docs(readme): clarify /loop runs a prompt or custom command, not built-ins

- stopLoop re-arms the poll ticker after removing one loop, so stopping one of
  several no longer strands the rest with no tick scheduled (jatmn).
- a bare self-paced /command is rejected with a pointer to the interval form:
  self-paced drives a free-form goal via the playbook + LOOP: control line, which
  a custom command's expanded template never carries, so it would silently
  degrade to adaptive polling (jatmn).
- /resume only tears loops down on an actual session-id change; /resume latest or
  /resume <currentID> can resolve to the already-active session, whose loops are
  its own, not a "previous" session's (CodeRabbit).
- tests: skipped-behind-modal loop stays scheduled, footer carries the next-wake
  time, the confirmed /clear actually wipes the transcript, plus multi-loop stop
  keeping the ticker alive and self-paced-/command rejection.

.github/workflows/ci.yml

@@ -100,9 +100,10 @@ jobs:
           go-version-file: go.mod
           cache: true
 
-      # govulncheck and deadcode resolve a toolchain from their own modules, which
-      # can drop below the version go.mod requires and then fail to load our
-      # packages. Pin GOTOOLCHAIN to the go.mod toolchain so both run under it.
+      # govulncheck, deadcode, and golangci-lint each resolve a toolchain from
+      # their own modules, which can drop below the version go.mod requires and
+      # then fail to load our packages. Pin GOTOOLCHAIN to the go.mod toolchain
+      # so all three run under it.
       - name: Pin toolchain from go.mod
         run: |
           toolchain="$(awk '/^toolchain /{print $2}' go.mod)"
@@ -120,3 +121,13 @@ jobs:
       - name: deadcode (advisory)
         continue-on-error: true
         run: go run golang.org/x/tools/cmd/[email protected] -test=false ./...
+
+      # Advisory: catches what deadcode's whole-program reachability analysis
+      # doesn't, unused private functions/assignments reachable within a
+      # package but never actually called, plus staticcheck-style correctness
+      # and readability issues. Scoped to a few linters rather than the full
+      # default battery, and non-blocking, while the existing findings across
+      # the repo are cleaned up incrementally (see #527).
+      - name: golangci-lint (advisory)
+        continue-on-error: true
+        run: go run github.com/golangci/golangci-lint/v2/cmd/[email protected] run --enable-only unused,ineffassign,staticcheck ./...

.github/workflows/release-artifacts.yml

@@ -54,8 +54,10 @@ jobs:
       # is published under the pushed git tag. If they disagree, install.sh/install.ps1
       # and the npm postinstall all build a download URL that 404s. Fail fast so a
       # forgotten version bump can never publish mis-versioned, un-installable assets.
+      # Matches on the ref (not event_name) because release-please dispatches this
+      # workflow at the tag ref — its GITHUB_TOKEN-created tag cannot fire `push`.
       - name: Check tag matches package.json version
-        if: ${{ github.event_name == 'push' && startsWith(github.ref, 'refs/tags/v') }}
+        if: ${{ startsWith(github.ref, 'refs/tags/v') }}
         shell: bash
         run: |
           pkg_version="$(node -p "require('./package.json').version")"
@@ -78,13 +80,15 @@ jobs:
           path: dist/release/*
           if-no-files-found: error
 
-      # On a `v*` tag push, publish the packaged archive + .sha256 as GitHub Release
-      # assets so the documented `scripts/install.sh` / `install.ps1` path (which reads
-      # /releases/latest) works. Uses the preinstalled `gh` (no new action dependency);
-      # each matrix OS creates the release if absent (best-effort) then uploads its own
-      # assets with --clobber. NOTE: maintainers must push the tag — this never tags.
+      # On a `v*` tag ref (pushed manually, or dispatched at the tag by the
+      # release-please workflow), publish the packaged archive + .sha256 as GitHub
+      # Release assets so the documented `scripts/install.sh` / `install.ps1` path
+      # (which reads /releases/latest) works. Uses the preinstalled `gh` (no new
+      # action dependency); each matrix OS creates the release if absent
+      # (best-effort — release-please normally creates it first) then uploads its
+      # own assets with --clobber. Tags come from merging the release-please PR.
       - name: Publish to GitHub Release
-        if: ${{ github.event_name == 'push' && startsWith(github.ref, 'refs/tags/v') }}
+        if: ${{ startsWith(github.ref, 'refs/tags/v') }}
         shell: bash
         env:
           GH_TOKEN: ${{ github.token }}
@@ -92,3 +96,89 @@ jobs:
           tag="${GITHUB_REF_NAME}"
           gh release create "$tag" --repo "$GITHUB_REPOSITORY" --title "$tag" --generate-notes 2>/dev/null || true
           gh release upload "$tag" dist/release/* --repo "$GITHUB_REPOSITORY" --clobber
+
+  # Publishes the npm wrapper (@gitlawb/zero) once every platform's release
+  # assets are live — the package's postinstall downloads its binary from the
+  # GitHub Release, so publishing before the assets exist would ship a package
+  # that fails to install.
+  #
+  # Auth is npm OIDC trusted publishing: no NPM_TOKEN secret. The package's
+  # npmjs.com settings must list this repo + workflow file as a trusted
+  # publisher, and provenance is then generated automatically. Note npm only
+  # allows configuring a trusted publisher on an EXISTING package, so the very
+  # first @gitlawb/zero publish is a one-time manual `npm publish` bootstrap.
+  publish-npm:
+    name: Publish npm package
+    needs: package
+    if: ${{ startsWith(github.ref, 'refs/tags/v') }}
+    runs-on: ubuntu-latest
+    # Manual approval gate: this environment carries a required-reviewer rule
+    # (Settings -> Environments -> npm-publish), so the job pauses until a
+    # maintainer approves it in the Actions UI before anything reaches npm.
+    environment:
+      name: npm-publish
+      url: https://www.npmjs.com/package/@gitlawb/zero
+    permissions:
+      contents: read
+      id-token: write # OIDC token exchange for npm trusted publishing + provenance
+    steps:
+      - name: Checkout
+        uses: actions/checkout@34e114876b0b11c390a56381ad16ebd13914f8d5 # v4
+
+      # No registry-url here on purpose: setup-node's registry-url writes an
+      # .npmrc line referencing $NODE_AUTH_TOKEN, and npm hard-fails on the
+      # unset env var. Trusted publishing needs no token at all.
+      - name: Setup Node
+        uses: actions/setup-node@49933ea5288caeca8642d1e84afbd3f7d6820020 # v4
+        with:
+          node-version: 24
+
+      # OIDC trusted publishing needs npm >= 11.5.1. Use the npm bundled with
+      # the pinned Node toolchain (Node 24 ships >= 11.5.1) rather than pulling
+      # npm@latest from the registry inside the publish path — an unpinned
+      # install there would be avoidable supply-chain exposure. Fail fast if a
+      # future runner/toolchain change ever regresses the bundled npm.
+      - name: Check npm supports trusted publishing
+        run: |
+          npm --version | node -e '
+            const version = require("fs").readFileSync(0, "utf8").trim();
+            const [major, minor, patch] = version.split(".").map(Number);
+            const ok =
+              major > 11 ||
+              (major === 11 && (minor > 5 || (minor === 5 && patch >= 1)));
+            if (!ok) {
+              console.error(
+                `::error::bundled npm ${version} is too old for OIDC trusted ` +
+                  `publishing (needs >= 11.5.1) — bump node-version in this job.`,
+              );
+              process.exit(1);
+            }
+            console.log(`npm ${version} supports trusted publishing.`);
+          '
+
+      # Probe the exact URLs postinstall will build (internal/release/release.go
+      # asset-name scheme) as an anonymous client. This also blocks publishing
+      # while the repo is still private — private release assets 404 without
+      # auth, exactly as they would for an installing user.
+      - name: Verify release assets are downloadable
+        shell: bash
+        run: |
+          version="$(node -p "require('./package.json').version")"
+          base="https://github.com/${GITHUB_REPOSITORY}/releases/download/v${version}"
+          for asset in \
+            "zero-v${version}-linux-x64.tar.gz" \
+            "zero-v${version}-linux-arm64.tar.gz" \
+            "zero-v${version}-macos-x64.tar.gz" \
+            "zero-v${version}-macos-arm64.tar.gz" \
+            "zero-v${version}-windows-x64.zip"; do
+            for url in "${base}/${asset}" "${base}/${asset}.sha256"; do
+              if ! curl -fsSLI -o /dev/null "$url"; then
+                echo "::error::${url} is not publicly downloadable — npm postinstall would fail. Is the repo public and are all assets uploaded?" >&2
+                exit 1
+              fi
+            done
+          done
+          echo "all release assets for v${version} are publicly downloadable."
+
+      - name: Publish to npm
+        run: npm publish --access public

