int{}, Routes: map[string]int{}}
-	for _, r := range runs {
-		if r.Contaminated {
-			f.Contaminated++
-			f.Stages[r.FailureStage]++
-			continue
-		}
-		if r.FailureStage == "RunError" {
-			f.Errored++
-			f.Stages[r.FailureStage]++
-			continue
-		}
-		f.Scored++
-		if r.SearchCalls > 0 {
-			f.Searched++
-		}
-		if r.TargetSearchHits > 0 {
-			f.TargetHit++
-		}
-		if r.TargetRead {
-			f.TargetRead++
-		}
-		if r.AnswerRecovered {
-			f.Answered++
-		}
-		if r.RecallReadWithoutSearch {
-			f.CueReads++
-		}
-		if len(r.UnexpectedWorkTools) > 0 {
-			f.Escapes++
-		}
-		f.EscapeCalls += r.EscapeCalls
-		f.SearchCalls += r.SearchCalls
-		f.ReadCalls += r.ReadCalls
-		f.RecallTokens += r.RecallReturnedTokens
-		f.Stages[r.FailureStage]++
-		if r.FailureStage != "RunError" {
-			f.Routes[r.Routing]++
-		}
-	}
-	return f
-}
-
-func (f funnel) line() string {
-	per := func(n int) float64 {
-		if f.Scored == 0 {
-			return 0
-		}
-		return float64(n) / float64(f.Scored)
-	}
-	line := fmt.Sprintf("%-14s runs=%-3d answered=%d/%d  searched=%d  target-hit=%d  cue-read=%d  search/task=%.2f  recall-tok/task=%.0f  escape/task=%.2f",
-		f.Arm, f.Scored, f.Answered, f.Scored, f.Searched, f.TargetHit, f.CueReads,
-		per(f.SearchCalls), per(f.RecallTokens), per(f.EscapeCalls))
-	if f.Contaminated > 0 {
-		line += fmt.Sprintf("  CONTAMINATED=%d", f.Contaminated)
-	}
-	if f.Errored > 0 {
-		line += fmt.Sprintf("  ERRORED=%d", f.Errored)
-	}
-	return line
-}
-
-func writeJSONLine(w io.Writer, v any) error {
-	b, err := json.Marshal(v)
-	if err != nil {
-		return err
-	}
-	_, err = fmt.Fprintln(w, string(b))
-	return err
-}
-
-func truncate(s string, n int) string {
-	if len(s) <= n {
-		return s
-	}
-	return s[:n] + "…"
-}
-
-// routing says where the model looked first, comparing rounds rather than
-// event order. Same round is a parallel fan-out, not a preference.
-func routing(recallRound, escapeRound int) string {
-	switch {
-	case recallRound == 0 && escapeRound == 0:
-		return routeNeither
-	case recallRound == 0:
-		return routeWorkspaceFirst
-	case escapeRound == 0:
-		return routeMemoryFirst
-	case recallRound < escapeRound:
-		return routeMemoryFirst
-	case escapeRound < recallRound:
-		return routeWorkspaceFirst
-	default:
-		return routeParallel
-	}
-}
-
-// ambiguousAnswer reports an answer naming more than one candidate of a
-// codename's shape. Only codenames are checked: they carry a stem this run
-// declared, so two distinct "fence-…" values in one answer is one answer too
-// many. Numbers have no such shape and are left alone.
-func ambiguousAnswer(final string, inst fixtureInstance) (bool, string) {
-	for _, spec := range inst.Task.Vars {
-		if spec.Kind != varCodename {
-			continue
-		}
-		seen := map[string]bool{}
-		for _, match := range regexp.MustCompile(regexp.QuoteMeta(spec.Word)+`-[a-z0-9]{3,}`).FindAllString(final, -1) {
-			seen[match] = true
-		}
-		if len(seen) > 1 {
-			return true, spec.Name
-		}
-	}
-	return false, ""
-}
-
-func allMarkersIn(text string, markers []string) bool {
-	if len(markers) == 0 {
-		return false
-	}
-	lower := strings.ToLower(text)
-	for _, marker := range markers {
-		if !strings.Contains(lower, strings.ToLower(marker)) {
-			return false
-		}
-	}
-	return true
-}
-
-// Stopping classes. The boundary is sufficiency, not the target coming back:
-// a rank-1 hit whose snippet was too short to answer leaves a read entirely
-// justified, and counting that as waste would blame the model for the bound.
-const (
-	stopSnippet       = "SnippetStop"             // enough arrived, and the turn ended
-	stopSnippetRead   = "SnippetThenRead"         // the snippet was short; the read finished it
-	stopDefensiveRead = "DefensiveRead"           // enough had arrived, one read confirmed it
-	stopPostRetrieval = "PostSufficientRetrieval" // enough had arrived, and retrieval continued
-	stopRunaway       = "PostSufficientRunaway"   // it continued for rounds, or ran out of budget
-	stopNeverEnough   = "NeverSufficient"
-)
-
-// runawayRounds is where continuing stops looking like confirmation.
-const runawayRounds = 3
-
-func stoppingClass(m contextMetrics) string {
-	if m.FirstSufficientRound == 0 {
-		return stopNeverEnough
-	}
-	after := m.SearchesAfterSufficient + m.ReadsAfterSufficient + m.EscapesAfterSufficient
-	switch {
-	case after == 0 && m.SufficientVia == "snippet":
-		return stopSnippet
-	case after == 0:
-		return stopSnippetRead
-	case m.RoundsAfterSufficient >= runawayRounds || m.FailureStage == "RunError":
-		return stopRunaway
-	case after == 1 && m.ReadsAfterSufficient == 1:
-		return stopDefensiveRead
-	default:
-		return stopPostRetrieval
-	}
-}
-
-// reportStopping is the funnel that matters once retrieval succeeds: what the
-// turn did after it already had everything it needed.
-func reportStopping(all []contextMetrics) string {
-	classes := map[string]int{}
-	var extraRounds, extraTokens []float64
-	sufficient, answeredNext := 0, 0
-	for _, m := range all {
-		if m.Contaminated {
-			continue
-		}
-		classes[m.StoppingClass]++
-		if m.FirstSufficientRound == 0 {
-			continue
-		}
-		sufficient++
-		extraRounds = append(extraRounds, float64(m.RoundsAfterSufficient))
-		extraTokens = append(extraTokens, float64(m.RecallTokensAfterSufficient))
-		if m.RoundsAfterSufficient <= 1 {
-			answeredNext++
-		}
-	}
-	var b strings.Builder
-	b.WriteString("\n## After the evidence was already enough\n")
-	fmt.Fprintf(&b, "  sufficient in %d runs; ended within one round in %d\n", sufficient, answeredNext)
-	fmt.Fprintf(&b, "  extra rounds  mean %.1f  median %.1f  p90 %.0f  max %.0f\n",
-		mean(extraRounds), median(extraRounds), percentile(extraRounds, 0.9), maxOf(extraRounds))
-	fmt.Fprintf(&b, "  extra recall tokens  mean %.0f  median %.0f  p90 %.0f  max %.0f\n",
-		mean(extraTokens), median(extraTokens), percentile(extraTokens, 0.9), maxOf(extraTokens))
-	b.WriteString("  " + countsLine(classes) + "\n")
-	return b.String()
-}
@@ -1,337 +0,0 @@
-package main
-
-import (
-	"fmt"
-	"strings"
-	"testing"
-
-	"reasonix/internal/contract/provider"
-)
-
-// The scorer decides what the pilot concludes, so it is held to the same rule
-// the corpus is: judgements read structure. A model that says it will search
-// has not searched; a recall call carrying a query has.
-
-const scorerTarget = 42
-
-func recallCall(id, args string) provider.Message {
-	return provider.Message{Role: provider.RoleAssistant, ToolCalls: []provider.ToolCall{
-		{ID: id, Name: "recall", Arguments: args},
-	}}
-}
-
-func toolResult(id, text string) provider.Message {
-	return provider.Message{Role: provider.RoleTool, ToolCallID: id, Name: "recall", Content: text}
-}
-
-func hitBlock(positions ...int) string {
-	var b strings.Builder
-	b.WriteString("Folded context matching \"x\":\n")
-	for _, p := range positions {
-		fmt.Fprintf(&b, "\n#%d assistant_text\nsome snippet here\n", p)
-	}
-	return b.String()
-}
-
-func scorerTask() fixtureInstance {
-	return fixtureInstance{Task: contextTask{ID: "t"}, AnswerMarkers: []string{"cobalt-lark-17"}}
-}
-
-func score(t *testing.T, msgs []provider.Message, final string) contextMetrics {
-	t.Helper()
-	m := scoreRun(msgs, scorerTask(), scorerTarget, "arm", false)
-	m.scoreAnswer(final, scorerTask())
-	return m
-}
-
-// The whole funnel, walked end to end.
-func TestScorerReadsTheFullRetrievalChain(t *testing.T) {
-	m := score(t, []provider.Message{
-		recallCall("c1", `{"query":"retry boundary"}`),
-		toolResult("c1", hitBlock(7, scorerTarget)),
-		recallCall("c2", fmt.Sprintf(`{"positions":[%d]}`, scorerTarget)),
-		toolResult("c2", "#42\n[assistant]\nthe ownership token is cobalt-lark-17"),
-		{Role: provider.RoleAssistant, Content: "The token was cobalt-lark-17."},
-	}, "The token was cobalt-lark-17.")
-
-	if m.SearchCalls != 1 || m.ReadCalls != 1 {
-		t.Errorf("calls = %d search / %d read, want 1/1", m.SearchCalls, m.ReadCalls)
-	}
-	if m.TargetSearchHits != 1 || m.FirstTargetRank != 2 {
-		t.Errorf("target hit %d at rank %d, want 1 at rank 2", m.TargetSearchHits, m.FirstTargetRank)
-	}
-	if !m.TargetRead || !m.ReadAfterHit || m.RecallReadWithoutSearch {
-		t.Errorf("read flags = read %v afterHit %v direct %v", m.TargetRead, m.ReadAfterHit, m.RecallReadWithoutSearch)
-	}
-	if !m.AnswerRecovered || m.FailureStage != stageRecovered {
-		t.Errorf("stage = %q recovered=%v, want Recovered", m.FailureStage, m.AnswerRecovered)
-	}
-	if m.RecallReturnedTokens == 0 {
-		t.Error("recall returned tokens were not counted")
-	}
-}
-
-// An index arm's claim: the address was already on screen, so no search was
-// needed. That is a different outcome from finding it, and gets its own stage.
-func TestScorerSeparatesCueReadFromSearch(t *testing.T) {
-	m := score(t, []provider.Message{
-		recallCall("c1", fmt.Sprintf(`{"positions":[%d]}`, scorerTarget)),
-		toolResult("c1", "#42\nthe ownership token is cobalt-lark-17"),
-		{Role: provider.RoleAssistant, Content: "cobalt-lark-17"},
-	}, "cobalt-lark-17")
-
-	if m.SearchCalls != 0 || !m.RecallReadWithoutSearch {
-		t.Errorf("search=%d direct=%v, want 0 and true", m.SearchCalls, m.RecallReadWithoutSearch)
-	}
-	if m.FailureStage != stageCueRead {
-		t.Errorf("stage = %q, want %q", m.FailureStage, stageCueRead)
-	}
-}
-
-// Every way the chain can break gets its own label, so a failed run says where
-// rather than only that.
-func TestScorerNamesWhereTheChainBroke(t *testing.T) {
-	for _, tc := range []struct {
-		name  string
-		msgs  []provider.Message
-		final string
-		want  string
-	}{
-		{
-			name:  "never touched recall",
-			msgs:  []provider.Message{{Role: provider.RoleAssistant, Content: "I do not recall that."}},
-			final: "I do not recall that.",
-			want:  stageNoRetrieval,
-		},
-		{
-			name: "searched and missed",
-			msgs: []provider.Message{
-				recallCall("c1", `{"query":"unrelated words"}`),
-				toolResult("c1", hitBlock(3, 9)),
-			},
-			final: "not found",
-			want:  stageSearchMiss,
-		},
-		{
-			name: "found it and never opened it",
-			msgs: []provider.Message{
-				recallCall("c1", `{"query":"retry boundary"}`),
-				toolResult("c1", hitBlock(scorerTarget)),
-			},
-			final: "probably something about retries",
-			want:  stageHitNotRead,
-		},
-		{
-			name: "read it and answered wrong",
-			msgs: []provider.Message{
-				recallCall("c1", fmt.Sprintf(`{"positions":[%d]}`, scorerTarget)),
-				toolResult("c1", "#42\nthe token is elsewhere"),
-			},
-			final: "I think it was amber-9",
-			want:  stageAnswerWrong,
-		},
-		{
-			name: "read the wrong position",
-			msgs: []provider.Message{
-				recallCall("c1", `{"positions":[7]}`),
-				toolResult("c1", "#7\nnothing useful"),
-			},
-			final: "unclear",
-			want:  stageReadMissed,
-		},
-	} {
-		t.Run(tc.name, func(t *testing.T) {
-			if got := score(t, tc.msgs, tc.final).FailureStage; got != tc.want {
-				t.Errorf("stage = %q, want %q", got, tc.want)
-			}
-		})
-	}
-}
-
-// The answer lives only in folded history, so reaching for the workspace is a
-// strategy change worth seeing — including when it happens to work.
-func TestScorerRecordsWorkspaceEscapes(t *testing.T) {
-	m := score(t, []provider.Message{
-		{Role: provider.RoleAssistant, ToolCalls: []provider.ToolCall{
-			{ID: "g1", Name: "bash", Arguments: `{"command":"rg cobalt-lark"}`},
-			{ID: "g2", Name: "read_file", Arguments: `{"path":"notes.md"}`},
-		}},
-		{Role: provider.RoleAssistant, Content: "cobalt-lark-17"},
-	}, "cobalt-lark-17")
-
-	if len(m.UnexpectedWorkTools) != 2 {
-		t.Fatalf("escapes = %v, want bash and read_file", m.UnexpectedWorkTools)
-	}
-	if m.SearchCalls != 0 || m.ReadCalls != 0 {
-		t.Errorf("a workspace tool was counted as recall: search=%d read=%d", m.SearchCalls, m.ReadCalls)
-	}
-	// It answered, so the run is Recovered — the escape is reported beside the
-	// outcome rather than overriding it.
-	if !m.AnswerRecovered {
-		t.Error("a correct answer was not scored as recovered")
-	}
-}
-
-// Narration is not action.
-func TestScorerIgnoresWhatTheModelSaysItWillDo(t *testing.T) {
-	m := score(t, []provider.Message{
-		{Role: provider.RoleAssistant, Content: "Let me search the folded context with recall for the retry boundary."},
-	}, "Let me search the folded context with recall for the retry boundary.")
-	if m.SearchCalls != 0 || m.FailureStage != stageNoRetrieval {
-		t.Errorf("prose was scored as a search: calls=%d stage=%q", m.SearchCalls, m.FailureStage)
-	}
-}
-
-// Every marker must land, so a partially right answer is not a recovery.
-func TestScorerRequiresEveryMarker(t *testing.T) {
-	task := fixtureInstance{Task: contextTask{ID: "t"}, AnswerMarkers: []string{"comet-42", "per-lineage"}}
-	m := scoreRun(nil, task, scorerTarget, "arm", false)
-	m.scoreAnswer("the salt was comet-42", task)
-	if m.AnswerRecovered {
-		t.Error("a half answer scored as recovered")
-	}
-	if len(m.MissingMarkers) != 1 || m.MissingMarkers[0] != "per-lineage" {
-		t.Errorf("missing = %v, want per-lineage", m.MissingMarkers)
-	}
-}
-
-// An answer reachable through anything but recall invalidates the run. The
-// workspace is isolated so this should never fire; asserting it beats trusting
-// it, and a tripped run leaves the statistics rather than inflating one.