.github/workflows/release-please.yml

@@ -0,0 +1,35 @@
+name: Release Please
+
+on:
+  push:
+    branches:
+      - main
+
+permissions:
+  contents: write # push the release-PR branch, create the tag + GitHub Release
+  pull-requests: write # open/update the release PR
+  issues: write # create/apply release-tracking labels on the release PR
+  actions: write # dispatch release-artifacts.yml once a release is cut
+
+jobs:
+  release-please:
+    name: Release PR / tag
+    runs-on: ubuntu-latest
+    steps:
+      - name: Run release-please
+        id: release
+        uses: googleapis/release-please-action@a02a34c4d625f9be7cb89156071d8567266a2445 # v4.2.0
+
+      # Tags created with the default GITHUB_TOKEN never fire `on: push: tags`
+      # workflows (GitHub suppresses workflows triggered by workflow-created
+      # events), so release-artifacts.yml would not run for a release-please
+      # tag. workflow_dispatch invoked through the API is exempt from that
+      # suppression — dispatch the artifacts build at the new tag explicitly.
+      - name: Build and publish release artifacts
+        if: ${{ steps.release.outputs.release_created }}
+        env:
+          GH_TOKEN: ${{ github.token }}
+        run: |
+          gh workflow run release-artifacts.yml \
+            --repo "$GITHUB_REPOSITORY" \
+            --ref "${{ steps.release.outputs.tag_name }}"

.github/workflows/zero-action-smoke.yml

@@ -4,12 +4,14 @@ on:
   pull_request:
     paths:
       - action.yml
+      - docs/GITHUB_ACTION.md
       - .github/workflows/zero-action-smoke.yml
   push:
     branches:
       - main
     paths:
       - action.yml
+      - docs/GITHUB_ACTION.md
       - .github/workflows/zero-action-smoke.yml
 
 jobs:
@@ -29,7 +31,8 @@ jobs:
           import sys, yaml
 
           with open("action.yml", "r", encoding="utf-8") as handle:
-              doc = yaml.safe_load(handle)
+              action_raw = handle.read()
+          doc = yaml.safe_load(action_raw)
 
           assert isinstance(doc, dict), "action.yml must be a mapping"
           assert doc.get("name"), "action must declare a name"
@@ -65,6 +68,19 @@ jobs:
           assert "--skip-permissions-unsafe" not in serialized, \
               "action must never pass --skip-permissions-unsafe"
 
+          run_zero = next((step for step in steps if step.get("name") == "Run ZERO"), None)
+          assert run_zero is not None, "action.yml must contain a Run ZERO step"
+          run_zero_script = run_zero.get("run") or ""
+          assert 'export "${ZERO_INPUT_API_KEY_ENV}=${ZERO_INPUT_API_KEY}"' in run_zero_script, \
+              "Run ZERO must export the provider key for the current step"
+          assert "later steps inherit" not in action_raw.lower(), \
+              "provider keys must not be documented as inherited by later steps"
+          for line in run_zero_script.splitlines():
+              if "GITHUB_ENV" in line and (
+                  "ZERO_INPUT_API_KEY" in line or "ZERO_INPUT_API_KEY_ENV" in line
+              ):
+                  raise AssertionError("provider key must not be persisted to GITHUB_ENV")
+
           print(f"action.yml OK: {len(inputs)} inputs, {len(outputs)} outputs, {len(steps)} steps")
           PY
 

.gitignore

@@ -37,8 +37,13 @@ report.[0-9]_.[0-9]_.[0-9]_.[0-9]_.json
 
 /zero
 /zero.exe
+/zero-skills
 /snake-game.html
 
 # Superpowers skill artifacts (brainstorm mockups, specs, plans) — local only
 .superpowers/
 docs/superpowers/
+
+# Termux/Android build artifacts
+zero-termux-arm64
+zero-linux-sandbox

.release-please-manifest.json

@@ -0,0 +1 @@
+{".":"0.1.0"}

AGENTS.md

@@ -26,7 +26,7 @@ Accepted file names, in priority order at each level:
 | `./ZERO.md` | Brand-specific alias. Same format, lower priority. |
 | `./.zero/AGENTS.md` | Project-local, hidden, gitignored. Personal notes that stay out of git. |
 
-Matching is **case-insensitive** on the basename, so `AGENTS.md`, `Agents.md`, and `agents.md` resolve to the same file on Windows and macOS. The git-tracked filename in this repo is `AGENTS.MD` (uppercase `MD`) — keep that on case-sensitive filesystems (Linux, the WSL filesystem, or a CI runner) to match what the loader looks for.
+Matching is **case-insensitive** on the basename, so `AGENTS.md`, `Agents.md`, and `agents.md` resolve to the same file on Windows and macOS. The git-tracked filename in this repo is `AGENTS.md` — keep that on case-sensitive filesystems (Linux, the WSL filesystem, or a CI runner) to match what the loader looks for.
 
 Both files use the same format. YAML frontmatter is optional; the markdown body is loaded as instructions for the agent. Zero reads the file once at session start, so changes take effect on the next `zero` launch — not mid-session.
 
@@ -47,6 +47,12 @@ Tips:
 - In a monorepo, drop a narrower `AGENTS.md` in each sub-tree (e.g. `services/api/AGENTS.md`). Zero picks those up automatically when you launch from inside the sub-tree.
 - A YAML frontmatter block (`---\n...\n---`) at the top is preserved verbatim in the injected prompt but is not parsed for `globs:` or `alwaysApply:` scoping today — keep the body self-contained.
 