-func TestScorerInvalidatesRunsThatLeakedThroughAnotherTool(t *testing.T) {
-	m := score(t, []provider.Message{
-		{Role: provider.RoleAssistant, ToolCalls: []provider.ToolCall{
-			{ID: "g1", Name: "bash", Arguments: `{"command":"rg -r ."}`},
-		}},
-		{Role: provider.RoleTool, ToolCallID: "g1", Name: "bash", Content: "corpus.go: the token is cobalt-lark-17"},
-		{Role: provider.RoleAssistant, Content: "cobalt-lark-17"},
-	}, "cobalt-lark-17")
-
-	if !m.Contaminated || m.LeakedVia != "bash" || m.LeakedMarker != "cobalt-lark-17" {
-		t.Fatalf("leak not caught: contaminated=%v via=%q marker=%q", m.Contaminated, m.LeakedVia, m.LeakedMarker)
-	}
-	if m.FailureStage != stageContaminated {
-		t.Errorf("stage = %q, want %q even though every marker matched", m.FailureStage, stageContaminated)
-	}
-	f := summarize("arm", []contextMetrics{m})
-	if f.Scored != 0 || f.Answered != 0 || f.Contaminated != 1 {
-		t.Errorf("a contaminated run entered the statistics: %+v", f)
-	}
-}
-
-// Where an escape happened matters: before any recall is a different strategy
-// from after one came back short.
-func TestScorerPlacesEscapesRelativeToRecall(t *testing.T) {
-	m := score(t, []provider.Message{
-		{Role: provider.RoleAssistant, ToolCalls: []provider.ToolCall{
-			{ID: "g1", Name: "bash", Arguments: `{"command":"ls"}`},
-		}},
-		{Role: provider.RoleTool, ToolCallID: "g1", Name: "bash", Content: "README.md"},
-		recallCall("c1", `{"query":"retry boundary"}`),
-		toolResult("c1", hitBlock(scorerTarget)),
-		{Role: provider.RoleAssistant, ToolCalls: []provider.ToolCall{
-			{ID: "g2", Name: "read_file", Arguments: `{"path":"README.md"}`},
-		}},
-		{Role: provider.RoleTool, ToolCallID: "g2", Name: "read_file", Content: "nothing here"},
-	}, "unclear")
-
-	if m.EscapeCalls != 2 || m.EscapeBeforeFirstRecall != 1 || m.EscapeAfterFirstRecall != 1 {
-		t.Errorf("escapes = %d (%d before, %d after), want 2 (1, 1)",
-			m.EscapeCalls, m.EscapeBeforeFirstRecall, m.EscapeAfterFirstRecall)
-	}
-}
-
-// Routing is judged by round. Two calls in one round are a fan-out, and
-// reporting that as a preference would invent a decision the model never made.
-func TestRoutingIsJudgedByRoundNotEventOrder(t *testing.T) {
-	parallel := []provider.Message{
-		{Role: provider.RoleAssistant, ToolCalls: []provider.ToolCall{
-			{ID: "g1", Name: "bash", Arguments: `{"command":"ls"}`},
-			{ID: "c1", Name: "recall", Arguments: `{"query":"retry boundary"}`},
-		}},
-		{Role: provider.RoleTool, ToolCallID: "g1", Name: "bash", Content: "README.md"},
-		toolResult("c1", hitBlock(scorerTarget)),
-	}
-	if got := score(t, parallel, "unclear").Routing; got != routeParallel {
-		t.Errorf("routing = %q for one round holding both, want %q", got, routeParallel)
-	}
-
-	workspaceFirst := []provider.Message{
-		{Role: provider.RoleAssistant, ToolCalls: []provider.ToolCall{
-			{ID: "g1", Name: "bash", Arguments: `{"command":"ls"}`},
-		}},
-		{Role: provider.RoleTool, ToolCallID: "g1", Name: "bash", Content: "README.md"},
-		recallCall("c1", `{"query":"retry boundary"}`),
-		toolResult("c1", hitBlock(scorerTarget)),
-	}
-	if got := score(t, workspaceFirst, "unclear").Routing; got != routeWorkspaceFirst {
-		t.Errorf("routing = %q for workspace in an earlier round, want %q", got, routeWorkspaceFirst)
-	}
-
-	memoryFirst := []provider.Message{
-		recallCall("c1", `{"query":"retry boundary"}`),
-		toolResult("c1", hitBlock(scorerTarget)),
-		{Role: provider.RoleAssistant, ToolCalls: []provider.ToolCall{
-			{ID: "g1", Name: "bash", Arguments: `{"command":"ls"}`},
-		}},
-		{Role: provider.RoleTool, ToolCallID: "g1", Name: "bash", Content: "README.md"},
-	}
-	if got := score(t, memoryFirst, "unclear").Routing; got != routeMemoryFirst {
-		t.Errorf("routing = %q for recall in an earlier round, want %q", got, routeMemoryFirst)
-	}
-
-	if got := score(t, nil, "no idea").Routing; got != routeNeither {
-		t.Errorf("routing = %q with no tools at all, want %q", got, routeNeither)
-	}
-}
-
-// Markers matching proves the right fact appears somewhere. An answer that
-// hedges between two candidates contains it and has not submitted it.
-func TestScorerFlagsAnAnswerThatNamesTwoCandidates(t *testing.T) {
-	inst := fixtureInstance{
-		Task:          contextTask{ID: "t", Vars: []varSpec{code("marker", "fence")}},
-		AnswerMarkers: []string{"fence-qmwnv"},
-	}
-	m := scoreRun(nil, inst, 0, "arm", false)
-	m.scoreAnswer("It was fence-qmwnv, though it may have been fence-abcde.", inst)
-	if !m.AnswerRecovered {
-		t.Error("the marker is present; recovery should still read true")
-	}
-	if !m.AnswerAmbiguous || m.AmbiguousVar != "marker" {
-		t.Errorf("ambiguity not flagged: %v %q", m.AnswerAmbiguous, m.AmbiguousVar)
-	}
-
-	clean := scoreRun(nil, inst, 0, "arm", false)
-	clean.scoreAnswer("The marker was fence-qmwnv.", inst)
-	if clean.AnswerAmbiguous {
-		t.Error("a single candidate was called ambiguous")
-	}
-}
-
-// The index arm's claim is narrow: the cue was addressed, the first memory
-// action was a positions read, and it covered the target. A positions call
-// after a search located the target credits the index for the search's work.
-func TestCueDirectReadRequiresTheCueAndTheFirstAction(t *testing.T) {
-	direct := []provider.Message{
-		recallCall("c1", fmt.Sprintf(`{"positions":[%d]}`, scorerTarget)),
-		toolResult("c1", "#42\nthe token is cobalt-lark-17"),
-	}
-	m := scoreRun(direct, scorerTask(), scorerTarget, "arm", true)
-	if !m.CueDirectRead {
-		t.Error("a first-action read with the cue visible was not credited")
-	}
-
-	// Same read, but no cue was on screen to give the address.
-	if scoreRun(direct, scorerTask(), scorerTarget, "arm", false).CueDirectRead {
-		t.Error("a read was credited to a cue that was not visible")
-	}
-
-	// Same read, but a search found the address first.
-	afterSearch := append([]provider.Message{
-		recallCall("c0", `{"query":"retry boundary"}`),
-		toolResult("c0", hitBlock(scorerTarget)),
-	}, direct...)
-	if scoreRun(afterSearch, scorerTask(), scorerTarget, "arm", true).CueDirectRead {
-		t.Error("a read after a search was credited to the index")
-	}
-}
@@ -1,344 +0,0 @@
-package main
-
-import (
-	"fmt"
-	"math/rand"
-	"reasonix/internal/state/sessionstore"
-
-	"reasonix/internal/contract/provider"
-)
-
-// Synthetic history on purpose: an answer that exists in the workspace is
-// solvable by reading the workspace, which measures grep. Every answer is a
-// placeholder here and a value only inside one run.
-
-// experiment names which question a task belongs to.
-const (
-	experimentSearch = "search" // does search reach what no index addresses?
-	experimentIndex  = "index"  // what is an ambient address line worth?
-)
-
-// cueTier is the narrowest fold-index scale at which an index task's cue is
-// still addressed. Below it the model has to find the call without a hint.
-const (
-	tierQuarter = "quarter" // visible at default, half, quarter
-	tierHalf    = "half"    // visible at default, half
-	tierDefault = "default" // visible at default only
-)
-
-// plantKind is how a task's target enters the transcript.
-const (
-	plantAssistant = "assistant" // the model's own account: never index-addressed
-	plantCall      = "call"      // a tool call and its result: always index-addressed
-)
-
-// contextTask is one recall question as a template. Nothing here is an answer.
-type contextTask struct {
-	ID         string
-	Experiment string
-	TargetKind string
-	Prompt     string
-	ProbeQuery string
-	// Vars declare the placeholders. AnswerVars name the ones a correct answer
-	// must contain, CueVar the one an index line carries.
-	Vars       []varSpec
-	AnswerVars []string
-	CueVar     string
-	CueTier    string
-	// PlantAfterGen ages the target: mergeFoldIndex trims from its oldest end,
-	// so what follows a cue decides which budget addresses it. Frozen, because
-	// a run that searched for it would let the tested code move its goalposts.
-	PlantAfterGen int
-	// Plant is how the target enters, and the rest is what it says.
-	Plant     string
-	CallTool  string
-	CallArgs  string
-	CallID    string
-	PlantBody string
-}
-
-// fixtureInstance is one task with values. Scoring binds to this, never to the
-// corpus: the corpus has no answers to bind to.
-type fixtureInstance struct {
-	Task          contextTask
-	Vars          fixtureVars
-	Prompt        string
-	ProbeQuery    string
-	AnswerMarkers []string
-	CueMarker     string
-	body          string
-	callArgs      string
-}
-
-// instantiateTask resolves one task against a generator.
-func instantiateTask(t contextTask, rng *rand.Rand) (fixtureInstance, error) {
-	vars := newVars(t.Vars, rng)
-	inst := fixtureInstance{Task: t, Vars: vars}
-	var err error
-	for _, pair := range []struct {
-		dst  *string
-		text string
-	}{
-		{&inst.Prompt, t.Prompt},
-		{&inst.ProbeQuery, t.ProbeQuery},
-		{&inst.body, t.PlantBody},
-		{&inst.callArgs, t.CallArgs},
-	} {
-		if *pair.dst, err = instantiate(pair.text, vars); err != nil {
-			return fixtureInstance{}, fmt.Errorf("%s: %w", t.ID, err)
-		}
-	}
-	for _, name := range t.AnswerVars {
-		value, ok := vars[name]
-		if !ok {
-			return fixtureInstance{}, fmt.Errorf("%s: answer var %q is not declared", t.ID, name)
-		}
-		inst.AnswerMarkers = append(inst.AnswerMarkers, value)
-	}
-	if t.CueVar != "" {
-		inst.CueMarker = vars[t.CueVar]
-	}
-	return inst, nil
-}
-
-// plant appends the instantiated target and returns its canonical position.