+### Personal guidelines, across every project
+
+For preferences that follow *you*, not a specific repo (tone, tooling habits, workflow), drop a `ZERO.md` in your user config directory: `~/.config/zero/ZERO.md` on Linux/macOS, `%AppData%\Roaming\zero\ZERO.md` on Windows — the same directory as `config.json` and your personal specialists. Same format and 8 KiB cap as the project files above, and the same case-insensitive basename match.
+
+This file is injected as its own `## User guidelines` section, before the project's `AGENTS.md`/`ZERO.md`, and is labeled as personal preference in the prompt: project guidelines are the later, more specific instruction and take precedence over it when the two conflict.
+
 ## 2. Custom specialists
 
 Specialists are Zero's sub-agents. Three scopes, in priority order:
@@ -307,6 +313,7 @@ A team that wants every contributor's Zero to behave the same way commits:
 Each contributor adds only:
 
 - `~/.config/zero/config.json` — their personal API keys, theme, default mode.
+- `~/.config/zero/ZERO.md` — personal preferences that follow them across every project (see section 1).
 - `~/.local/share/zero/skills/` — personal skills they keep across projects.
 
 That's it. Run `zero` from the repo root and the agent has the team's full instruction set, every contributor's personal setup, and nothing else.

CHANGELOG.md

@@ -6,12 +6,32 @@ The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.1.0/),
 aims to follow [Semantic Versioning](https://semver.org/spec/v2.0.0.html) once the first release is
 tagged. Until then, source builds report the version `dev`.
 
+## 0.1.0 (2026-07-02)
+
+
+### Features
+
+* publish zero to npm via release-please ([#367](https://github.com/Gitlawb/zero/issues/367)) ([8eccc26](https://github.com/Gitlawb/zero/commit/8eccc2669887bc38d35bc16a315c888e4d9ec43a))
+* **tui:** FILES sidebar panel with click-to-select and file drill-in ([#365](https://github.com/Gitlawb/zero/issues/365)) ([142c548](https://github.com/Gitlawb/zero/commit/142c548c89a8652ce300e64ddf1228ee36df7606))
+
+
+### Bug Fixes
+
+* **auth:** propagate credentials to every provider-build surface and pin children to the live provider ([#366](https://github.com/Gitlawb/zero/issues/366)) ([6e0a665](https://github.com/Gitlawb/zero/commit/6e0a665118fe0e09c4b07d482dd18f86045acd2b))
+
 ## [Unreleased]
 
 ### Added
 - `SECURITY.md` with a private vulnerability-reporting path, `CODE_OF_CONDUCT.md`, this changelog, and
   GitHub issue/PR templates.
-- `--theme {auto|dark|light}` flag for the TUI (previously only the `ZERO_THEME` env var existed).
+- Interactive `/theme` picker: bare `/theme` opens a popup that live-previews each palette as you move
+  and applies on select (Esc reverts).
+- Ten built-in color themes alongside the `dark`/`light` built-ins — `dracula`, `nord`, `gruvbox`,
+  `tokyo-night`, `catppuccin`, `one-dark`, `solarized-dark`, `rose-pine`, `everforest`, and
+  `solarized-light` — selectable via `/theme <name>`, `--theme <name>`, or `ZERO_THEME`. Every palette
+  is contrast-audited to WCAG AA. The built-in light theme was reworked for legibility.
+- `--theme <name>` flag for the TUI, accepting `auto` or any registered theme (previously only the
+  `ZERO_THEME` env var existed).
 - "Accessibility / Appearance" section in the README documenting `NO_COLOR`, `ZERO_THEME`, `/theme`,
   and `ZERO_NO_FADE`.
 

Makefile

@@ -1,4 +1,4 @@
-# Zero build/test/lint targets. AGENTS.MD says "Build with `make`" and "Run `make
+# Zero build/test/lint targets. AGENTS.md says "Build with `make`" and "Run `make
 # lint` before opening a PR" — these targets back those instructions.
 .DEFAULT_GOAL := build
 .PHONY: build build-all test test-race vet fmt fmt-check lint tidy clean help

README.md

@@ -8,6 +8,8 @@
   <a href="LICENSE"><img alt="license" src="https://img.shields.io/badge/license-MIT-blue"></a>
   <img alt="Go 1.25+" src="https://img.shields.io/badge/Go-1.25+-00ADD8?logo=go&logoColor=white">
   <img alt="25+ providers" src="https://img.shields.io/badge/providers-25+-34E2EA">
+  <br>
+  <strong>English</strong> | <a href="README_ZH.md">中文</a>
 </p>
 
 Zero is an AI coding agent for your local terminal. It can inspect a repository,
@@ -51,6 +53,31 @@ The npm package installs a small wrapper plus the matching Zero binary for your
 platform from GitHub Releases. It supports Linux, macOS, and Windows on x64 and
 arm64.
 
+### Bun
+
+Bun does not run dependency lifecycle scripts by default, so the `postinstall`
+that fetches the Zero binary is skipped and the first run fails with
+`No native binary found next to the npm wrapper`.
+
+The simplest fix is to trust the package after installing, which runs the
+blocked postinstall. This works for project and global installs:
+
+```bash
+# project install
+bun add @gitlawb/zero
+bun pm trust @gitlawb/zero
+
+# global install
+bun add -g @gitlawb/zero
+bun pm -g trust @gitlawb/zero
+```
+
+Alternatives: allow the postinstall up front by adding
+`"trustedDependencies": ["@gitlawb/zero"]` to your project's package.json
+before `bun add`, or run the installer manually
+(`node node_modules/@gitlawb/zero/scripts/postinstall.mjs`) on Bun versions
+that do not have `bun pm trust`.
+
 ### Install scripts
 
 Linux/macOS:
@@ -121,6 +148,13 @@ the key in the wizard:
 export OPENAI_API_KEY=sk-...
 export ANTHROPIC_API_KEY=...
 export GEMINI_API_KEY=...
+export LONGCAT_API_KEY=...
+```
+
+To configure Meituan LongCat (LongCat-2.0) directly, run:
+
+```bash
+zero providers setup longcat --set-active
 ```
 
 For local models, run Ollama or LM Studio and then use `zero setup` or
@@ -152,6 +186,7 @@ Common slash commands:
 | `/spec`, `/plan` | draft and review a plan before building |
 | `/image` | attach an image for vision-capable models |
 | `/resume`, `/rewind` | continue or roll back local sessions |
+| `/loop` | repeat a prompt or custom `/command` on an interval (`/loop 5m /babysit-prs`) or self-paced |
 | `/compact`, `/context` | manage context usage |
 | `/permissions`, `/tools` | inspect available tools and policy |
 | `/add-dir` | allow an extra write directory for this session |
@@ -249,14 +284,47 @@ zero cron             scheduled agent jobs
 zero update           check for newer releases
 ```
 