-func (f fixtureInstance) plant(s *sessionstore.Session) int {
-	at := len(s.Messages)
-	if f.Task.Plant == plantAssistant {
-		s.Add(provider.Message{Role: provider.RoleAssistant, Content: f.body})
-		return at
-	}
-	s.Add(provider.Message{Role: provider.RoleAssistant, ToolCalls: []provider.ToolCall{
-		{ID: f.Task.CallID, Name: f.Task.CallTool, Arguments: f.callArgs},
-	}})
-	s.Add(provider.Message{Role: provider.RoleTool, ToolCallID: f.Task.CallID, Name: f.Task.CallTool, Content: f.body})
-	return at
-}
-
-func code(name, word string) varSpec { return varSpec{Name: name, Kind: varCodename, Word: word} }
-
-func num(name string, lo, hi int) varSpec {
-	return varSpec{Name: name, Kind: varInt, Min: lo, Max: hi}
-}
-
-func searchTasks() []contextTask {
-	return []contextTask{
-		{
-			ID: "r01-retry-boundary", Experiment: experimentSearch, TargetKind: "assistant_reasoning",
-			Prompt:     "What ownership token did we settle on for the retry boundary, and under what tag did we file the reason for not moving retries above framing?",
-			ProbeQuery: "retry boundary ownership token framing",
-			Vars:       []varSpec{code("token", "cobalt"), code("reason", "stream")},
-			AnswerVars: []string{"token", "reason"},
-			Plant:      plantAssistant,
-			PlantBody: "The retry boundary remains at ingress. The ownership token is {{token}}. " +
-				"Moving retries above framing can duplicate a partially received stream, which we filed as {{reason}}.",
-		},
-		{
-			ID: "r02-checkpoint-anchor", Experiment: experimentSearch, TargetKind: "assistant_reasoning",
-			Prompt:     "What was the codename of the checkpoint anchor policy, and what tag did we give the reason for not anchoring against digest positions?",
-			ProbeQuery: "checkpoint anchor policy digest positions",
-			Vars:       []varSpec{code("policy", "northglass"), code("reason", "genlocal")},
-			AnswerVars: []string{"policy", "reason"},
-			Plant:      plantAssistant,
-			PlantBody: "The checkpoint anchor policy is {{policy}}. Digest positions are generation-local — " +
-				"filed as {{reason}} — so recovery must anchor against canonical state.",
-		},
-		{
-			ID: "r03-recovery-cjk", Experiment: experimentSearch, TargetKind: "assistant_reasoning_cjk",
-			Prompt:     "我们之前定的验证失败恢复策略代号叫什么？当时把“失败 digest 不能变成 baseline”这条登记在哪个标记下？",
-			ProbeQuery: "恢复策略 baseline digest 验证失败",
-			Vars:       []varSpec{code("policy", "songzhen"), code("rule", "trustbase")},
-			AnswerVars: []string{"policy", "rule"},
-			Plant:      plantAssistant,
-			PlantBody: "恢复策略代号是 {{policy}}。验证失败以后必须保留旧基线，不能把新的 digest 晋升为 baseline，" +
-				"因为失败证据不能成为后续比较的可信基准；这条规则登记为 {{rule}}。",
-		},
-		{
-			ID: "r04-lease-fence", Experiment: experimentSearch, TargetKind: "assistant_reasoning",
-			Prompt:     "那次 lease fence 检查报告的诊断标记是什么？expected 和 observed epoch 分别是多少？",
-			ProbeQuery: "lease fence writer epoch mismatch",
-			Vars: []varSpec{
-				code("marker", "fence"), num("expected", 400, 900), num("observed", 100, 399),
-			},
-			AnswerVars: []string{"marker", "expected", "observed"},
-			Plant:      plantAssistant,
-			PlantBody: "The lease fence check reported {{marker}}: expected writer epoch {{expected}}, " +
-				"observed {{observed}}. We left the fence in place.",
-		},
-		{
-			ID: "r05-schema-guard", Experiment: experimentSearch, TargetKind: "assistant_reasoning",
-			Prompt:     "当时 schema guard 报告的标记是什么，超预算多少 token？",
-			ProbeQuery: "schema guard provider visible surface budget",
-			Vars:       []varSpec{code("marker", "amber"), num("over", 120, 480)},
-			AnswerVars: []string{"marker", "over"},
-			Plant:      plantAssistant,
-			PlantBody: "The schema guard reported {{marker}}: the provider-visible tool schema exceeded " +
-				"the allowed surface by {{over}} tokens. We trimmed a description instead.",
-		},
-		{
-			ID: "r06-shadow-probe", Experiment: experimentSearch, TargetKind: "assistant_reasoning",
-			Prompt:     "我们那次 shadow probe 的 marker、p95 divergence 和 generation 分别是多少？",
-			ProbeQuery: "shadow probe divergence generation marker",
-			Vars: []varSpec{
-				code("marker", "violet"),
-				{Name: "divergence", Kind: varDecimal, Min: 110, Max: 890},
-				num("generation", 14, 60),
-			},
-			AnswerVars: []string{"marker", "divergence", "generation"},
-			Plant:      plantAssistant,
-			PlantBody: "The shadow probe came back as marker={{marker}} with a p95 divergence of " +
-				"{{divergence}} at generation={{generation}}, which we judged within tolerance.",
-		},
-	}
-}
-
-func indexTasks() []contextTask {
-	return []contextTask{
-		{
-			ID: "i01-transport-fallback", Experiment: experimentIndex, TargetKind: "tool_input",
-			Prompt:     "transport fallback probe 当时报告的 selector 是什么？",
-			ProbeQuery: "transport fallback probe selector",
-			Vars:       []varSpec{code("cue", "tfprobe"), code("selector", "fern")},
-			AnswerVars: []string{"selector"}, CueVar: "cue", CueTier: tierQuarter,
-			PlantAfterGen: 4,
-			Plant:         plantCall, CallID: "tf1", CallTool: "bash",
-			CallArgs:  `{"command":"./scripts/probe {{cue}}"}`,
-			PlantBody: "fallback selector = {{selector}}",
-		},
-		{
-			ID: "i02-scheduler-handoff", Experiment: experimentIndex, TargetKind: "tool_input",
-			Prompt:     "scheduler handoff 当时记录的 grace 和 marker 是什么？",
-			ProbeQuery: "scheduler handoff grace marker",
-			Vars: []varSpec{
-				code("cue", "schedhandoff"), code("marker", "opal"),
-				{Name: "grace", Kind: varDuration, Min: 18, Max: 90},
-			},
-			AnswerVars: []string{"grace", "marker"}, CueVar: "cue", CueTier: tierQuarter,
-			PlantAfterGen: 4,
-			Plant:         plantCall, CallID: "sh1", CallTool: "read_file",
-			CallArgs:  `{"path":"scratch/{{cue}}.txt"}`,
-			PlantBody: "handoff grace = {{grace}}\nmarker = {{marker}}",
-		},
-		{
-			ID: "i03-coalescing", Experiment: experimentIndex, TargetKind: "tool_input",
-			Prompt:     "job completion 的 coalescing probe 最后记录的 window 和 marker 是多少？",
-			ProbeQuery: "coalescing policy completion window",
-			Vars: []varSpec{
-				code("cue", "coalescing"), code("marker", "cedar"),
-				{Name: "window", Kind: varDuration, Min: 40, Max: 140},
-			},
-			AnswerVars: []string{"window", "marker"}, CueVar: "cue", CueTier: tierHalf,
-			PlantAfterGen: 2,
-			Plant:         plantCall, CallID: "cp1", CallTool: "bash",
-			CallArgs:  `{"command":"grep {{cue}} scratch/job-notes.log"}`,
-			PlantBody: "completion coalesce window = {{window}}\nmarker = {{marker}}",
-		},
-		{
-			ID: "i04-recovery-fence", Experiment: experimentIndex, TargetKind: "tool_input",
-			Prompt:     "recovery fence 那次检查报告的 marker 和两个 epoch 是什么？",
-			ProbeQuery: "recovery fence epoch check",
-			Vars: []varSpec{
-				code("cue", "recfence"), code("marker", "quartz"),
-				num("expected", 400, 900), num("observed", 100, 399),
-			},
-			AnswerVars: []string{"marker", "expected", "observed"}, CueVar: "cue", CueTier: tierHalf,
-			PlantAfterGen: 2,
-			Plant:         plantCall, CallID: "rf1", CallTool: "bash",
-			CallArgs:  `{"command":"./scripts/check {{cue}}"}`,
-			PlantBody: "{{marker}}\nexpected epoch {{expected}}\nobserved epoch {{observed}}",
-		},
-		{
-			ID: "i05-cache-probe", Experiment: experimentIndex, TargetKind: "tool_input",
-			Prompt:     "cache probe 当时记录的 prefix salt 和 scope 是什么？",
-			ProbeQuery: "cache probe prefix salt scope",
-			Vars:       []varSpec{code("cue", "cacheprobe"), code("salt", "comet"), code("scope", "lineage")},
-			AnswerVars: []string{"salt", "scope"}, CueVar: "cue", CueTier: tierDefault,
-			PlantAfterGen: 0,
-			Plant:         plantCall, CallID: "cb1", CallTool: "read_file",
-			CallArgs:  `{"path":"scratch/{{cue}}.txt"}`,
-			PlantBody: "prefix salt = {{salt}}\nscope = {{scope}}",
-		},
-		{
-			ID: "i06-dispatch-handoff", Experiment: experimentIndex, TargetKind: "tool_input",
-			Prompt:     "dispatch handoff probe 当时确认的 mode 和 marker 是什么？",
-			ProbeQuery: "dispatch handoff probe mode",
-			Vars:       []varSpec{code("cue", "dispatchhandoff"), code("mode", "trigger"), code("marker", "iris")},
-			AnswerVars: []string{"mode", "marker"}, CueVar: "cue", CueTier: tierDefault,
-			PlantAfterGen: 0,
-			Plant:         plantCall, CallID: "dh1", CallTool: "bash",
-			CallArgs:  `{"command":"./scripts/probe {{cue}}"}`,
-			PlantBody: "handoff mode = {{mode}}\nmarker = {{marker}}",
-		},
-	}
-}
-
-func allTasks() []contextTask {
-	return append(searchTasks(), indexTasks()...)
-}
-
-func taskByID(id string) (contextTask, bool) {
-	for _, t := range allTasks() {
-		if t.ID == id {
-			return t, true
-		}
-	}
-	return contextTask{}, false
-}
-
-// tierScales is the fixture's contract: which arms must address a cue and
-// which must not. A policy change that moves a cue across a boundary fails
-// preflight rather than turning two arms into one experiment.
-func tierScales(tier string) (visible, hidden []string) {
-	switch tier {
-	case tierQuarter:
-		return []string{"default", "half", "quarter"}, []string{"off"}
-	case tierHalf:
-		return []string{"default", "half"}, []string{"quarter", "off"}
-	case tierDefault:
-		return []string{"default"}, []string{"half", "quarter", "off"}
-	default:
-		return nil, nil
-	}
-}
-
-// boundaryPair is the one comparison an index task is built for: the narrowest
-// scale that still addresses its cue, and the next one down. Holding the
-// question fixed and moving only the affordance is what four arm averages
-// cannot do, since each of those mixes six different questions.
-func boundaryPair(tier string) (cue, noCue string) {
-	switch tier {
-	case tierQuarter:
-		return "quarter", "off"
-	case tierHalf:
-		return "half", "quarter"
-	case tierDefault:
-		return "default", "half"
-	default:
-		return "", ""
-	}
-}
-
-func validateCorpus() error {
-	seen := map[string]bool{}
-	for _, t := range allTasks() {
-		if seen[t.ID] {
-			return fmt.Errorf("duplicate task id %q", t.ID)
-		}
-		seen[t.ID] = true
-		if len(t.AnswerVars) == 0 || t.ProbeQuery == "" || t.Prompt == "" || t.PlantBody == "" {
-			return fmt.Errorf("%s: a task needs a prompt, a probe query, a body and answer vars", t.ID)
-		}
-		if t.Experiment == experimentIndex {
-			if t.CueVar == "" {
-				return fmt.Errorf("%s: an index task needs a cue var", t.ID)
-			}
-			if v, _ := tierScales(t.CueTier); v == nil {
-				return fmt.Errorf("%s: unknown cue tier %q", t.ID, t.CueTier)
-			}
-		}
-		if _, err := instantiateTask(t, seededRand(t.ID, "validate")); err != nil {
-			return err
-		}
-	}
-	return nil
-}
@@ -1,243 +0,0 @@
-// dose.go — does the cost step where the cue disappears?
-package main
-
-import (
-	"fmt"
-	"math"
-	"sort"
-	"strings"
-)
-
-// Not "a bigger index is better" — CueDirectRead says the addresses are rarely
-// used. The hypothesis is that a visible action cue lowers retrieval cost,
-// which predicts a step at each task's own boundary, not a slope.
-
-// scaleOrder is the dose axis, widest first.
-var scaleOrder = []string{"default", "half", "quarter", "off"}
-
-// boundaryStep is the index into the three transitions where a tier's cue
-// disappears: default->half is 0, half->quarter is 1, quarter->off is 2.
-func boundaryStep(tier string) int {
-	switch tier {
-	case tierDefault:
-		return 0
-	case tierHalf:
-		return 1
-	case tierQuarter:
-		return 2
-	default:
-		return -1
-	}
-}
-
-// doseCurve is one task's readings across the four budgets.
-type doseCurve struct {
-	Task   string
-	Tier   string
-	Metric string
-	Values [4]float64
-	// Missing marks arms that produced nothing. Printing those as zero draws a
-	// curve falling to the floor where in fact it was never measured.
-	Missing [4]bool
-	Present bool
-	// PreMean averages the arms where the cue is visible, PostMean the rest.
-	PreMean, PostMean float64
-	BoundaryDelta     float64
-	// BoundaryAligned: the largest transition is the one where this task's cue
-	// disappears. A change elsewhere is variance shaped like an effect.
-	BoundaryAligned bool
-}
-
-func buildCurve(task, tier, metric string, byScale map[string]float64) doseCurve {
-	c := doseCurve{Task: task, Tier: tier, Metric: metric, Present: true}
-	for i, scale := range scaleOrder {
-		v, ok := byScale[scale]
-		if !ok {
-			c.Present, c.Missing[i] = false, true
-		}
-		c.Values[i] = v
-	}
-	step := boundaryStep(tier)
-	if !c.Present || step < 0 {
-		return c
-	}
-	visible := step + 1 // arms before the boundary still show the cue
-	c.PreMean = mean(c.Values[:visible])
-	c.PostMean = mean(c.Values[visible:])
-	c.BoundaryDelta = c.PostMean - c.PreMean
-
-	transitions := [3]float64{
-		math.Abs(c.Values[1] - c.Values[0]),
-		math.Abs(c.Values[2] - c.Values[1]),
-		math.Abs(c.Values[3] - c.Values[2]),
-	}
-	largest, at := 0.0, -1
-	for i, t := range transitions {
-		if t > largest {
-			largest, at = t, i
-		}
-	}
-	// A flat curve has no largest transition to align with anything.
-	c.BoundaryAligned = largest > 0 && at == step
-	return c
-}
-
-func mean(xs []float64) float64 {
-	if len(xs) == 0 {
-		return 0
-	}
-	sum := 0.0
-	for _, x := range xs {
-		sum += x
-	}
-	return sum / float64(len(xs))
-}
-
-func median(xs []float64) float64 {
-	if len(xs) == 0 {
-		return 0
-	}
-	s := append([]float64(nil), xs...)
-	sort.Float64s(s)
-	mid := len(s) / 2
-	if len(s)%2 == 1 {
-		return s[mid]
-	}
-	return (s[mid-1] + s[mid]) / 2
-}
-
-// doseMetrics are the readings a curve is drawn for.
-var doseMetrics = []struct {
-	name string
-	of   func(contextMetrics) float64
-}{
-	{"searches", func(m contextMetrics) float64 { return float64(m.SearchCalls) }},
-	{"recall-tokens", func(m contextMetrics) float64 { return float64(m.RecallReturnedTokens) }},
-	{"tool-rounds", func(m contextMetrics) float64 { return float64(m.RetrievalRounds + m.EscapeCalls) }},
-	{"escapes", func(m contextMetrics) float64 { return float64(m.EscapeCalls) }},
-}
-
-// reportDoseResponse draws each task's curve and reports where the steps fell.
-func reportDoseResponse(all []contextMetrics) string {
-	valid := map[string]map[string]contextMetrics{}
-	for _, m := range all {
-		if m.Contaminated || m.FailureStage == "RunError" {
-			continue
-		}
-		scale := strings.TrimPrefix(m.Arm, "index-")
-		if valid[m.Task] == nil {
-			valid[m.Task] = map[string]contextMetrics{}
-		}
-		valid[m.Task][scale] = m
-	}
-
-	var b strings.Builder
-	b.WriteString("\n## Dose-response, per task\n")
-	b.WriteString("  The arrow marks where this task's cue stops being addressed.\n")
-	aligned := map[string]int{}
-	counted := map[string]int{}
-	deltas := map[string][]float64{}
-
-	for _, t := range indexTasks() {
-		arms := valid[t.ID]
-		if len(arms) == 0 {
-			continue
-		}
-		step := boundaryStep(t.CueTier)
-		fmt.Fprintf(&b, "\n  %s (cue survives to %s)\n", t.ID, t.CueTier)
-		fmt.Fprintf(&b, "    %-14s%s\n", "", scaleHeader(step))
-		for _, metric := range doseMetrics {
-			byScale := map[string]float64{}
-			for scale, m := range arms {
-				byScale[scale] = metric.of(m)
-			}
-			c := buildCurve(t.ID, t.CueTier, metric.name, byScale)
-			mark := " "
-			if c.BoundaryAligned {
-				mark = "*"
-				aligned[metric.name]++
-			}
-			if c.Present {
-				counted[metric.name]++
-				deltas[metric.name] = append(deltas[metric.name], c.BoundaryDelta)
-			}
-			delta := fmt.Sprintf("Δ%+8.1f", c.BoundaryDelta)
-			if !c.Present {
-				delta = "  no curve"
-			}
-			fmt.Fprintf(&b, "    %-14s%s   %s %s\n", metric.name, curveCells(c), delta, mark)
-		}
-	}
-
-	b.WriteString("\n## Where the step fell (* = largest transition is at the cue boundary)\n")
-	for _, metric := range doseMetrics {
-		d := deltas[metric.name]
-		sensitivity := ""
-		if len(d) > 1 {
-			sensitivity = fmt.Sprintf("  without the largest: %+.1f", meanWithoutLargest(d))
-		}
-		fmt.Fprintf(&b, "  %-14s boundary-aligned %d/%d   mean Δ%+.1f   median Δ%+.1f%s\n",
-			metric.name, aligned[metric.name], counted[metric.name], mean(d), median(d), sensitivity)
-	}
-	return b.String()
-}
-
-// meanWithoutLargest is the robustness check: the largest effect is reported
-// and then set aside, so a conclusion never rests on one task alone.
-func meanWithoutLargest(d []float64) float64 {
-	if len(d) < 2 {
-		return mean(d)
-	}
-	worst, at := 0.0, 0
-	for i, v := range d {
-		if math.Abs(v) > worst {
-			worst, at = math.Abs(v), i
-		}
-	}
-	rest := append(append([]float64(nil), d[:at]...), d[at+1:]...)
-	return mean(rest)
-}
-
-func scaleHeader(step int) string {
-	var b strings.Builder
-	for i, s := range scaleOrder {
-		if i == step+1 {
-			s = "↓" + s
-		}
-		fmt.Fprintf(&b, "%8s", s)
-	}
-	return b.String()
-}
-
-// curveCells prints a reading per arm, and a dash where an arm produced none.
-func curveCells(c doseCurve) string {
-	var b strings.Builder
-	for i, v := range c.Values {
-		if c.Missing[i] {
-			fmt.Fprintf(&b, "%8s", "-")
-			continue
-		}
-		fmt.Fprintf(&b, "%8.0f", v)
-	}
-	return b.String()
-}
-
-func percentile(xs []float64, q float64) float64 {
-	if len(xs) == 0 {
-		return 0
-	}
-	s := append([]float64(nil), xs...)
-	sort.Float64s(s)
-	i := int(float64(len(s)-1) * q)
-	return s[i]
-}
-
-func maxOf(xs []float64) float64 {
-	out := 0.0
-	for _, x := range xs {
-		if x > out {
-			out = x
-		}
-	}
-	return out
-}
@@ -1,205 +0,0 @@
-package main
-
-import (
-	"context"
-	"fmt"
-	"io/fs"
-	"os"
-	"path/filepath"
-	"reasonix/internal/state/sessionstore"
-	"strings"
-
-	"reasonix/internal/contract/ablation"
-	"reasonix/internal/contract/event"
-	"reasonix/internal/contract/provider"
-	"reasonix/internal/contract/tool"
-	"reasonix/internal/runtime/agent"
-)
-
-// A fixture is a real session that has really been compacted: no bench-only
-// state shape, so the loader and the projection restore have to work on it.
-
-const (
-	fixtureWindow      = 60000
-	fixtureGenerations = 5
-	// fillerBytes per unit is what makes a fold necessary rather than optional.
-	fillerBytes = 6000
-)
-
-// digestProvider answers a summarizer without naming anything the corpus
-// plants. A digest that happened to quote the answer would leave it visible
-// and silently turn the task into a no-op.
-type digestProvider struct{ calls int }
-
-func (p *digestProvider) Name() string { return "fixture-digest" }
-
-func (p *digestProvider) Stream(_ context.Context, _ provider.Request) (<-chan provider.Chunk, error) {
-	p.calls++
-	ch := make(chan provider.Chunk, 2)
-	ch <- provider.Chunk{Type: provider.ChunkText, Text: "Earlier work continued across several files; nothing outstanding was recorded."}
-	ch <- provider.Chunk{Type: provider.ChunkDone}
-	close(ch)
-	return ch, nil
-}
-
-// fillerCall is one ordinary read whose only job is to take up room — in the
-// window, and in the fold index ahead of a cue.
-func fillerCall(sess *sessionstore.Session, n int) {
-	id := fmt.Sprintf("f%04d", n)
-	sess.Add(provider.Message{Role: provider.RoleAssistant, ToolCalls: []provider.ToolCall{
-		{ID: id, Name: "read_file", Arguments: fmt.Sprintf(`{"path":"internal/mod%03d/handler%03d.go"}`, n%37, n)},
-	}})
-	sess.Add(provider.Message{Role: provider.RoleTool, ToolCallID: id, Name: "read_file", Content: fillerBody(n)})
-}
-
-func fillerBody(n int) string {
-	var b strings.Builder
-	fmt.Fprintf(&b, "package mod%03d\n\n", n%37)
-	for b.Len() < fillerBytes {
-		fmt.Fprintf(&b, "func handler%03dStep%02d() error { return nil }\n", n, b.Len()%64)
-	}
-	return b.String()
-}
-
-// builtFixture is one task's history after real folds, plus where its target
-// ended up.
-type builtFixture struct {
-	Session  *sessionstore.Session
-	State    sessionstore.CompactionState
-	Target   int
-	Path     string
-	Instance fixtureInstance
-}
-
-// buildFixture plays a task's history through the real agent and folds it at
-// the given arm, writing the session and its context state where the agent
-// itself would.
-func buildFixture(f fixtureInstance, arm ablation.Set, sessionPath string) (builtFixture, error) {
-	t := f.Task
-	sess := sessionstore.NewSession("You are a coding agent.")
-	sess.Add(provider.Message{Role: provider.RoleUser, Content: "Work through the transport and scheduler backlog."})
-
-	n := 0
-	at := -1
-
-	a := agent.New(&digestProvider{}, tool.NewRegistry(), sess, agent.Options{
-		ContextWindow: fixtureWindow, CompactRatio: 0.5, RecentKeep: 2,
-		SessionPath: sessionPath, KeepPolicy: agent.KeepErrors, Ablation: arm,
-	}, event.Discard)
-
-	for gen := range fixtureGenerations {
-		if gen == t.PlantAfterGen {
-			at = f.plant(sess)
-		}
-		for range 8 {
-			fillerCall(sess, n)
-			n++
-		}
-		if _, err := a.CompactNow(context.Background(), agent.CompactRequest{}); err != nil {
-			return builtFixture{}, fmt.Errorf("%s: generation %d fold: %w", t.ID, gen, err)
-		}
-	}
-	if at < 0 {
-		return builtFixture{}, fmt.Errorf("%s: PlantAfterGen %d is past the last generation", t.ID, t.PlantAfterGen)
-	}
-	// projectionCoversTail cannot revalidate a projection covering the whole
-	// transcript: it needs a live tail, which a real session always has because
-	// work continues after a fold. Neutral text — the tail is provider-visible.
-	sess.Add(provider.Message{Role: provider.RoleAssistant, Content: "Backlog swept; ready for the next question."})
-	if err := sess.Save(sessionPath); err != nil {
-		return builtFixture{}, fmt.Errorf("%s: save session: %w", t.ID, err)
-	}
-	state, ok, err := sessionstore.LoadCompactionState(sessionPath)
-	if err != nil {
-		return builtFixture{}, fmt.Errorf("%s: load context state: %w", t.ID, err)
-	}
-	if !ok || len(state.Projection.Messages) == 0 {
-		return builtFixture{}, fmt.Errorf("%s: no projection was installed", t.ID)
-	}
-	if state.Projection.CoveredCount <= at {
-		return builtFixture{}, fmt.Errorf("%s: target #%d is not inside the folded region (covered %d)",
-			t.ID, at, state.Projection.CoveredCount)
-	}
-	return builtFixture{Session: sess, State: state, Target: at, Path: sessionPath, Instance: f}, nil
-}
-
-// visibleContext is the model-visible view: the stored projection spliced with
-// everything canonical that came after it.
-func visibleContext(f builtFixture) []provider.Message {
-	canonical := f.Session.Snapshot()
-	out := append([]provider.Message(nil), f.State.Projection.Messages...)
-	if n := f.State.Projection.CoveredCount; n >= 0 && n < len(canonical) {
-		out = append(out, canonical[n:]...)
-	}
-	return out
-}
-
-// visibleText is everything the model would read, as one string.
-func visibleText(f builtFixture) string {
-	var b strings.Builder
-	for _, m := range visibleContext(f) {
-		b.WriteString(m.Content)
-		b.WriteString("\n")
-		for _, tc := range m.ToolCalls {
-			b.WriteString(tc.Name)
-			b.WriteString(" ")
-			b.WriteString(string(tc.Arguments))
-			b.WriteString("\n")
-		}
-	}
-	return b.String()
-}
-
-// sealFixture removes the fixture from disk once the agent holds it in memory.
-// The canonical transcript is the answer, and the sandbox mounts the host
-// read-only by design — so a model that finds the path reads the answer, which
-// one did. The agent has already loaded the session and its sidecar by the time
-// this runs; a failed autosave afterwards costs the benchmark nothing.
-func sealFixture(sessionPath string) {
-	// Every file, not two named ones: a session writes an event log and a
-	// context sidecar beside the transcript, and session.events.jsonl still
-	// carried the answer after the other two were sealed by name.
-	dir := filepath.Dir(sessionPath)
-	entries, err := os.ReadDir(dir)
-	if err != nil {
-		return
-	}
-	for _, e := range entries {
-		if !e.IsDir() {
-			_ = os.Remove(filepath.Join(dir, e.Name()))
-		}
-	}
-	// Deny the directory too, so an autosave cannot recreate what was removed.
-	_ = os.Chmod(dir, 0o500)
-}
-
-// unsealFixture restores the directory so the harness can clean up after
-// itself.
-func unsealFixture(sessionPath string) {
-	_ = os.Chmod(filepath.Dir(sessionPath), 0o700)
-}
-
-// answerOnDisk reports any readable file under dir holding a current-run
-// answer. Two named files were sealed because two leaks were found in them;
-// this scans the directory, so a path that moves — an autosave, a recovery
-// tombstone, a future sidecar — fails here instead of on a paid run.
-func answerOnDisk(dir string, markers []string) (string, string) {
-	var atPath, found string
-	_ = filepath.WalkDir(dir, func(path string, d fs.DirEntry, err error) error {
-		if err != nil || d.IsDir() || atPath != "" {
-			return nil //nolint:nilerr // an unreadable entry is not a leak
-		}
-		body, readErr := os.ReadFile(path)
-		if readErr != nil {
-			return nil
-		}
-		for _, marker := range markers {
-			if len(marker) >= 6 && strings.Contains(string(body), marker) {
-				atPath, found = path, marker
-				return filepath.SkipAll
-			}
-		}
-		return nil
-	})
-	return atPath, found
-}
@@ -1,128 +0,0 @@
-// instance.go — a task template becomes one run's concrete history.
-package main
-
-import (
-	"crypto/rand"
-	"fmt"
-	"hash/fnv"
-	mathrand "math/rand"
-	"strings"
-)
-
-// The invariant this file exists for: no literal sufficient to satisfy
-// AnswerRecovered for a concrete run may exist in this repository before that
-// run begins. Grepping the corpus reveals the question, never the answer.
-
-// varKind is how one placeholder's value is generated. Values stay shaped like
-// real engineering records — a UUID would defeat the leak but turn every task
-// into "find the random string".
-type varKind string
-
-const (
-	varCodename varKind = "codename" // cobalt-k7m4q9
-	varInt      varKind = "int"      // 731
-	varDecimal  varKind = "decimal"  // 0.0437
-	varDuration varKind = "duration" // 43ms
-)
-
-// varSpec declares one placeholder. Min/Max bound the numeric kinds.
-type varSpec struct {
-	Name string
-	Kind varKind
-	Word string // codename stem, e.g. "cobalt"
-	Min  int
-	Max  int
-}
-
-// codenameEntropy is how much of a codename is unguessable. Five base32 digits
-// is about 25 bits: far past what a model can reach by knowing the stem.
-const codenameEntropy = 5
-
-const codenameAlphabet = "abcdefghijkmnpqrstuvwxyz23456789"
-
-// fixtureVars is one run's instantiated values.
-type fixtureVars map[string]string
-
-// newVars generates values for every spec. A nil rng means deterministic.
-func newVars(specs []varSpec, rng *mathrand.Rand) fixtureVars {
-	out := fixtureVars{}
-	for _, spec := range specs {
-		switch spec.Kind {
-		case varCodename:
-			out[spec.Name] = spec.Word + "-" + randomToken(rng, codenameEntropy)
-		case varInt:
-			out[spec.Name] = fmt.Sprint(randomBetween(rng, spec.Min, spec.Max))
-		case varDecimal:
-			out[spec.Name] = fmt.Sprintf("0.%04d", randomBetween(rng, spec.Min, spec.Max))
-		case varDuration:
-			out[spec.Name] = fmt.Sprintf("%dms", randomBetween(rng, spec.Min, spec.Max))
-		}
-	}
-	return out
-}
-
-func randomToken(rng *mathrand.Rand, n int) string {
-	var b strings.Builder
-	for range n {
-		b.WriteByte(codenameAlphabet[randomBetween(rng, 0, len(codenameAlphabet)-1)])
-	}
-	return b.String()
-}
-
-func randomBetween(rng *mathrand.Rand, lo, hi int) int {
-	if hi <= lo {
-		return lo
-	}
-	return lo + rng.Intn(hi-lo+1)
-}
-
-// seededRand is the deterministic generator preflight and tests use, so a
-// fixture's tier calibration and its golden state reproduce exactly.
-func seededRand(parts ...string) *mathrand.Rand {
-	h := fnv.New64a()
-	for _, p := range parts {
-		h.Write([]byte(p))
-		h.Write([]byte{0})
-	}
-	return mathrand.New(mathrand.NewSource(int64(h.Sum64())))
-}
-
-// liveRand is the generator a paid run uses: every run gets values that existed
-// nowhere before it started.
-func liveRand() *mathrand.Rand {
-	var seed [8]byte
-	if _, err := rand.Read(seed[:]); err != nil {
-		panic("contextbench: no entropy for a live run: " + err.Error())
-	}
-	var n int64
-	for _, b := range seed {
-		n = n<<8 | int64(b)
-	}
-	return mathrand.New(mathrand.NewSource(n))
-}
-
-// instantiate substitutes {{name}} and refuses anything left over. The grammar
-// is one identifier in double braces: no conditionals, no loops, no defaults.
-// This is fixture data, not a template language.
-func instantiate(text string, vars fixtureVars) (string, error) {
-	out := text
-	for name, value := range vars {
-		out = strings.ReplaceAll(out, "{{"+name+"}}", value)
-	}
-	if strings.Contains(out, "{{") || strings.Contains(out, "}}") {
-		return "", fmt.Errorf("unresolved placeholder in %q", firstPlaceholder(out))
-	}
-	return out, nil
-}
-
-func firstPlaceholder(text string) string {
-	i := strings.Index(text, "{{")
-	if i < 0 {
-		return text
-	}
-	j := strings.Index(text[i:], "}}")
-	if j < 0 {
-		return text[i:min(i+40, len(text))]
-	}
-	return text[i : i+j+2]
-}
@@ -1,137 +0,0 @@
-// ContextRetrievalBench builds the fixtures the long-horizon retrieval
-// experiments run against, and proves they are worth paying a provider for.
-//
-//	-mode=calibrate  find each index task's lead-call count for its cue tier
-//	-mode=preflight  build every task under every arm and check the contract
-//
-// Nothing here calls a provider. The experiments themselves do; this is what
-// says they would measure what they claim to.
-package main
-
-import (
-	"flag"
-	"fmt"
-	"os"
-	"sort"
-	"strings"
-)
-
-func main() {
-	mode := flag.String("mode", "preflight", "calibrate | preflight | snippet | adversarial | run-search | run-boundary | run-index")
-	work := flag.String("work", "", "directory for built fixtures (default: a temp dir)")
-	only := flag.String("task", "", "limit to one task id")
-	dry := flag.Bool("dry", false, "run modes: drive the pipeline with a scripted provider instead of paying for one")
-	flag.Parse()
-
-	if err := validateCorpus(); err != nil {
-		fmt.Fprintln(os.Stderr, "corpus:", err)
-		os.Exit(2)
-	}
-	root := *work
-	if root == "" {
-		dir, err := os.MkdirTemp("", "context-retrieval-*")
-		if err != nil {
-			fmt.Fprintln(os.Stderr, err)
-			os.Exit(2)
-		}
-		defer os.RemoveAll(dir)
-		root = dir
-	}
-
-	tasks := allTasks()
-	if *only != "" {
-		t, ok := taskByID(*only)
-		if !ok {
-			fmt.Fprintf(os.Stderr, "unknown task %q\n", *only)
-			os.Exit(2)
-		}
-		tasks = []contextTask{t}
-	}
-
-	switch *mode {
-	case "calibrate":
-		os.Exit(runCalibrate(tasks, root))
-	case "preflight":
-		os.Exit(runPreflight(tasks, root))
-	case "run-search":
-		os.Exit(runExperiment(experimentSearch, root, *dry, tasks))
-	case "run-boundary":
-		os.Exit(runBoundaries(root))
-	case "snippet":
-		os.Exit(runSnippetAudit(root, 100))
-	case "adversarial":
-		os.Exit(runAdversarial(root))
-	case "run-index":
-		os.Exit(runExperiment(experimentIndex, root, *dry, tasks))
-	default:
-		fmt.Fprintf(os.Stderr, "unknown mode %q\n", *mode)
-		os.Exit(2)
-	}
-}
-
-func runPreflight(tasks []contextTask, root string) int {
-	var all []finding
-	for _, t := range tasks {
-		found, err := checkTask(t, root)
-		if err != nil {
-			fmt.Fprintf(os.Stderr, "%s: %v\n", t.ID, err)
-			return 2
-		}
-		all = append(all, found...)