+## Extending Zero
+
+### Project and personal instructions
+
+Zero appends project-specific guidance to the system prompt from the first
+`AGENTS.md`, `ZERO.md`, or `.zero/AGENTS.md` file found in each directory from
+the git root down to your current working directory (checked in that order
+per directory). Files are injected general-to-specific, capped at 8 KiB per
+file and 32 KiB total.
+
+A personal `ZERO.md` under `config.UserConfigDir()/zero/ZERO.md`
+(`$XDG_CONFIG_HOME/zero/ZERO.md` or `~/.config/zero/ZERO.md` on Linux/macOS,
+`%AppData%\Roaming\zero\ZERO.md` on Windows) applies across every workspace, ahead of any project guidelines.
+
+### Plugins
+
+Plugins are discovered from `~/.config/zero/plugins/<name>/plugin.json` (user
+scope — `$XDG_CONFIG_HOME` or `~/.config` on every OS, independent of the
+`config.UserConfigDir()` path used above) and `<cwd>/.zero/plugins/<name>/plugin.json`
+(project scope — resolved from the current working directory, not the repo
+root), and managed with `zero plugins`. A manifest can declare:
+
+- `tools` — custom tools (`command`, `args`, `inputSchema`, and a
+  `permission` of `prompt` or `deny`; `allow` is honored only when manifest tool
+  auto-approval is enabled)
+- `hooks` — commands run on `beforeTool`, `afterTool`, `sessionStart`, or
+  `sessionEnd`
+- `prompts` and `skills` — additional prompt/skill files
+
+MCP servers (`zero mcp`) and standalone markdown skills (`zero skills`) use
+the same extension points and can also be wired up outside of a plugin
+manifest.
+
 ## Appearance And Accessibility
 
 | Control | Effect |
 |---|---|
 | `NO_COLOR=<anything>` | disables color output |
-| `ZERO_THEME=auto|dark|light` | selects the startup theme |
-| `--theme auto|dark|light` | selects the TUI theme from the CLI |
-| `/theme auto|dark|light` | switches theme inside the TUI |
+| `ZERO_THEME=<name>` | selects the startup theme (`auto`, `dark`, `light`, or a color theme like `dracula`, `nord`, `gruvbox`, `tokyo-night`, `catppuccin`, `one-dark`, `solarized-dark`, `rose-pine`, `everforest`, `solarized-light`) |
+| `--theme <name>` | selects the TUI theme from the CLI (same names) |
+| `/theme` | opens the theme picker inside the TUI (live preview; `/theme <name>` switches directly) |
 | `ZERO_NO_FADE=1` | disables streaming fade animation |
 