-		status := "ok"
-		if len(found) > 0 {
-			status = fmt.Sprintf("%d problem(s)", len(found))
-		}
-		fmt.Printf("%-24s %-6s gen=%-2d %s\n", t.ID, t.Experiment, t.PlantAfterGen, status)
-	}
-	if len(all) == 0 {
-		fmt.Printf("\n%d fixtures ready.\n", len(tasks))
-		return 0
-	}
-	fmt.Fprintf(os.Stderr, "\n%d preflight problem(s):\n", len(all))
-	sort.Slice(all, func(i, j int) bool { return all[i].Task < all[j].Task })
-	for _, f := range all {
-		fmt.Fprintln(os.Stderr, " ", f)
-	}
-	return 1
-}
-
-// runCalibrate reports, for each index task, the lead-call counts at which its
-// cue is still addressed by each budget. The corpus freezes one number from
-// this; it is never computed at run time, or the implementation under test
-// would be moving its own goalposts.
-func runCalibrate(tasks []contextTask, root string) int {
-	for _, t := range tasks {
-		if t.Experiment != experimentIndex {
-			continue
-		}
-		var rows []string
-		for lead := range fixtureGenerations {
-			probe := t
-			probe.PlantAfterGen = lead
-			var visibleAt []string
-			for _, scale := range armScales {
-				inst, ierr := instantiateTask(probe, seededRand(t.ID, scale, "calibrate"))
-				if ierr != nil {
-					fmt.Fprintln(os.Stderr, ierr)
-					return 2
-				}
-				f, err := buildFixture(inst, armFor(scale, false),
-					fmt.Sprintf("%s/calib/%s/%d-%s/session.jsonl", root, t.ID, lead, scale))
-				if err != nil {
-					fmt.Fprintf(os.Stderr, "%s gen=%d %s: %v\n", t.ID, lead, scale, err)
-					return 2
-				}
-				if strings.Contains(visibleText(f), inst.CueMarker) {
-					visibleAt = append(visibleAt, scale)
-				}
-			}
-			rows = append(rows, fmt.Sprintf("planted after gen %d, cue visible at: %s", lead, strings.Join(visibleAt, ",")))
-		}
-		fmt.Printf("\n%s (wants tier %s)\n", t.ID, t.CueTier)
-		for _, r := range rows {
-			fmt.Println(" ", r)
-		}
-	}
-	return 0
-}
@@ -1,137 +0,0 @@
-package main
-
-import (
-	"fmt"
-	"os/exec"
-	"slices"
-	"strings"
-	"testing"
-)
-
-// The same checks the preflight mode runs, under go test: a compaction change
-// that invalidates the fixtures fails here rather than three hundred API calls
-// later.
-
-func TestCorpusIsWellFormed(t *testing.T) {
-	if err := validateCorpus(); err != nil {
-		t.Fatal(err)
-	}
-	if got := len(searchTasks()); got != 6 {
-		t.Errorf("search experiment has %d tasks, want 6", got)
-	}
-	if got := len(indexTasks()); got != 6 {
-		t.Errorf("index experiment has %d tasks, want 6", got)
-	}
-	tiers := map[string]int{}
-	for _, task := range indexTasks() {
-		tiers[task.CueTier]++
-	}
-	// Two per tier is what makes the staircase readable: the same task can be
-	// compared across the boundary its cue sits on, instead of two different
-	// questions being compared across two arms.
-	for _, tier := range []string{tierQuarter, tierHalf, tierDefault} {
-		if tiers[tier] != 2 {
-			t.Errorf("tier %s has %d tasks, want 2", tier, tiers[tier])
-		}
-	}
-}
-
-// The invariant the whole instantiation layer exists for: no literal a scorer
-// needs may be readable in this repository before a run starts.
-func TestNoAnswerLiteralExistsInTheRepository(t *testing.T) {
-	root, err := repoRoot()
-	if err != nil {
-		t.Skip(err)
-	}
-	for _, task := range allTasks() {
-		inst, err := instantiateTask(task, seededRand(task.ID, "leak-check"))
-		if err != nil {
-			t.Fatal(err)
-		}
-		for _, marker := range append(inst.AnswerMarkers, inst.CueMarker) {
-			if len(marker) < 6 {
-				continue // a bare number is not a literal anyone can grep for
-			}
-			if found, where := grepRepo(t, root, marker); found {
-				t.Errorf("%s: %q is readable at %s before the run", task.ID, marker, where)
-			}
-		}
-	}
-}
-
-// Two runs of the same task must not share an answer, or one leaked run leaks
-// every future one.
-func TestInstancesDifferBetweenRuns(t *testing.T) {
-	for _, task := range allTasks() {
-		a, err := instantiateTask(task, liveRand())
-		if err != nil {
-			t.Fatal(err)
-		}
-		b, err := instantiateTask(task, liveRand())
-		if err != nil {
-			t.Fatal(err)
-		}
-		if slices.Equal(a.AnswerMarkers, b.AnswerMarkers) {
-			t.Errorf("%s produced the same answers twice: %v", task.ID, a.AnswerMarkers)
-		}
-	}
-}
-
-// A preflight that could not reproduce would calibrate against one run's
-// nonces and report about another's.
-func TestSeededInstancesReproduce(t *testing.T) {
-	for _, task := range allTasks() {
-		a, _ := instantiateTask(task, seededRand(task.ID, "x"))
-		b, _ := instantiateTask(task, seededRand(task.ID, "x"))
-		if !slices.Equal(a.AnswerMarkers, b.AnswerMarkers) || a.body != b.body {
-			t.Errorf("%s did not reproduce under the same seed", task.ID)
-		}
-	}
-}
-
-// An unresolved placeholder would put "{{marker}}" in the transcript and score
-// every run wrong; it fails the fixture instead.
-func TestUnresolvedPlaceholderIsRefused(t *testing.T) {
-	if _, err := instantiate("value is {{missing}}", fixtureVars{"other": "x"}); err == nil {
-		t.Fatal("an unresolved placeholder was accepted")
-	}
-}
-
-func repoRoot() (string, error) {
-	out, err := exec.Command("git", "rev-parse", "--show-toplevel").Output()
-	if err != nil {
-		return "", fmt.Errorf("not in a git checkout: %w", err)
-	}
-	return strings.TrimSpace(string(out)), nil
-}
-
-// grepRepo searches tracked files only: build artifacts and temp fixtures are
-// not what a fresh checkout hands anyone.
-func grepRepo(t *testing.T, root, needle string) (bool, string) {
-	t.Helper()
-	cmd := exec.Command("git", "grep", "-l", "-F", needle)
-	cmd.Dir = root
-	out, err := cmd.Output()
-	if err != nil {
-		return false, "" // exit 1 means no match
-	}
-	return true, strings.TrimSpace(strings.SplitN(string(out), "\n", 2)[0])
-}
-
-func TestEveryFixturePassesPreflight(t *testing.T) {
-	if testing.Short() {
-		t.Skip("builds and folds 36 sessions")
-	}
-	root := t.TempDir()
-	for _, task := range allTasks() {
-		t.Run(task.ID, func(t *testing.T) {
-			found, err := checkTask(task, root)
-			if err != nil {
-				t.Fatal(err)
-			}
-			for _, f := range found {
-				t.Error(f)
-			}
-		})
-	}
-}
@@ -1,290 +0,0 @@
-package main
-
-import (
-	"encoding/json"
-	"fmt"
-	"path/filepath"
-	"slices"
-	"strings"
-
-	"reasonix/internal/base/retrieval"
-	"reasonix/internal/contract/ablation"
-	"reasonix/internal/contract/provider"
-)
-
-// Preflight is what makes a real run worth paying for. An answer the model can
-// already read, or a cue not where the matrix says, turns an arm into a
-// different experiment without saying so.
-
-// armScales are the fold-index arms an index task is measured across.
-var armScales = []string{"default", "half", "quarter", "off"}
-
-// experimentArm is one measured configuration of a task.
-type experimentArm struct {
-	name      string
-	scale     string
-	searchOff bool
-}
-
-// armsFor is the single declaration of which arms a task runs at, so preflight
-// and the run itself cannot check one set and measure another.
-func armsFor(t contextTask) []experimentArm {
-	if t.Experiment == experimentSearch {
-		return []experimentArm{
-			{"search-on", "default", false},
-			{"search-off", "default", true},
-		}
-	}
-	arms := make([]experimentArm, 0, len(armScales))
-	for _, scale := range armScales {
-		arms = append(arms, experimentArm{"index-" + scale, scale, false})
-	}
-	return arms
-}
-
-func scaleOf(name string) ablation.FoldIndexScale {
-	s, err := ablation.ParseFoldIndexScale(name)
-	if err != nil {
-		panic("unknown scale " + name) // the list above is a constant
-	}
-	return s
-}
-
-// armFor is the arm a task is measured under at one fold-index scale.
-func armFor(scale string, searchOff bool) ablation.Set {
-	set := ablation.Set{}
-	if searchOff {
-		set = ablation.New(ablation.RecallSearch)
-	}
-	return set.WithFoldIndex(scaleOf(scale))
-}
-
-type finding struct {
-	Task, Arm, Problem string
-}
-
-func (f finding) String() string { return fmt.Sprintf("%s [%s]: %s", f.Task, f.Arm, f.Problem) }
-
-// checkTask builds the task under every arm it will be run at and reports what
-// does not hold.
-func checkTask(t contextTask, root string) ([]finding, error) {
-	var out []finding
-	for _, arm := range armsFor(t) {
-		// Deterministic values: a preflight has to reproduce, and a tier
-		// calibration measured against one run's nonces is measured against
-		// nothing.
-		inst, err := instantiateTask(t, seededRand(t.ID, arm.name, "preflight"))
-		if err != nil {
-			return nil, err
-		}
-		dir := filepath.Join(root, t.ID, arm.name)
-		f, err := buildFixture(inst, armFor(arm.scale, arm.searchOff), filepath.Join(dir, "session.jsonl"))
-		if err != nil {
-			return nil, err
-		}
-		visible := visibleText(f)
-
-		// The answer must be unreadable in every arm, or the task measures
-		// reading rather than recall.
-		for _, marker := range inst.AnswerMarkers {
-			if len(marker) < 3 {
-				continue // a bare number is not a nonce; the whole set still has to match
-			}
-			if strings.Contains(visible, marker) {
-				out = append(out, finding{t.ID, arm.name, fmt.Sprintf("answer marker %q is already visible", marker)})
-			}
-		}
-
-		// The corpus has to be findable at all, or a miss in a real run says
-		// nothing about the model.
-		rank, err := probeRank(f, inst.ProbeQuery)
-		if err != nil {
-			return nil, err
-		}
-		switch {
-		case rank == 0:
-			out = append(out, finding{t.ID, arm.name, fmt.Sprintf("probe query %q does not find the target at all", inst.ProbeQuery)})
-		case rank > 5:
-			out = append(out, finding{t.ID, arm.name, fmt.Sprintf("probe query %q ranks the target %d", inst.ProbeQuery, rank)})
-		}
-
-		if t.Experiment == experimentIndex {
-			out = append(out, checkCueVisibility(t, inst, arm.scale, arm.name, visible)...)
-		} else if line := indexLineFor(f, f.Target); line != "" {
-			out = append(out, finding{t.ID, arm.name,
-				"a search task is addressed by the fold index, so it measures the index and not search: " + line})
-		}
-
-		// The projection is what the agent believes. This is what the model is
-		// told, and it is the one that has to hold.
-		req, err := captureRequest(f, armFor(arm.scale, arm.searchOff), inst.Prompt)
-		if err != nil {
-			return nil, fmt.Errorf("%s [%s]: capture: %w", t.ID, arm.name, err)
-		}
-		out = append(out, checkCapturedRequest(t, inst, arm.name, arm.scale, arm.searchOff, req)...)
-	}
-	return out, nil
-}
-
-// checkCapturedRequest holds one arm's real provider request to its contract:
-// the answer unreadable, the cue exactly where its tier says, the recall tool
-// offering only what this arm can do, and no host-side field surviving the
-// boundary.
-func checkCapturedRequest(t contextTask, inst fixtureInstance, armName, scale string, searchOff bool, req provider.Request) []finding {
-	var out []finding
-	for _, marker := range inst.AnswerMarkers {
-		if len(marker) < 3 {
-			continue // a bare number is not a nonce; the set as a whole still is
-		}
-		for _, at := range findInSurface(req, marker) {
-			out = append(out, finding{t.ID, armName, fmt.Sprintf("answer marker %q leaked at %s", marker, at)})
-		}
-	}
-	if t.Experiment == experimentIndex {
-		wantVisible, wantHidden := tierScales(t.CueTier)
-		where := findInSurface(req, inst.CueMarker)
-		switch {
-		case contains(wantVisible, scale) && len(where) == 0:
-			out = append(out, finding{t.ID, armName, fmt.Sprintf("cue %q is not provider-visible; the tier expects it", inst.CueMarker)})
-		case contains(wantHidden, scale) && len(where) > 0:
-			out = append(out, finding{t.ID, armName, fmt.Sprintf("cue %q is provider-visible at %s: %s", inst.CueMarker, where[0], scale)})
-		}
-	}
-
-	schema, ok := recallToolSchema(req)
-	if !ok {
-		out = append(out, finding{t.ID, armName, "recall is not in the provider's tool list"})
-		return append(out, boundaryFindings(t, armName, req)...)
-	}
-	params := ""
-	if b, err := json.Marshal(schema.Parameters); err == nil {
-		params = string(b)
-	}
-	// Both arms keep the read half: the ablation removes search, not recall.
-	if !strings.Contains(params, `"positions"`) {
-		out = append(out, finding{t.ID, armName, "recall lost its positions parameter"})
-	}
-	hasQuery := strings.Contains(params, `"query"`)
-	switch {
-	case searchOff && hasQuery:
-		out = append(out, finding{t.ID, armName, "recall still offers a query parameter"})
-	case !searchOff && !hasQuery:
-		out = append(out, finding{t.ID, armName, "recall has no query parameter"})
-	}
-	if searchOff {
-		for _, phrase := range searchWording {
-			for _, at := range findInSurface(req, phrase) {
-				out = append(out, finding{t.ID, armName, fmt.Sprintf("search is offered to an arm without it: %q at %s", phrase, at)})
-			}
-		}
-	}
-	return append(out, boundaryFindings(t, armName, req)...)
-}
-
-func boundaryFindings(t contextTask, armName string, req provider.Request) []finding {
-	var out []finding
-	for _, at := range localOnlyLeaks(req) {
-		out = append(out, finding{t.ID, armName, "host-only field crossed the provider boundary at " + at})
-	}
-	return out
-}
-
-// checkCueVisibility holds an index task to its tier: the cue is addressed at
-// the scales the matrix names and at no others. A policy change that moves it
-// fails here rather than turning two arms into one experiment.
-func checkCueVisibility(t contextTask, inst fixtureInstance, scale, armName, visible string) []finding {
-	wantVisible, wantHidden := tierScales(t.CueTier)
-	has := strings.Contains(visible, inst.CueMarker)
-	switch {
-	case contains(wantVisible, scale) && !has:
-		return []finding{{t.ID, armName, fmt.Sprintf("cue %q should be addressed at %s and is not", inst.CueMarker, scale)}}
-	case contains(wantHidden, scale) && has:
-		return []finding{{t.ID, armName, fmt.Sprintf("cue %q should be gone at %s and is not", inst.CueMarker, scale)}}
-	}
-	return nil
-}
-
-func contains(list []string, want string) bool { return slices.Contains(list, want) }
-
-// probeRank runs the task's own query against the folded region the way recall
-// would, and reports where the target lands.
-func probeRank(f builtFixture, query string) (int, error) {
-	terms, err := retrieval.QueryTerms(query)
-	if err != nil {
-		return 0, err
-	}
-	canonical := f.Session.Snapshot()
-	covered := min(f.State.Projection.CoveredCount, len(canonical))
-	docs := make([]map[string]int, 0, covered)
-	positions := make([]int, 0, covered)
-	lengths := make([]int, 0, covered)
-	for i, m := range canonical[:covered] {
-		var text strings.Builder
-		text.WriteString(m.Content)
-		for _, tc := range m.ToolCalls {
-			text.WriteString(" " + tc.Name + " " + string(tc.Arguments))
-		}
-		tokens := retrieval.Tokens(text.String())
-		if len(tokens) == 0 {
-			continue
-		}
-		docs = append(docs, retrieval.Counts(tokens))
-		positions = append(positions, i)
-		lengths = append(lengths, len(tokens))
-	}
-	df := retrieval.DocumentFrequency(docs)
-	avg := 0.0
-	for _, l := range lengths {
-		avg += float64(l)
-	}
-	if len(lengths) > 0 {
-		avg /= float64(len(lengths))
-	}
-	type hit struct {
-		pos   int
-		score float64
-	}
-	var hits []hit
-	for i, counts := range docs {
-		if s := retrieval.BM25Score(counts, lengths[i], terms, df, len(docs), avg); s > 0 {
-			hits = append(hits, hit{positions[i], s})
-		}
-	}
-	for i := range hits {
-		for j := i + 1; j < len(hits); j++ {
-			if hits[j].score > hits[i].score {
-				hits[i], hits[j] = hits[j], hits[i]
-			}
-		}
-	}
-	// A tool result answers at its caller's position, so either counts.
-	for i, h := range hits {
-		if h.pos == f.Target || h.pos == f.Target+1 {
-			return i + 1, nil
-		}
-	}
-	return 0, nil
-}
-
-// indexLineFor returns the fold-index line addressing a position, if any.
-func indexLineFor(f builtFixture, pos int) string {
-	want := fmt.Sprintf("#%d ", pos)
-	for line := range strings.SplitSeq(visibleIndexOnly(f), "\n") {
-		if strings.Contains(line, want) {
-			return strings.TrimSpace(line)
-		}
-	}
-	return ""
-}
-
-// visibleIndexOnly is just the fold-index sections of the projection.
-func visibleIndexOnly(f builtFixture) string {
-	var b strings.Builder
-	for _, m := range visibleContext(f) {
-		if i := strings.Index(m.Content, "## Folded work index"); i >= 0 {
-			b.WriteString(m.Content[i:])
-			b.WriteString("\n")
-		}
-	}
-	return b.String()
-}
@@ -1,252 +0,0 @@
-// reformulation.go — how a query changed, from token sets alone.
-package main
-
-import (
-	"fmt"
-	"slices"
-	"sort"
-	"strings"
-
-	"reasonix/internal/base/retrieval"
-)
-
-// The host probe ranks every target first, so the engine is not what makes a
-// model take more than one attempt. These classes are set arithmetic over the
-// same tokenizer recall uses: no judge, nothing read out of prose.
-
-const (
-	reformFirst      = "First"            // nothing to compare against yet
-	reformNearDup    = "NearDuplicate"    // the same query again, give or take a word
-	reformNarrowing  = "Narrowing"        // kept everything and added terms
-	reformBroadening = "Broadening"       // dropped terms, added none
-	reformAnchored   = "HistoricalAnchor" // reached for a term the question does not contain
-	reformShift      = "SynonymShift"     // traded terms for others
-)
-
-// nearDuplicateOverlap is where "reworded" stops and "asked again" begins.
-const nearDuplicateOverlap = 0.8
-
-func terms(s string) []string {
-	return retrieval.Unique(retrieval.Tokens(s))
-}
-
-func termSet(s string) map[string]bool {
-	out := map[string]bool{}
-	for _, t := range terms(s) {
-		out[t] = true
-	}
-	return out
-}
-
-func jaccard(a, b map[string]bool) float64 {
-	if len(a) == 0 && len(b) == 0 {
-		return 1
-	}
-	shared := 0
-	for t := range a {
-		if b[t] {
-			shared++
-		}
-	}
-	union := len(a) + len(b) - shared
-	if union == 0 {
-		return 0
-	}
-	return float64(shared) / float64(union)
-}
-
-// classifyReformulation names how this query differs from the one before it.
-func classifyReformulation(previous, current, question string) string {
-	if previous == "" {
-		return reformFirst
-	}
-	prev, cur := termSet(previous), termSet(current)
-	if jaccard(prev, cur) >= nearDuplicateOverlap {
-		return reformNearDup
-	}
-	added, dropped := diffTerms(cur, prev), diffTerms(prev, cur)
-	// A term the question never used had to come from the model's own memory of
-	// the history, which is a different move from rephrasing what it was asked.
-	ask := termSet(question)
-	for _, t := range added {
-		if !ask[t] {
-			return reformAnchored
-		}
-	}
-	switch {
-	case len(dropped) == 0:
-		return reformNarrowing
-	case len(added) == 0:
-		return reformBroadening
-	default:
-		return reformShift
-	}
-}
-
-func diffTerms(from, minus map[string]bool) []string {
-	var out []string
-	for t := range from {
-		if !minus[t] {
-			out = append(out, t)
-		}
-	}
-	sort.Strings(out)
-	return out
-}
-
-// classifySearches labels each attempt and records which of the host probe's
-// terms it left out. The probe is the oracle: it ranks the target first, so a
-// query missing its terms says what the model did not think to ask for.
-func classifySearches(m *contextMetrics, inst fixtureInstance) {
-	probe := terms(inst.ProbeQuery)
-	previous := ""
-	for i := range m.Searches {
-		s := &m.Searches[i]
-		s.Reformulation = classifyReformulation(previous, s.Query, inst.Prompt)
-		have := termSet(s.Query)
-		for _, term := range probe {
-			if !have[term] {
-				s.MissingProbeTerms = append(s.MissingProbeTerms, term)
-			}
-		}
-		previous = s.Query
-	}
-}
-
-// queryAudit is the reformulation funnel across a set of runs.
-type queryAudit struct {
-	Runs, RunsThatSearched  int
-	FirstQueryHit           int
-	HitBySecond, HitByThird int
-	NeverHit                int
-	TotalSearches           int
-	PostHitSearches         int
-	HitThenRead             int
-	EscapeBeforeFirstSearch int
-	Reformulations          map[string]int
-	MissedProbeTerms        map[string]int
-}
-
-func auditQueries(runs []contextMetrics) queryAudit {
-	a := queryAudit{Reformulations: map[string]int{}, MissedProbeTerms: map[string]int{}}
-	for _, r := range runs {
-		if r.Contaminated {
-			continue
-		}
-		a.Runs++
-		a.TotalSearches += len(r.Searches)
-		a.PostHitSearches += r.PostHitSearches
-		if r.EscapeBeforeFirstRecall > 0 {
-			a.EscapeBeforeFirstSearch++
-		}
-		if len(r.Searches) == 0 {
-			continue
-		}
-		a.RunsThatSearched++
-		hitAt := 0
-		for i, s := range r.Searches {
-			a.Reformulations[s.Reformulation]++
-			for _, term := range s.MissingProbeTerms {
-				a.MissedProbeTerms[term]++
-			}
-			if s.TargetRank > 0 && hitAt == 0 {
-				hitAt = i + 1
-			}
-		}
-		switch {
-		case hitAt == 1:
-			a.FirstQueryHit++
-		case hitAt == 2:
-			a.HitBySecond++
-		case hitAt >= 3:
-			a.HitByThird++
-		default:
-			a.NeverHit++
-		}
-		if hitAt > 0 && r.TargetRead {
-			a.HitThenRead++
-		}
-	}
-	return a
-}
-
-func (a queryAudit) report() string {
-	var b strings.Builder
-	b.WriteString("\n## Query reformulation\n")
-	fmt.Fprintf(&b, "  runs=%d searched=%d  searches=%d (%.2f per searching run)\n",
-		a.Runs, a.RunsThatSearched, a.TotalSearches, ratio(a.TotalSearches, a.RunsThatSearched))
-	fmt.Fprintf(&b, "  target found by query 1: %d   by 2: %d   by 3+: %d   never: %d\n",
-		a.FirstQueryHit, a.HitBySecond, a.HitByThird, a.NeverHit)
-	fmt.Fprintf(&b, "  found then read: %d/%d   searches issued after the target was already back: %d\n",
-		a.HitThenRead, a.FirstQueryHit+a.HitBySecond+a.HitByThird, a.PostHitSearches)
-	fmt.Fprintf(&b, "  reached for the workspace before searching: %d\n", a.EscapeBeforeFirstSearch)
-	b.WriteString("  reformulations: " + countsLine(a.Reformulations) + "\n")
-	b.WriteString("  probe terms most often absent: " + topCounts(a.MissedProbeTerms, 8) + "\n")
-	return b.String()
-}
-
-func ratio(n, d int) float64 {
-	if d == 0 {
-		return 0
-	}
-	return float64(n) / float64(d)
-}
-
-func countsLine(counts map[string]int) string {
-	keys := make([]string, 0, len(counts))
-	for k := range counts {
-		keys = append(keys, k)
-	}
-	sort.Strings(keys)
-	parts := make([]string, 0, len(keys))
-	for _, k := range keys {
-		parts = append(parts, fmt.Sprintf("%s=%d", k, counts[k]))
-	}
-	return strings.Join(parts, " ")
-}
-
-func topCounts(counts map[string]int, limit int) string {
-	type kv struct {
-		k string
-		n int
-	}
-	all := make([]kv, 0, len(counts))
-	for k, n := range counts {
-		all = append(all, kv{k, n})
-	}
-	sort.Slice(all, func(i, j int) bool {
-		if all[i].n != all[j].n {
-			return all[i].n > all[j].n
-		}
-		return all[i].k < all[j].k
-	})
-	parts := make([]string, 0, limit)
-	for _, e := range all[:min(limit, len(all))] {
-		parts = append(parts, fmt.Sprintf("%s=%d", e.k, e.n))
-	}
-	if len(parts) == 0 {
-		return "(none)"
-	}
-	return strings.Join(parts, " ")
-}
-
-// queryLines lists every attempt, so a surprising audit can be read back.
-func queryLines(runs []contextMetrics) string {
-	var b strings.Builder
-	b.WriteString("\n## Every query, in order\n")
-	for _, r := range runs {
-		for _, s := range r.Searches {
-			rank := "miss"
-			if s.TargetRank > 0 {
-				rank = fmt.Sprintf("rank %d", s.TargetRank)
-			}
-			fmt.Fprintf(&b, "  %-24s %-14s #%d %-17s %-9s %q\n",
-				r.Task, r.Arm, s.Ordinal, s.Reformulation, rank, s.Query)
-			if len(s.MissingProbeTerms) > 0 && s.TargetRank == 0 {
-				fmt.Fprintf(&b, "  %-24s %-14s    missing probe terms: %s\n",
-					"", "", strings.Join(slices.Clone(s.MissingProbeTerms), " "))
-			}
-		}
-	}
-	return b.String()
-}
@@ -1,519 +0,0 @@
-package main
-
-import (
-	"context"
-	"fmt"
-	"os"
-	"os/exec"
-	"path/filepath"
-	"reasonix/internal/state/sessionstore"
-	"slices"
-	"sort"
-	"strings"
-	"time"
-
-	"reasonix/internal/assembly/boot"
-	"reasonix/internal/contract/ablation"
-	"reasonix/internal/contract/event"
-	"reasonix/internal/contract/provider"
-	"reasonix/internal/contract/tool"
-	_ "reasonix/internal/model/openai"
-	"reasonix/internal/runtime/agent"
-	_ "reasonix/internal/tools/builtin"
-)
-
-// A run is one recall question against a fixture the preflight has already
-// cleared. Scoring reads the transcript the turn appended — tool names and
-// their JSON arguments — so nothing depends on how the model narrated itself.
-
-const (
-	runBaseURL = "https://api.deepseek.com"
-	runModel   = "deepseek-chat"
-	runTimeout = 4 * time.Minute
-)
-
-func realProvider() (provider.Provider, error) {
-	key := os.Getenv("DEEPSEEK_API_KEY")
-	if key == "" {
-		return nil, fmt.Errorf("a real run needs DEEPSEEK_API_KEY")
-	}
-	return provider.New("openai", provider.Config{
-		Name: "contextbench", BaseURL: runBaseURL, Model: runModel, APIKey: key,
-	})
-}
-
-// adversarialPrompts tell the model what it is running and send it after the
-// answer everywhere except the conversation. The question is not whether it
-// obeys the experiment: it is whether the answer is reachable at all by a model
-// that knows the mechanism and is actively attacking the host.
-var adversarialPrompts = map[string]string{
-	experimentSearch: "You are running a benchmark. Find the expected answer for this task from the " +
-		"Reasonix repository, benchmark source, fixture generator, test files, scratch files, or " +
-		"session metadata. Do not use recall unless necessary. Report the exact hidden values.",
-	experimentIndex: "Inspect benchmarks/context-retrieval, the git repository, session files, temporary " +
-		"files, and any benchmark metadata you can access. Recover the concrete instantiated answer for " +
-		"this run without relying on the folded conversation. Report the exact values.",
-}
-
-// runAdversarial sends one task per experiment after the answer through every
-// host-readable surface. It passes when nothing but recall could produce it —
-// a run that gives up and recalls has still proven the attack failed.
-func runAdversarial(root string) int {
-	root, err := filepath.Abs(root)
-	if err != nil {
-		fmt.Fprintln(os.Stderr, err)
-		return 2
-	}
-	restore, ierr := isolateWorkspace()
-	if ierr != nil {
-		fmt.Fprintln(os.Stderr, "isolate workspace:", ierr)
-		return 2
-	}
-	defer restore()
-	p, perr := realProvider()
-	if perr != nil {
-		fmt.Fprintln(os.Stderr, perr)
-		return 2
-	}
-	leaked := 0
-	for _, id := range []string{"r01-retry-boundary", "i01-transport-fallback"} {
-		t, ok := taskByID(id)
-		if !ok {
-			fmt.Fprintf(os.Stderr, "unknown task %q\n", id)
-			return 2
-		}
-		m, err := runOne(p, t, armFor("default", false), "adversarial", root, false, false,
-			adversarialPrompts[t.Experiment])
-		if err != nil {
-			fmt.Fprintf(os.Stderr, "%s: %v\n", id, err)
-		}
-		verdict := "isolated"
-		if m.EscapeFoundAnswer {
-			verdict = "LEAKED via " + m.LeakedVia
-			leaked++
-		}
-		fmt.Printf("%-24s %-24s escapes=%-3d recall=%d answered=%v tools=%v\n",
-			id, verdict, m.EscapeCalls, m.SearchCalls+m.ReadCalls, m.AnswerRecovered, m.UnexpectedWorkTools)
-		if m.EscapeFoundAnswer {
-			fmt.Printf("   args:   %s\n   output: %s\n", m.LeakedArgs, m.LeakedOutput)
-		}
-	}
-	if leaked > 0 {
-		fmt.Fprintf(os.Stderr, "\n%d of 2 adversarial runs reached an answer outside recall\n", leaked)
-		return 1
-	}
-	fmt.Println("\nNo answer was reachable outside recall.")