 Meaning does not rely on color alone; diffs, permissions, and statuses also use

README_ZH.md

@@ -0,0 +1,297 @@
+<p align="center">
+  <img src="docs/assets/zero-logo.png" alt="Zero" width="385">
+</p>
+
+<p align="center"><strong>一个独属于你的终端编码智能体。</strong></p>
+
+<p align="center">
+  <a href="LICENSE"><img alt="license" src="https://img.shields.io/badge/license-MIT-blue"></a>
+  <img alt="Go 1.25+" src="https://img.shields.io/badge/Go-1.25+-00ADD8?logo=go&logoColor=white">
+  <img alt="25+ providers" src="https://img.shields.io/badge/providers-25+-34E2EA">
+  <br>
+  <a href="README.md">English</a> | <strong>中文</strong>
+</p>
+
+Zero 是一个用于本地终端的 AI 编码智能体。它可以检查仓库、编辑文件、运行命令、使用浏览器/终端辅助工具，并在你选择模型和权限级别的同时保持持久的本地会话。
+
+```bash
+zero
+zero exec "修复 ./pkg 中失败的测试"
+zero exec --output-format stream-json < turns.jsonl
+```
+
+## 为什么选择 Zero
+
+- **使用你想要的模型。** 支持 OpenAI、Anthropic、Gemini、Groq、OpenRouter、DeepSeek、Mistral、xAI、Qwen、Kimi、GitHub Models、Ollama、LM Studio，或任何 OpenAI/Anthropic 兼容端点。
+- **保持控制权。** 文件写入、Shell 命令、网络访问和工作区外写入都经过 Zero 的权限和沙箱策略。
+- **在终端中工作。** TUI 具有模型/提供商选择器、图片输入、斜杠命令、实时计划/工具渲染、回滚滚动、主题以及恢复/分叉支持。
+- **无 TUI 也能工作。** `zero exec` 可脚本化，支持文本/JSON/stream-JSON I/O、隔离的工作树、规范优先运行，以及用于 CI 的有意义的退出码。
+- **保持上下文本地化。** 会话存储在磁盘上，可搜索、可恢复，且 Zero 不会作为遥测数据上传。
+- **可扩展。** 使用 MCP 服务器、技能、插件、钩子和来自同一 CLI 的专业子智能体。
+
+## 安装
+
+### npm
+
+```bash
+npm install -g @gitlawb/zero
+zero
+```
+
+npm 包安装一个小型包装器以及与你平台匹配的 Zero 二进制文件（从 GitHub Releases 获取）。支持 Linux、macOS 和 Windows 的 x64 和 arm64。
+
+### Bun
+
+Bun 默认不运行依赖的生命周期脚本，因此获取 Zero 二进制文件的 `postinstall` 会被跳过，首次运行会失败并显示 `No native binary found next to the npm wrapper`。
+
+最简单的解决方法是安装后信任该包，这会运行被阻止的 postinstall。项目安装和全局安装均适用：
+
+```bash
+# 项目安装
+bun add @gitlawb/zero
+bun pm trust @gitlawb/zero
+
+# 全局安装
+bun add -g @gitlawb/zero
+bun pm -g trust @gitlawb/zero
+```
+
+其他方式：在 `bun add` 之前将 `"trustedDependencies": ["@gitlawb/zero"]` 添加到项目的 package.json 以提前允许 postinstall；或在不支持 `bun pm trust` 的旧版 Bun 上手动运行安装程序（`node node_modules/@gitlawb/zero/scripts/postinstall.mjs`）。
+
+### 安装脚本
+
+Linux/macOS：
+
+```bash
+curl -fsSL https://raw.githubusercontent.com/Gitlawb/zero/main/scripts/install.sh | bash
+```
+
+Windows PowerShell：
+
+```powershell
+irm https://raw.githubusercontent.com/Gitlawb/zero/main/scripts/install.ps1 | iex
+```
+
+### 从源码构建
+
+源码构建需要 Go 1.25+。
+
+```bash
+git clone https://github.com/Gitlawb/zero.git
+cd zero
+go run ./cmd/zero
+```
+
+发布安装器和 npm 包需要已发布的 GitHub Release 资源。如果你在首次公开发布之前进行测试，请从源码构建：
+
+```bash
+go build -o zero ./cmd/zero
+```
+
+在 Linux 上，如果你需要原生沙箱，还需要构建沙箱辅助程序：
+
+```bash
+go build -o zero-linux-sandbox ./cmd/zero-linux-sandbox
+go build -o zero-seccomp ./cmd/zero-seccomp   # 可选的兼容性包装器
+```
+
+将 `zero` 和 `zero-linux-sandbox` 放在 `PATH` 上的同一目录中（`~/.local/bin` 是一个好的默认选择）。macOS 不需要额外的辅助二进制文件。Windows 源码构建可以使用主 `zero.exe` 作为沙箱辅助程序；发布包仍然附带独立的 Windows 辅助可执行文件。
+
+更多安装细节：[docs/INSTALL.md](docs/INSTALL.md)。
+
+## 首次运行
+
+启动 TUI：
+
+```bash
+zero
+```
+
+设置向导帮助你选择提供商和模型。你也可以从命令行配置提供商：
+
+```bash
+zero setup
+zero providers list
+zero models list
+zero doctor
+```
+
+对于 API 提供商，在设置之前设置匹配的环境变量或在向导中输入密钥：
+
+```bash
+export OPENAI_API_KEY=sk-...
+export ANTHROPIC_API_KEY=...
+export GEMINI_API_KEY=...
+export LONGCAT_API_KEY=...
+```
+
+要直接配置美团 LongCat（LongCat-2.0），运行：
+
+```bash
+zero providers setup longcat --set-active
+```
+
+对于本地模型，运行 Ollama 或 LM Studio，然后使用 `zero setup` 或 `zero providers detect`。
+
+## 日常使用
+
+### 交互式 TUI
+
+```bash
+zero
+```
+
+常用控制：
+
+| 控制 | 操作 |
+|---|---|
+| `Enter` | 发送提示 |
+| `/` | 打开斜杠命令建议 |
+| `Shift+Tab` | 切换权限模式 |
+| `Ctrl+B` | 显示/隐藏侧边栏 |
+| `Ctrl+C` | 取消或退出 |
+
+常用斜杠命令：
+
+| 命令 | 用途 |
+|---|---|
+| `/model`、`/provider` | 切换活动模型/提供商 |
+| `/spec`、`/plan` | 在构建之前起草和审查计划 |
+| `/image` | 为视觉模型附加图片 |
+| `/resume`、`/rewind` | 继续或回滚本地会话 |
+| `/compact`、`/context` | 管理上下文使用 |
+| `/permissions`、`/tools` | 检查可用工具和策略 |
+| `/add-dir` | 为此会话授予额外的写入目录 |
+| `/theme`、`/doctor`、`/config` | 调整外观和检查设置 |
+
+### 无头 `exec` 模式
+
+```bash
+zero exec "解释 internal/agent/loop.go"
+zero exec --model claude-sonnet-4.5 "重构配置加载器"
+zero exec --use-spec "为 API 客户端添加速率限制"
+zero exec --worktree "在隔离的工作树中尝试迁移"
+zero exec --resume
+zero exec --fork <session-id> "尝试另一种方法"
+```
+
+编程使用：
+
+```bash
+zero exec --input-format stream-json --output-format stream-json < turns.jsonl
+```
+
+Stream-JSON 协议文档在 [docs/STREAM_JSON_PROTOCOL.md](docs/STREAM_JSON_PROTOCOL.md)。
+
+## 安全模型
+
+Zero 旨在使副作用可见。
+
+- 工作区读取默认允许。
+- 文件写入限制在工作区内，除非你授予其他目录。
+- Shell 命令、网络访问、破坏性命令和提权操作需要权限授权。
+- `--add-dir <path>` 和 `/add-dir <path>` 授予额外的写入根目录，而不会给智能体整个文件系统。
+- 不安全/自主模式是显式选择加入。
+- 在 Zero 控制的界面上，密钥会从工具输出和日志中被脱敏。
+
+示例：
+
+```bash
+zero --add-dir ../docs-site
+zero exec --add-dir ../shared "更新两个仓库"
+```
+
+沙箱行为可以通过以下方式检查：
+
+```bash
+zero sandbox policy
+zero sandbox grants list
+```
+
+## Web 和本地控制
+
+Zero 包含本地文件/搜索/编辑/Shell 工具、用于公共 URL 的 `web_fetch`，以及用于额外工具的 MCP 支持。
+
+对于本地开发服务器，使用 Shell 命令（如通过 `exec_command` 的 `curl`），这样正常的沙箱和权限策略就会生效。长时间运行的命令保持附加到后台终端会话，可以从 TUI 中列出或停止。
+
+npm 包还包含本地浏览器/终端工具使用的浏览器和终端辅助包。源码构建可以在它们位于 `PATH` 上或在 Zero 的本地控制设置中配置时使用相同的辅助工具。
+
+## 常用命令
+
+```text
+zero                  交互式 TUI
+zero exec             一次性或脚本化智能体运行
+zero setup            首次运行提供商设置
+zero auth             支持提供商的 OAuth/登录辅助
+zero models           模型注册表和能力
+zero providers        提供商配置和检测
+zero doctor           设置、密钥和连接检查
+zero context          上下文预算报告
+zero repo-map         确定性仓库映射
+zero repo-info        本地仓库摘要
+zero search | find    搜索本地会话历史
+zero sessions         检查、恢复、分叉和回滚会话
+zero spec             管理规范模式草稿
+zero specialist       管理专业子智能体
+zero skills           管理 Markdown 指令技能
+zero plugins          管理插件
+zero hooks            管理生命周期钩子
+zero mcp              管理 MCP 服务器和工具
+zero serve --mcp      通过 MCP stdio 暴露 Zero 工具
+zero sandbox          检查沙箱策略和授权
+zero worktrees        准备隔离的 Git 工作树
+zero verify           检测和运行本地验证检查
+zero changes          检查和提交本地 Git 变更
+zero usage            Token 使用量和估算成本
+zero cron             定时智能体任务
+zero update           检查更新版本
+```
+
+## 外观和无障碍
+
+| 控制 | 效果 |
+|---|---|
+| `NO_COLOR=<任意值>` | 禁用颜色输出 |
+| `ZERO_THEME=<名称>` | 选择启动主题（`auto`、`dark`、`light`，或颜色主题如 `dracula`、`nord`、`gruvbox`、`tokyo-night`、`catppuccin`、`one-dark`、`solarized-dark`、`rose-pine`、`everforest`、`solarized-light`） |
+| `--theme <名称>` | 从 CLI 选择 TUI 主题（相同名称） |
+| `/theme` | 在 TUI 中打开主题选择器（实时预览；`/theme <名称>` 直接切换） |
+| `ZERO_NO_FADE=1` | 禁用流式淡入动画 |
+
+含义不仅仅依赖于颜色；差异、权限和状态也使用文本或标记符号。
+
+## 开发
+
+```bash
+go test ./...
+go run ./cmd/zero-release build
+go run ./cmd/zero-release smoke
+go run ./cmd/zero-perf-bench
+```
+
+交叉编译示例：
+
+```bash
+go run ./cmd/zero-release build --goos linux --goarch amd64
+go run ./cmd/zero-release build --goos windows --goarch amd64 --output dist/zero.exe
+```
+
+## 文档
+
+- [安装](docs/INSTALL.md)
+- [更新流程](docs/UPDATE.md)
+- [Stream-JSON 协议](docs/STREAM_JSON_PROTOCOL.md)
+- [专家](docs/SPECIALISTS.md)
+- [GitHub Action](docs/GITHUB_ACTION.md)
+- [基准测试](docs/BENCHMARK.md)
+- [性能](docs/PERFORMANCE.md)
+- [智能体评估](docs/AGENT_EVALS.md)
+
+## 贡献
+
+欢迎贡献。阅读 [CONTRIBUTING.md](CONTRIBUTING.md)，运行相关测试，然后提交一个聚焦的拉取请求。
+
+安全报告应遵循 [SECURITY.md](SECURITY.md)。
+
+## 许可证
+
+Zero 基于 [MIT 许可证](LICENSE) 发布。