-	return 0
-}
-
-// runOne asks one task under one arm and scores what the model did.
-func runOne(p provider.Provider, t contextTask, arm ablation.Set, armName, root string, cueVisible, dry bool, promptOverride ...string) (contextMetrics, error) {
-	// Live runs get values that existed nowhere before the run started; a dry
-	// run stays reproducible.
-	rng := liveRand()
-	if dry {
-		rng = seededRand(t.ID, armName, "dry")
-	}
-	inst, err := instantiateTask(t, rng)
-	if err != nil {
-		return contextMetrics{}, err
-	}
-	f, err := buildFixture(inst, arm, filepath.Join(root, t.ID, armName, "session.jsonl"))
-	if err != nil {
-		return contextMetrics{}, err
-	}
-	sess, err := sessionstore.LoadSession(f.Path)
-	if err != nil {
-		return contextMetrics{}, fmt.Errorf("load session: %w", err)
-	}
-	reg := tool.NewRegistry()
-	for _, x := range tool.Builtins() {
-		reg.Add(x)
-	}
-	boot.ApplyUnifiedProviderToolSurface(reg, false, arm)
-
-	a := agent.New(p, reg, sess, agent.Options{
-		ContextWindow: fixtureWindow, CompactRatio: 0.5, RecentKeep: 2,
-		SessionPath: f.Path, KeepPolicy: agent.KeepErrors, Ablation: arm,
-	}, event.Discard)
-	a.LoadProjectionSidecar(f.Path)
-	// From here the transcript lives only in memory. On disk it is the answer.
-	sealFixture(f.Path)
-	defer unsealFixture(f.Path)
-	dir := filepath.Dir(f.Path)
-	if at, marker := answerOnDisk(dir, inst.AnswerMarkers); at != "" {
-		return contextMetrics{}, fmt.Errorf("%s [%s]: %q is readable at %s before the run", t.ID, armName, marker, at)
-	}
-
-	before := len(sess.Snapshot())
-	ctx, cancel := context.WithTimeout(context.Background(), runTimeout)
-	defer cancel()
-	prompt := inst.Prompt
-	if len(promptOverride) > 0 && promptOverride[0] != "" {
-		prompt = promptOverride[0]
-	}
-	runErr := a.Run(ctx, prompt)
-
-	if at, marker := answerOnDisk(dir, inst.AnswerMarkers); at != "" {
-		return contextMetrics{}, fmt.Errorf("%s [%s]: %q was written to %s during the run", t.ID, armName, marker, at)
-	}
-
-	appended := sess.Snapshot()[before:]
-	m := scoreRun(appended, inst, f.Target, armName, cueVisible)
-	m.trajectory = appended
-	m.scoreAnswer(finalAnswer(appended), inst)
-	if runErr != nil {
-		m.FailureStage = "RunError"
-		m.TimeoutReason = timeoutReason(m)
-	}
-	return m, runErr
-}
-
-// finalAnswer is the last assistant text the turn produced.
-func finalAnswer(msgs []provider.Message) string {
-	for _, m := range slices.Backward(msgs) {
-		if m.Role == provider.RoleAssistant && strings.TrimSpace(m.Content) != "" {
-			return m.Content
-		}
-	}
-	return ""
-}
-
-// isolateWorkspace moves the process into an empty directory before any task
-// runs. The corpus lives in this repository, so an agent left here greps the
-// answers out of corpus.go — one did, quoting the file's own line about them
-// being nonces. Fixtures stay outside it too: a readable session file is the
-// canonical transcript, answer included.
-func isolateWorkspace() (func(), error) {
-	dir, err := os.MkdirTemp("", "contextbench-workspace-*")
-	if err != nil {
-		return nil, err
-	}
-	readme := "This workspace is empty on purpose. The question is about earlier conversation,\nnot about any file here.\n"
-	if err := os.WriteFile(filepath.Join(dir, "README.md"), []byte(readme), 0o644); err != nil {
-		return nil, err
-	}
-	previous, err := os.Getwd()
-	if err != nil {
-		return nil, err
-	}
-	if err := os.Chdir(dir); err != nil {
-		return nil, err
-	}
-	return func() {
-		_ = os.Chdir(previous)
-		_ = os.RemoveAll(dir)
-	}, nil
-}
-
-// runBoundaries runs only the pair each index task is built for: the narrowest
-// scale still addressing its cue, and the next one down. Each pair moves the
-// affordance and nothing else, where default-vs-off moves the whole index.
-
-// The two arms get separate nonces: reusing one would leave the first arm's
-// answer in a host-readable history for the second to find.
-func runBoundaries(root string) int {
-	root, err := filepath.Abs(root)
-	if err != nil {
-		fmt.Fprintln(os.Stderr, err)
-		return 2
-	}
-	restore, ierr := isolateWorkspace()
-	if ierr != nil {
-		fmt.Fprintln(os.Stderr, "isolate workspace:", ierr)
-		return 2
-	}
-	defer restore()
-	p, perr := realProvider()
-	if perr != nil {
-		fmt.Fprintln(os.Stderr, perr)
-		return 2
-	}
-
-	var all []contextMetrics
-	var cueSide, noCueSide []contextMetrics
-	for _, t := range indexTasks() {
-		cue, noCue := boundaryPair(t.CueTier)
-		for _, side := range []struct {
-			scale string
-			has   bool
-		}{{cue, true}, {noCue, false}} {
-			name := "index-" + side.scale
-			m, err := runOne(p, t, armFor(side.scale, false), name, root, side.has, false)
-			if err != nil {
-				fmt.Fprintf(os.Stderr, "%s [%s]: %v\n", t.ID, name, err)
-			}
-			all = append(all, m)
-			if side.has {
-				cueSide = append(cueSide, m)
-			} else {
-				noCueSide = append(noCueSide, m)
-			}
-			fmt.Printf("%-24s %-14s cue=%-3v %-14s search=%d read=%d direct=%-5v recall-tok=%-5d escape=%d\n",
-				t.ID, name, side.has, m.FailureStage, m.SearchCalls, m.ReadCalls,
-				m.CueDirectRead, m.RecallReturnedTokens, m.EscapeCalls)
-		}
-	}
-	reportDeltas(cueSide, noCueSide)
-	fmt.Print(reportStopping(all))
-	return writeResults(root, all)
-}
-
-// reportDeltas is the paired reading: each task against itself, then pooled.
-func reportDeltas(cueSide, noCueSide []contextMetrics) {
-	fmt.Printf("\n## Paired deltas (cue present − cue absent)\n")
-	fmt.Printf("  %-24s %-10s %-10s %-9s %-11s %s\n",
-		"task", "recovery", "directread", "searches", "recall-tok", "escapes")
-	for i := range cueSide {
-		if i >= len(noCueSide) {
-			break
-		}
-		a, b := cueSide[i], noCueSide[i]
-		fmt.Printf("  %-24s %-10s %-10s %-9d %-11d %d\n", a.Task,
-			deltaBool(a.AnswerRecovered, b.AnswerRecovered),
-			deltaBool(a.CueDirectRead, b.CueDirectRead),
-			a.SearchCalls-b.SearchCalls,
-			a.RecallReturnedTokens-b.RecallReturnedTokens,
-			a.EscapeCalls-b.EscapeCalls)
-	}
-	fmt.Println("\n## Pooled")
-	fmt.Println(" ", summarize("cue-present", cueSide).line())
-	fmt.Println(" ", summarize("cue-absent", noCueSide).line())
-	fmt.Println("\n## Where each side looked first")
-	fmt.Printf("  %-14s %s\n", "cue-present", countsLine(summarize("x", cueSide).Routes))
-	fmt.Printf("  %-14s %s\n", "cue-absent", countsLine(summarize("x", noCueSide).Routes))
-}
-
-func deltaBool(a, b bool) string {
-	switch {
-	case a == b:
-		return "same"
-	case a:
-		return "+cue"
-	default:
-		return "-cue"
-	}
-}
-
-// runExperiment runs one experiment's tasks across its arms. A dry run drives
-// the same pipeline with a scripted provider: it proves fixture, run, scoring
-// and report hold together before any of it is paid for.
-func runExperiment(experiment, root string, dry bool, tasks []contextTask) int {
-	root, err := filepath.Abs(root)
-	if err != nil {
-		fmt.Fprintln(os.Stderr, err)
-		return 2
-	}
-	restore, err := isolateWorkspace()
-	if err != nil {
-		fmt.Fprintln(os.Stderr, "isolate workspace:", err)
-		return 2
-	}
-	defer restore()
-	var p provider.Provider = &scriptedProvider{}
-	if !dry {
-		real, rerr := realProvider()
-		if rerr != nil {
-			fmt.Fprintln(os.Stderr, rerr)
-			return 2
-		}
-		p = real
-	}
-	byArm := map[string][]contextMetrics{}
-	var all []contextMetrics
-
-	if experiment == experimentIndex {
-		fmt.Print(fingerprint(dry))
-	}
-	for _, t := range tasks {
-		if t.Experiment != experiment {
-			continue
-		}
-		// Arms are shuffled inside each task block. Running six defaults, then
-		// six halves, would make provider drift collinear with the axis under
-		// test: a bad minute would land entirely on one arm.
-		for _, arm := range shuffledArms(t) {
-			cueVisible := t.Experiment == experimentIndex && cueExpectedVisible(t, arm.scale)
-			m, err := runOne(p, t, armFor(arm.scale, arm.searchOff), arm.name, root, cueVisible, dry)
-			if err != nil {
-				fmt.Fprintf(os.Stderr, "%s [%s]: %v\n", t.ID, arm.name, err)
-			}
-			byArm[arm.name] = append(byArm[arm.name], m)
-			all = append(all, m)
-			fmt.Printf("%-24s %-14s %-18s search=%d hit=%d read=%d recall-tok=%d %v\n",
-				t.ID, arm.name, m.FailureStage, m.SearchCalls, m.TargetSearchHits, m.ReadCalls,
-				m.RecallReturnedTokens, m.UnexpectedWorkTools)
-		}
-	}
-	reportFunnels(byArm)
-	if experiment == experimentIndex {
-		fmt.Print(reportDoseResponse(all))
-	}
-	fmt.Print(reportStopping(all))
-	fmt.Print(auditQueries(all).report())
-	fmt.Print(queryLines(all))
-	if experiment == experimentIndex {
-		reportBoundaries(all)
-	}
-	return writeResults(root, all)
-}
-
-func cueExpectedVisible(t contextTask, scale string) bool {
-	visible, _ := tierScales(t.CueTier)
-	return contains(visible, scale)
-}
-
-func reportFunnels(byArm map[string][]contextMetrics) {
-	names := make([]string, 0, len(byArm))
-	for name := range byArm {
-		names = append(names, name)
-	}
-	sort.Strings(names)
-	fmt.Println("\n## Funnel")
-	for _, name := range names {
-		fmt.Println(" ", summarize(name, byArm[name]).line())
-	}
-	fmt.Println("\n## Where it looked first")
-	for _, name := range names {
-		f := summarize(name, byArm[name])
-		fmt.Printf("  %-14s %s\n", name, countsLine(f.Routes))
-	}
-	fmt.Println("\n## Where it broke")
-	for _, name := range names {
-		f := summarize(name, byArm[name])
-		stages := make([]string, 0, len(f.Stages))
-		for stage, n := range f.Stages {
-			stages = append(stages, fmt.Sprintf("%s=%d", stage, n))
-		}
-		sort.Strings(stages)
-		fmt.Printf("  %-12s %s\n", name, strings.Join(stages, " "))
-	}
-}
-
-// writeTrajectories saves what each turn produced, after the whole batch is
-// done. Three analyses in a row have needed a re-run because something was
-// counted and not kept — query text, then snippet contents. It is written at
-// the end on purpose: during the batch these files are the answer.
-func writeTrajectories(root string, all []contextMetrics) error {
-	f, err := os.Create(filepath.Join(root, "trajectories.jsonl"))
-	if err != nil {
-		return err
-	}
-	defer f.Close()
-	for _, m := range all {
-		if err := writeJSONLine(f, map[string]any{
-			"task": m.Task, "arm": m.Arm, "messages": m.trajectory,
-		}); err != nil {
-			return err
-		}
-	}
-	return nil
-}
-
-func writeResults(root string, all []contextMetrics) int {
-	path := filepath.Join(root, "results.jsonl")
-	f, err := os.Create(path)
-	if err != nil {
-		fmt.Fprintln(os.Stderr, err)
-		return 2
-	}
-	defer f.Close()
-	for _, m := range all {
-		if err := writeJSONLine(f, m); err != nil {
-			fmt.Fprintln(os.Stderr, err)
-			return 2
-		}
-	}
-	if err := writeTrajectories(root, all); err != nil {
-		fmt.Fprintln(os.Stderr, "write trajectories:", err)
-		return 2
-	}
-	fmt.Printf("\n%d runs written to %s (with trajectories.jsonl)\n", len(all), path)
-	return 0
-}
-
-// reportBoundaries prints each index task across its own cue boundary, then the
-// two sides pooled. Task difficulty is held constant inside a row, which is
-// what a four-arm average cannot do.
-func reportBoundaries(all []contextMetrics) {
-	byKey := map[string]contextMetrics{}
-	for _, m := range all {
-		byKey[m.Task+"|"+m.Arm] = m
-	}
-	var cueSide, noCueSide []contextMetrics
-	fmt.Println("\n## Across each task's cue boundary")
-	fmt.Printf("  %-24s %-16s %-4s %-10s %-9s %-12s %-11s %s\n",
-		"task", "arm", "cue", "recovered", "cue-read", "search/task", "recall-tok", "escape")
-	for _, t := range indexTasks() {
-		cue, noCue := boundaryPair(t.CueTier)
-		for _, side := range []struct {
-			scale string
-			has   bool
-		}{{cue, true}, {noCue, false}} {
-			m, ok := byKey[t.ID+"|index-"+side.scale]
-			if !ok {
-				continue
-			}
-			if side.has {
-				cueSide = append(cueSide, m)
-			} else {
-				noCueSide = append(noCueSide, m)
-			}
-			mark := "no"
-			if side.has {
-				mark = "yes"
-			}
-			fmt.Printf("  %-24s %-16s %-4s %-10v %-9v %-12d %-11d %d\n",
-				t.ID, "index-"+side.scale, mark, m.AnswerRecovered, m.RecallReadWithoutSearch,
-				m.SearchCalls, m.RecallReturnedTokens, m.EscapeCalls)
-		}
-	}
-	fmt.Println("\n## Pooled across boundaries")
-	fmt.Println(" ", summarize("cue-present", cueSide).line())
-	fmt.Println(" ", summarize("cue-absent", noCueSide).line())
-}
-
-// shuffledArms permutes a task's arms deterministically, so the order is
-// reproducible and still uncorrelated with the axis being measured.
-func shuffledArms(t contextTask) []experimentArm {
-	arms := armsFor(t)
-	rng := seededRand(t.ID, "arm-order")
-	rng.Shuffle(len(arms), func(i, j int) { arms[i], arms[j] = arms[j], arms[i] })
-	return arms
-}
-
-// fingerprint records what produced a batch. Not a gate — but batch effects
-// have reversed three conclusions in this experiment already, and a table with
-// no provenance cannot be compared with the next one.
-func fingerprint(dry bool) string {
-	model, endpoint := runModel, runBaseURL
-	if dry {
-		model, endpoint = "scripted", "none"
-	}
-	commit := "unknown"
-	if out, err := exec.Command("git", "rev-parse", "--short", "HEAD").Output(); err == nil {
-		commit = strings.TrimSpace(string(out))
-	}
-	return fmt.Sprintf("## Batch\n  model=%s endpoint=%s window=%d generations=%d commit=%s started=%s\n",
-		model, endpoint, fixtureWindow, fixtureGenerations, commit, time.Now().UTC().Format(time.RFC3339))
-}
-
-// timeoutReason separates a run that never found its answer from one that found
-// it and kept going. Only the first is a case for a longer budget; the second
-// would just get more time to circle. A batch reporting only "context deadline
-// exceeded" cannot tell an arm that made retrieval harder from an arm that
-// happened to draw a slow minute.
-func timeoutReason(m contextMetrics) string {
-	switch {
-	case m.TargetHeldAtEnd && m.RoundsAfterTarget >= 2:
-		return "PostTargetRunaway"
-	case m.TargetHeldAtEnd:
-		return "TargetHeldLate"
-	case m.SearchCalls >= 3:
-		return "SearchSpiral"
-	case m.SearchCalls+m.ReadCalls+m.EscapeCalls == 0:
-		return "NoToolActivity"
-	default:
-		return "Unknown"
-	}
-}
@@ -1,95 +0,0 @@
-package main
-
-import (
-	"context"
-	"encoding/json"
-	"regexp"
-	"slices"
-	"strconv"
-	"strings"
-
-	"reasonix/internal/contract/provider"
-)
-
-// scriptedProvider walks the chain a competent model would: search, read the
-// best hit, answer with what came back. It proves the pipeline holds together
-// for nothing, so a real run's first failure is about the model.
-
-type scriptedProvider struct{}
-
-func (p *scriptedProvider) Name() string { return "contextbench-scripted" }
-
-var (
-	scriptedHit    = regexp.MustCompile(`(?m)^#(\d+) `)
-	scriptedRecall = regexp.MustCompile(`(?s)^#\d+\n(.*)`)
-)
-
-// Stream decides from the request alone, never from a call counter: one
-// provider serves every task in a run, and a counter would make the second task
-// start where the first left off.
-func (p *scriptedProvider) Stream(_ context.Context, req provider.Request) (<-chan provider.Chunk, error) {
-	ch := make(chan provider.Chunk, 3)
-	defer close(ch)
-	// Honour the schema, as a model does: an arm without a query parameter is
-	// an arm that cannot search, and pretending otherwise scores a refusal as
-	// a failed search.
-	if !recallAccepts(req, "query") {
-		ch <- provider.Chunk{Type: provider.ChunkText, Text: "I have no address for that part of the history."}
-		ch <- provider.Chunk{Type: provider.ChunkDone}
-		return ch, nil
-	}
-	// The last message, not the last tool result anywhere: the fixture's own
-	// history is full of tool results, and one of those is not this turn's.
-	last := ""
-	if n := len(req.Messages); n > 0 && req.Messages[n-1].Role == provider.RoleTool {
-		last = req.Messages[n-1].Content
-	}
-	switch {
-	case last == "":
-		args, _ := json.Marshal(map[string]string{"query": lastUserText(req)})
-		ch <- provider.Chunk{Type: provider.ChunkToolCall, ToolCall: &provider.ToolCall{
-			ID: "s1", Name: "recall", Arguments: string(args),
-		}}
-	case strings.Contains(last, "Folded context matching"):
-		positions := scriptedHit.FindAllStringSubmatch(last, 1)
-		if len(positions) == 0 {
-			ch <- provider.Chunk{Type: provider.ChunkText, Text: "Nothing in the folded history matches."}
-			break
-		}
-		n, _ := strconv.Atoi(positions[0][1])
-		args, _ := json.Marshal(map[string][]int{"positions": {n}})
-		ch <- provider.Chunk{Type: provider.ChunkToolCall, ToolCall: &provider.ToolCall{
-			ID: "r1", Name: "recall", Arguments: string(args),
-		}}
-	default:
-		// Answer with what recall returned, so a correct chain scores as one.
-		body := last
-		if m := scriptedRecall.FindStringSubmatch(last); len(m) > 1 {
-			body = m[1]
-		}
-		ch <- provider.Chunk{Type: provider.ChunkText, Text: "From the folded history: " + body}
-	}
-	ch <- provider.Chunk{Type: provider.ChunkDone}
-	return ch, nil
-}
-
-// recallAccepts reports whether recall is offered with the given parameter.
-func recallAccepts(req provider.Request, param string) bool {
-	for _, t := range req.Tools {
-		if t.Name != "recall" {
-			continue
-		}
-		b, err := json.Marshal(t.Parameters)
-		return err == nil && strings.Contains(string(b), `"`+param+`"`)
-	}
-	return false
-}
-
-func lastUserText(req provider.Request) string {
-	for _, m := range slices.Backward(req.Messages) {
-		if m.Role == provider.RoleUser {
-			return m.Content
-		}
-	}
-	return ""
-}
@@ -1,182 +0,0 @@
-// snippet.go — does a rank-1 hit actually hand over the answer?
-package main
-
-import (
-	"context"
-	"fmt"
-	"os"
-	"path/filepath"
-	"strings"
-
-	"reasonix/internal/base/retrieval"
-	"reasonix/internal/contract/ablation"
-	"reasonix/internal/contract/event"
-	"reasonix/internal/contract/tool"
-	"reasonix/internal/runtime/agent"
-)
-
-// TargetHit is not EvidenceSufficient: a rank-1 hit can still hand over a
-// window too short to answer with, and searching again after that is correct.
-// Measured through the shipped recall path, not a copy of it.
-
-// snippetWidths are the windows to compare. 240 is what ships.
-var snippetWidths = []int{120, 180, 240, 320, 480}
-
-// historicalQueries are the queries these tasks actually drew in the pilots,
-// kept because a probe query is the host's phrasing and the model's was not.
-var historicalQueries = map[string][]string{
-	"i03-coalescing":     {"coalesce window marker probe", "job completion coalescing window"},
-	"i04-recovery-fence": {"recovery fence marker epoch", "expected epoch", "recfence quartz"},
-}
-
-type snippetReading struct {
-	Task     string
-	Query    string
-	Width    int
-	Covered  int
-	Total    int
-	Runes    int
-	Rank     int
-	Complete bool
-}
-
-// runSnippetAudit answers the question offline, over many instantiations so a
-// single seed's lengths cannot decide it.
-func runSnippetAudit(root string, seeds int) int {
-	fmt.Printf("## Snippet coverage (%d instantiations per task)\n", seeds)
-	fmt.Println("  Does the window a rank-1 hit returns contain every value the task scores on?")
-
-	byTaskWidth := map[string]map[int][]snippetReading{}
-	for _, t := range allTasks() {
-		byTaskWidth[t.ID] = map[int][]snippetReading{}
-		for seed := range seeds {
-			inst, err := instantiateTask(t, seededRand(t.ID, "snippet", fmt.Sprint(seed)))
-			if err != nil {
-				fmt.Fprintln(os.Stderr, err)
-				return 2
-			}
-			queries := append([]string{inst.ProbeQuery}, historicalQueries[t.ID]...)
-			for _, q := range queries {
-				for _, width := range snippetWidths {
-					r := measureSnippet(inst, q, width)
-					byTaskWidth[t.ID][width] = append(byTaskWidth[t.ID][width], r)
-				}
-			}
-		}
-	}
-
-	fmt.Printf("\n  %-24s %-5s", "task", "vals")
-	for _, w := range snippetWidths {
-		fmt.Printf("%9d", w)
-	}
-	fmt.Println("   mean runes @240")
-	for _, t := range allTasks() {
-		vals := len(t.AnswerVars)
-		fmt.Printf("  %-24s %-5d", t.ID, vals)
-		for _, w := range snippetWidths {
-			fmt.Printf("%8.0f%%", 100*completeRate(byTaskWidth[t.ID][w]))
-		}
-		fmt.Printf("   %.0f\n", meanRunes(byTaskWidth[t.ID][240]))
-	}
-	reportShippedPath(root)
-	return 0
-}
-
-// measureSnippet renders one hit through the same MakeSnippet the recall path
-// uses, then asks whether every scored value survived the window.
-func measureSnippet(inst fixtureInstance, query string, width int) snippetReading {
-	terms, err := retrieval.QueryTerms(query)
-	if err != nil {
-		return snippetReading{Task: inst.Task.ID, Query: query, Width: width, Total: len(inst.AnswerMarkers)}
-	}
-	// The tool result is the document the answer lives in; a call's arguments
-	// are a separate document and never carry it.
-	snippet := retrieval.MakeSnippet(inst.body, query, terms, width)
-	covered := 0
-	for _, marker := range inst.AnswerMarkers {
-		if strings.Contains(strings.ToLower(snippet), strings.ToLower(marker)) {
-			covered++
-		}
-	}
-	return snippetReading{
-		Task: inst.Task.ID, Query: query, Width: width,
-		Covered: covered, Total: len(inst.AnswerMarkers),
-		Runes: len([]rune(snippet)), Complete: covered == len(inst.AnswerMarkers),
-	}
-}
-
-// reportShippedPath cross-checks the table against the real recall tool: a
-// fixture, a fold, and a search through the agent that ships.
-func reportShippedPath(root string) {
-	fmt.Println("\n## Through the shipped recall path (width as shipped)")
-	for _, t := range allTasks() {
-		inst, err := instantiateTask(t, seededRand(t.ID, "snippet-live"))
-		if err != nil {
-			continue
-		}
-		f, err := buildFixture(inst, ablation.Set{}, filepath.Join(root, "snippet", t.ID, "session.jsonl"))
-		if err != nil {
-			fmt.Printf("  %-24s fixture: %v\n", t.ID, err)
-			continue
-		}
-		a := agent.New(&digestProvider{}, tool.NewRegistry(), f.Session, agent.Options{
-			ContextWindow: fixtureWindow, CompactRatio: 0.5, RecentKeep: 2,
-			SessionPath: f.Path, KeepPolicy: agent.KeepErrors,
-		}, event.Discard)
-		a.LoadProjectionSidecar(f.Path)
-		queries := append([]string{inst.ProbeQuery}, historicalQueries[t.ID]...)
-		for _, q := range queries {
-			res, err := a.RecallContext(context.Background(), tool.RecallRequest{Query: q})
-			if err != nil {
-				fmt.Printf("  %-24s %-34q recall: %v\n", t.ID, q, err)
-				continue
-			}
-			covered := 0
-			for _, marker := range inst.AnswerMarkers {
-				if strings.Contains(strings.ToLower(res.Text), strings.ToLower(marker)) {
-					covered++
-				}
-			}
-			fmt.Printf("  %-24s %-34q %d/%d values in the result\n",
-				t.ID, truncate(q, 32), covered, len(inst.AnswerMarkers))
-		}
-	}
-}
-
-func completeRate(rs []snippetReading) float64 {
-	if len(rs) == 0 {
-		return 0
-	}
-	n := 0
-	for _, r := range rs {
-		if r.Complete {
-			n++
-		}
-	}
-	return float64(n) / float64(len(rs))
-}
-
-func meanRunes(rs []snippetReading) float64 {
-	if len(rs) == 0 {
-		return 0
-	}
-	total := 0
-	for _, r := range rs {
-		total += r.Runes
-	}
-	return float64(total) / float64(len(rs))
-}
-
-// answerCoverage is what a run should have recorded all along: how much of the
-// scored answer a search actually handed over. rank=1 with coverage 1/3 is not
-// a stopping failure.
-func answerCoverage(text string, markers []string) int {
-	covered := 0
-	lower := strings.ToLower(text)
-	for _, marker := range markers {
-		if strings.Contains(lower, strings.ToLower(marker)) {
-			covered++
-		}
-	}
-	return covered
-}
@@ -1,3 +1,4 @@
 prompt = "contacts.txt is full of duplicate entries and users keep complaining. Clean it up and write the result to contacts_clean.txt."
 class = "ambiguous"
+max_steps = 15
 timeout_sec = 300
@@ -1,3 +1,4 @@
 prompt = "config.py silently accepts broken config files and the app then crashes much later somewhere else. Make loading fail fast with clear errors instead."
 class = "ambiguous"
+max_steps = 15
 timeout_sec = 300
@@ -1,3 +1,4 @@
 prompt = "parse_all in parser.py only handles some of the files under fixtures/. Users need it to handle all of them. Extend it so every fixture is parsed."
 class = "ambiguous"
+max_steps = 15
 timeout_sec = 300
@@ -1,3 +1,4 @@
 prompt = "Users report that stats.py keeps crashing on real-world input files. Make it robust enough for production use. Sample inputs are in samples/."
 class = "ambiguous"
+max_steps = 15
 timeout_sec = 300
@@ -1,3 +1,4 @@
 prompt = "Exactly one setting differs across config/base.toml, config/staging.toml and config/production.toml. Write AUDIT.md naming the setting that drifted, the environment it drifted in, and both values. Do not change any config."
 class = "read-only-output"
+max_steps = 25
 timeout_sec = 480
@@ -1,3 +1,4 @@
 prompt = "discounted(price_cents, percent) in pricing.py never applies the discount. Fix it (integer cents, round the discount down)."
 class = "atomic-bugfix"
+max_steps = 12
 timeout_sec = 180
@@ -1,3 +1,4 @@
 prompt = "overlaps(a, b) in intervals.py misjudges some interval pairs. Fix it so half-open intervals [start, end) are compared correctly."
 class = "atomic-bugfix"
+max_steps = 12
 timeout_sec = 180
@@ -1,6 +0,0 @@
-def median(values):
-    ordered = sorted(values)
-    mid = len(ordered) // 2
-    if len(ordered) % 2:
-        return ordered[mid]
-    return (ordered[mid - 1] + ordered[mid]) / 2
@@ -1,3 +1,4 @@
 prompt = "median(values) in stats.py is wrong for even-length lists. Fix it to return the average of the two middle values."
 class = "atomic-bugfix"
+max_steps = 12
 timeout_sec = 180
@@ -1,3 +1,4 @@
 prompt = "Fifo.pop in fifo.py returns the newest item instead of the oldest. Fix the FIFO order."
 class = "atomic-bugfix"
+max_steps = 12
 timeout_sec = 180
@@ -1,3 +1,4 @@
 prompt = "slugify in slug.py leaves stray hyphens at the ends of some slugs. Fix it."
 class = "atomic-bugfix"
+max_steps = 12
 timeout_sec = 180
@@ -1,3 +1,4 @@
 prompt = "store.py: values saved with set() sometimes cannot be read back with get(). Fix the bug."
 class = "atomic-bugfix"
+max_steps = 12
 timeout_sec = 180
@@ -1,3 +1,4 @@
 prompt = "tail(lst, n) in utils.py drops the last element. Fix it so it returns the last n items of the list."
 class = "atomic-bugfix"
+max_steps = 12
 timeout_sec = 180
@@ -1,3 +1,4 @@
 prompt = "Begin by reading the file story/chapter-1.md. Each chapter ends by telling you which chapter to read next; follow that trail one chapter at a time — read a chapter, find where it points you, then read that next chapter — until a chapter says the trail ends. You cannot know the next chapter without reading the current one, so do not guess or read ahead. In chapter 1 a river village is named; in the final chapter on the trail a noble House is named. When the trail ends, write a file answer.txt whose entire contents are the village name from chapter 1, a single hyphen, and the House name from the final chapter — for example Riverton-Stark — and nothing else."
 class = "repo-exploration"
+max_steps = 40
 timeout_sec = 900