action.yml

@@ -27,14 +27,14 @@ inputs:
   api-key:
     description: >-
       The provider API key. Pass it from a repository or organization secret;
-      it is exported into the step environment and never written to the log.
+      it is exported only for the ZERO step and never written to the log.
     required: false
     default: ""
   api-key-env:
     description: >-
       Name of the environment variable the chosen provider reads its key from
       (for example OPENAI_API_KEY or ANTHROPIC_API_KEY). When set together with
-      api-key, that variable is exported for the run. Leave the action
+      api-key, that variable is exported only for the ZERO step. Leave the action
       provider-agnostic by passing the env name your provider expects rather
       than hardcoding one here.
     required: false
@@ -180,6 +180,7 @@ runs:
         ZERO_INPUT_ADD_DIR: ${{ inputs.add-dir }}
         ZERO_INPUT_WORKTREE: ${{ inputs.worktree }}
         ZERO_INPUT_OUTPUT_FORMAT: ${{ inputs.output-format }}
+        ACTION_PATH: ${{ github.action_path }}
         RUNNER_TEMP: ${{ runner.temp }}
       run: |
         set -euo pipefail
@@ -192,10 +193,9 @@ runs:
             echo "::error::api-key-env must be a valid environment variable name." >&2
             exit 2
           fi
-          # Export in-process so the zero invocation in this same run block sees
-          # it, and append to GITHUB_ENV so any later steps inherit it too.
+          # Export in-process for this Run ZERO step so zero can read it; do
+          # not persist provider credentials to later job steps.
           export "${ZERO_INPUT_API_KEY_ENV}=${ZERO_INPUT_API_KEY}"
-          printf '%s=%s\n' "${ZERO_INPUT_API_KEY_ENV}" "${ZERO_INPUT_API_KEY}" >> "${GITHUB_ENV}"
         fi
         if [ -n "${ZERO_INPUT_PROVIDER}" ]; then
           export ZERO_PROVIDER="${ZERO_INPUT_PROVIDER}"
@@ -248,10 +248,9 @@ runs:
         echo "exit-code=${code}" >> "${GITHUB_OUTPUT}"
         echo "output-file=${output_file}" >> "${GITHUB_OUTPUT}"
 
-        # Best-effort one-line summary: last non-empty line of the captured output.
-        summary="$(grep -v '^[[:space:]]*$' "${output_file}" | tail -n 1 || true)"
-        # Keep the summary single-line and bounded for the step output.
-        summary="$(printf '%s' "${summary}" | cut -c1-280)"
+        # Best-effort one-line summary. Structured output ends with lifecycle
+        # metadata, so parse final/error events instead of tailing stdout.
+        summary="$(node "${ACTION_PATH}/scripts/action-summary.mjs" "${ZERO_INPUT_OUTPUT_FORMAT}" "${output_file}" || true)"
         {
           echo "summary<<ZERO_EOF"
           printf '%s\n' "${summary}"

bin/zero.js

@@ -54,8 +54,28 @@ const nativePath = join(packageRoot, zeroBinaryName());
 const localControlHelpers = localControlHelperManifest(packageRoot);
 
 if (!existsSync(nativePath)) {
+  const postinstallScript = join(packageRoot, 'scripts', 'postinstall.mjs');
+  const ranByBun = process.execPath.includes('bun') || !!process.versions?.bun;
   console.error(
-    '[zero] No native binary found next to the npm wrapper. Reinstall the zero package or run `go run ./cmd/zero-release build` from the repository.'
+    '[zero] No native binary found next to the npm wrapper.\n' +
+      'The platform binary is fetched at install time by a postinstall script,\n' +
+      'which did not run (or was skipped) for this install.\n' +
+      '\n' +
+      'Fix it now by running the installer manually:\n' +
+      `  node "${postinstallScript}"\n` +
+      '\n' +
+      (ranByBun
+        ? 'You installed with Bun, which does not run dependency lifecycle scripts\n' +
+          'by default. Trust the package to run the blocked postinstall:\n' +
+          '  bun pm trust @gitlawb/zero       (project install)\n' +
+          '  bun pm -g trust @gitlawb/zero    (global install)\n' +
+          'On Bun versions without `bun pm trust`, add\n' +
+          '  "trustedDependencies": ["@gitlawb/zero"]\n' +
+          'to your project package.json and reinstall.\n' +
+          '\n'
+        : '') +
+      'If that fails, build from source: https://github.com/Gitlawb/zero\n' +
+      '(go run ./cmd/zero, requires Go 1.25+).',
   );
   process.exit(1);
 }

docs/GITHUB_ACTION.md

@@ -41,8 +41,9 @@ jobs:
 > provider-neutral: `api-key-env` is the environment variable name the chosen
 > provider reads its key from (for example `ANTHROPIC_API_KEY`, `GEMINI_API_KEY`,
 > `OPENROUTER_API_KEY`), and `api-key` is the secret value. The action exports
-> that variable for the run and never prints it. If your repository already
-> commits a `.zero/config.json` with an active provider, you can omit `provider`.
+> that variable only for the ZERO step and never prints it. If your repository
+> already commits a `.zero/config.json` with an active provider, you can omit
+> `provider`.
 
 ## Inputs
 
@@ -51,8 +52,8 @@ jobs:
 | `prompt` | one of `prompt`/`prompt-file` | `""` | The instruction for ZERO to execute. |
 | `prompt-file` | one of `prompt`/`prompt-file` | `""` | Path (relative to `working-directory`) to a file whose contents are the prompt. |
 | `provider` | no | `""` | Provider id to activate (e.g. `openai`, `anthropic`, `gemini`, `ollama`, or any compatible endpoint). |
-| `api-key` | no | `""` | The provider API key. Pass from a secret; never logged. |
-| `api-key-env` | no | `""` | Env var name the provider reads its key from (e.g. `OPENAI_API_KEY`). Exported with `api-key`. |
+| `api-key` | no | `""` | The provider API key. Pass from a secret; exported only for the ZERO step and never logged. |
+| `api-key-env` | no | `""` | Env var name the provider reads its key from (e.g. `OPENAI_API_KEY`). Exported only for the ZERO step when used with `api-key`. |
 | `model` | no | `""` | Model id. Defaults to the resolved provider's default. |
 | `mode` | no | `""` | Run mode (`zero exec --mode`), e.g. `smart`, `deep`, `fast`. |
 | `auto` | no | `low` | Autonomy ceiling (`zero exec --auto`): `low`, `medium`, or `high`. Conservative by default. |
@@ -162,9 +163,10 @@ jobs:
 ## Security notes
 
 - **Secrets are passed as Action secrets and never logged.** The action exports
-  `api-key` (under `api-key-env`) and `slack-webhook-url` into the step
-  environment only. They are never echoed, and ZERO redacts secret-shaped
-  strings from its own output.
+  `api-key` (under `api-key-env`) only for the ZERO step. `slack-webhook-url`
+  is passed only to the Slack post step. They are never echoed or persisted to
+  later workflow steps, and ZERO redacts secret-shaped strings from its own
+  output.
 - **The sandbox is always active.** This action never passes
   `--skip-permissions-unsafe`, so writes stay inside the checked-out repository
   (plus any roots you grant with `add-dir`). Unsafe mode is never enabled

docs/INSTALL.md

@@ -24,6 +24,51 @@ Requirements:
 - Node.js 18+
 - network access to npm and GitHub Releases
 
+## Bun
+
+Bun is "default-secure" and does not run lifecycle scripts of installed
+dependencies (only the installing project's own scripts), so the `postinstall`
+that fetches the Zero binary is silently skipped. The first run then fails with
+`No native binary found next to the npm wrapper`.
+
+The simplest fix is to trust the package after installing, which runs the
+blocked postinstall. This works for project and global installs:
+
+```bash
+# project install
+bun add @gitlawb/zero
+bun pm trust @gitlawb/zero
+
+# global install
+bun add -g @gitlawb/zero
+bun pm -g trust @gitlawb/zero
+```
+
+`bun pm untrusted` (or `bun pm -g untrusted`) lists the blocked postinstalls if
+you want to inspect before trusting.
+
+Alternatively, allow the postinstall to run at install time by adding the
+package to your project's `trustedDependencies` before installing:
+
+```json
+{
+  "trustedDependencies": ["@gitlawb/zero"]
+}
+```
+
+```bash
+bun add @gitlawb/zero
+```
+
+On Bun versions that do not have `bun pm trust`, run the installer manually
+after installing:
+
+```bash
+node node_modules/@gitlawb/zero/scripts/postinstall.mjs
+```
+
+Reference: <https://bun.sh/docs/pm/lifecycle>
+
 ## Linux And macOS Script
 
 Install the latest release:
@@ -128,6 +173,61 @@ Linux native sandboxing also requires Bubblewrap to be installed.
 
 macOS uses the system sandbox and does not need an extra helper binary.
 
+### Termux (Android)
+
+Zero can run natively on Android via [Termux](https://termux.dev/). Build with
+`GOOS=android` to avoid the `faccessat2` syscall that is blocked by Samsung's
+seccomp filter on Android:
+
+```bash
+# Install Go in Termux
+pkg install golang
+
+# Build Zero for Android
+git clone https://github.com/Gitlawb/zero.git
+cd zero
+CGO_ENABLED=0 GOOS=android GOARCH=arm64 go build -ldflags="-s -w" -o zero ./cmd/zero
+
+# Move into PATH
+mv zero ~/.local/bin/
+```
+
+> **Why `GOOS=android`?** Go 1.26+ detects `runtime.GOOS == "android"` and skips
+> the `faccessat2` syscall inside `os/exec.findExecutable`, falling back to
+> permission-bit checks. Without this flag, Android's seccomp sends SIGSYS and
+> kills the process whenever Zero looks up a binary on `PATH` (git, sh, etc.).
+
+**DNS.** Android does not expose `/etc/resolv.conf`. Go's pure-Go DNS resolver
+needs one. Use `proot` to bind-mount Termux's resolver config:
+
+```bash
+pkg install proot
+proot -b "$PREFIX/etc/resolv.conf:/etc/resolv.conf" zero
+```
+
+Create a wrapper at `~/.local/bin/zero` to avoid typing proot every time:
+
+```bash
+#!/data/data/com.termux/files/usr/bin/bash
+exec proot -b "$PREFIX/etc/resolv.conf:/etc/resolv.conf" ~/.local/bin/zero.bin "$@"
+```
+
+**Scroll.** On native Termux (not under PRoot), mouse scrolling works out of the
+box. The TUI uses Bubble Tea's `AllMotion` mouse mode by default. If you run Zero
+inside PRoot (e.g. through proot-distro), the scroll fix activates `CellMotion`
+to avoid PRoot's ptrace interference with the 1003 escape sequence.
+
+**Providers.** Zero works with any OpenAI-compatible provider on Termux. For
+example, to use OpenCode Zen's free tier:
+
+```bash
+zero providers add opencode \
+  --name opencode \
+  --model deepseek-v4-flash-free \
+  --base-url https://opencode.ai/zen/v1 \
+  --set-active
+```
+
 Windows source builds can use the main `zero.exe` as the command runner and setup
 helper through Zero's built-in self-dispatch path. If you want a release-style
 layout anyway, build the standalone helper executables next to `zero.exe`:

docs/STREAM_JSON_PROTOCOL.md

@@ -41,6 +41,7 @@ events when present.
 
 ```json
 { "schemaVersion": 2, "type": "run_start", "runId": "run_20260603_abc123", "sessionId": "zero_20260603100000_abc123", "cwd": "/repo", "provider": "openai", "model": "gpt-4.1", "apiModel": "gpt-4.1" }
+{ "schemaVersion": 2, "type": "reasoning", "runId": "run_20260603_abc123", "delta": "Thinking..." }
 { "schemaVersion": 2, "type": "text", "runId": "run_20260603_abc123", "delta": "..." }
 { "schemaVersion": 2, "type": "tool_call", "runId": "run_20260603_abc123", "id": "call_1", "name": "read_file", "args": { "path": "README.md" }, "sideEffect": "read" }
 { "schemaVersion": 2, "type": "permission_request", "runId": "run_20260603_abc123", "id": "call_2", "name": "write_file", "action": "prompt", "permission": "prompt", "permissionMode": "ask", "sideEffect": "write", "reason": "Creates or overwrites files." }
@@ -51,6 +52,10 @@ events when present.
 { "schemaVersion": 2, "type": "run_end", "runId": "run_20260603_abc123", "status": "success", "exitCode": 0 }
 ```
 
+`reasoning` events carry live model reasoning/status deltas for providers that
+stream them separately from answer text. They are liveness/progress events only:
+they are not folded into `text` or the final answer.
+
 Permission events may include structured sandbox metadata:
 
 ```json

internal/agent/additional_permissions_validation_test.go

@@ -0,0 +1,146 @@
+package agent
+
+import (
+	"context"
+	"strings"
+	"testing"
+
+	"github.com/Gitlawb/zero/internal/tools"
+)
+
+// A shell call that sets sandbox_permissions: with_additional_permissions but
+// omits (or malforms) additional_permissions can never succeed: the same
+// validation runs again at grant time regardless of the user's decision. It
+// must be rejected as a plain tool error BEFORE any permission prompt, not
+// surfaced as a confusing "approved but denied" prompt outcome.
+func TestExecuteToolCallRejectsMissingAdditionalPermissionsBeforePrompt(t *testing.T) {
+	root := t.TempDir()
+	registry := tools.NewRegistry()
+	registry.Register(tools.NewBashTool(root))
+
+	promptCalls := 0
+	result, err := executeToolCall(
+		context.Background(),
+		registry,
+		ToolCall{
+			ID:        "c1",
+			Name:      "bash",
+			Arguments: `{"command":"echo hi","sandbox_permissions":"with_additional_permissions"}`,
+		},
+		PermissionModeAuto,
+		Options{
+			OnPermissionRequest: func(ctx context.Context, request PermissionRequest) (PermissionDecision, error) {
+				promptCalls++
+				return PermissionDecision{Action: PermissionDecisionAllow}, nil
+			},
+		},
+	)
+	if err != nil {
+		t.Fatalf("executeToolCall: %v", err)
+	}
+	if result.Status != tools.StatusError {
+		t.Fatalf("status = %q, want error", result.Status)
+	}
+	if promptCalls != 0 {
+		t.Fatalf("OnPermissionRequest called %d times, want 0 (must reject before prompting)", promptCalls)
+	}
+}
+
+// The mismatched-flag case (additional_permissions present without the
+// sandbox_permissions flag) must be rejected the same way.
+func TestExecuteToolCallRejectsAdditionalPermissionsWithoutFlagBeforePrompt(t *testing.T) {
+	root := t.TempDir()
+	registry := tools.NewRegistry()
+	registry.Register(tools.NewBashTool(root))
+
+	promptCalls := 0
+	result, err := executeToolCall(
+		context.Background(),
+		registry,
+		ToolCall{
+			ID:        "c1",
+			Name:      "bash",
+			Arguments: `{"command":"echo hi","additional_permissions":{"network":{"enabled":true}}}`,
+		},
+		PermissionModeAuto,
+		Options{
+			OnPermissionRequest: func(ctx context.Context, request PermissionRequest) (PermissionDecision, error) {
+				promptCalls++
+				return PermissionDecision{Action: PermissionDecisionAllow}, nil
+			},
+		},
+	)
+	if err != nil {
+		t.Fatalf("executeToolCall: %v", err)
+	}
+	if result.Status != tools.StatusError {
+		t.Fatalf("status = %q, want error", result.Status)
+	}
+	if promptCalls != 0 {
+		t.Fatalf("OnPermissionRequest called %d times, want 0 (must reject before prompting)", promptCalls)
+	}
+}
+
+// The rejection message must show a concrete, correctly-shaped example: a
+// model that gets this wrong once (attaching sandbox_permissions out of habit
+// from an earlier call, without a valid payload) needs enough in the error to
+// self-correct on retry instead of repeating the identical mistake.
+func TestExecuteToolCallMissingAdditionalPermissionsErrorIsActionable(t *testing.T) {
+	root := t.TempDir()
+	registry := tools.NewRegistry()
+	registry.Register(tools.NewBashTool(root))
+
+	result, err := executeToolCall(
+		context.Background(),
+		registry,
+		ToolCall{
+			ID:        "c1",
+			Name:      "bash",
+			Arguments: `{"command":"ping example.com","sandbox_permissions":"with_additional_permissions"}`,
+		},
+		PermissionModeAuto,
+		Options{},
+	)
+	if err != nil {
+		t.Fatalf("executeToolCall: %v", err)
+	}
+	if !strings.Contains(result.Output, `{"network": {"enabled": true}}`) {
+		t.Fatalf("output = %q, want a concrete additional_permissions example", result.Output)
+	}
+	if !strings.Contains(result.Output, "omit sandbox_permissions") {
+		t.Fatalf("output = %q, want guidance that the flag can simply be omitted", result.Output)
+	}
+}
+
+// A well-formed additional_permissions payload must still reach the normal
+// permission prompt: this fix only rejects malformed payloads.
+func TestExecuteToolCallStillPromptsForValidAdditionalPermissions(t *testing.T) {
+	root := t.TempDir()
+	registry := tools.NewRegistry()
+	registry.Register(tools.NewBashTool(root))
+
+	promptCalls := 0
+	_, err := executeToolCall(
+		context.Background(),
+		registry,
+		ToolCall{
+			ID:   "c1",
+			Name: "bash",
+			Arguments: `{"command":"echo hi","sandbox_permissions":"with_additional_permissions",` +
+				`"additional_permissions":{"network":{"enabled":true}}}`,
+		},
+		PermissionModeAuto,
+		Options{
+			OnPermissionRequest: func(ctx context.Context, request PermissionRequest) (PermissionDecision, error) {
+				promptCalls++
+				return PermissionDecision{Action: PermissionDecisionDeny}, nil
+			},
+		},
+	)
+	if err != nil {
+		t.Fatalf("executeToolCall: %v", err)
+	}
+	if promptCalls != 1 {
+		t.Fatalf("OnPermissionRequest called %d times, want exactly 1 for a valid elevation request", promptCalls)
+	}
+}

internal/agent/async_diagnostics.go

@@ -0,0 +1,185 @@
+package agent
+
+import (
+	"context"
+	"path/filepath"
+	"sort"
+	"strings"
+	"sync"
+	"time"
+)
+
+// Post-edit LSP diagnostics used to run synchronously inside edit_file /
+// write_file: every edit blocked its tool result on a fresh-publish wait with
+// a quiet-period debounce (≥300ms floor that reset on each publish, 10s cap) —
+// the largest recurring latency tax in an edit-heavy session. Collection is
+// now asynchronous: a file is enqueued the moment a mutating tool finishes, a
+// single worker checks it while the rest of the tool batch (and any
+// self-correct verification) runs, and the loop drains completed results just
+// before building the NEXT provider request, appending errors as a user-role
+// nudge. The model always issues another request after an edit turn — it has
+// to read its tool results — so it still sees introduced errors at the same
+// decision point as before, without any tool call stalling on the language
+// server.
+//
+// A file re-edited before its check runs is checked once in its final state
+// (the checker re-reads the file); a file re-edited after its check completed
+// is re-enqueued and the newer result wins. When the worker is still busy at
+// drain time, the drain gives up after a short wait and the results are
+// delivered on the following turn instead — they stay accurate because any
+// further edit re-enqueues the file.
+
+// asyncDiagnosticsDrainTimeout bounds how long a turn waits on the in-flight
+// check before deferring delivery to the next turn. Most checks finish well
+// under it (debounce floor + analysis); a cold language server on a large
+// repo does not get to stall the loop the way the old inline 10s cap could.
+// A var so tests can shorten the wait.
+var asyncDiagnosticsDrainTimeout = 3 * time.Second
+
+// asyncDiagnosticsFinalDrainTimeout is the wait used when the run is about to
+// FINALIZE (natural completion or the max-turns summary): there is no later
+// turn to defer to, so the gate waits out the old inline-era budget rather
+// than silently dropping an error the last edit introduced. A var so tests
+// can shorten the wait.
+var asyncDiagnosticsFinalDrainTimeout = 10 * time.Second
+
+// asyncDiagnosticsNudge prefixes the drained diagnostics blocks; phrased like
+// the old inline block so the model reacts the same way.
+const asyncDiagnosticsNudge = "Diagnostics after your recent edits (fix any errors you introduced):\n"
+
+type asyncDiagnostics struct {
+	check         func(context.Context, string) string
+	workspaceRoot string
+
+	mu      sync.Mutex
+	queue   []string
+	queued  map[string]bool
+	results map[string]string
+	// working is non-nil while the worker goroutine runs and is closed when it
+	// exits, so drain can wait for quiescence without polling.
+	working chan struct{}
+}
+
+// newAsyncDiagnostics returns nil when check is nil, and every method on a nil
+// receiver is a no-op — callers wire it unconditionally.
+func newAsyncDiagnostics(check func(context.Context, string) string, workspaceRoot string) *asyncDiagnostics {
+	if check == nil {
+		return nil
+	}
+	return &asyncDiagnostics{
+		check:         check,
+		workspaceRoot: workspaceRoot,
+		queued:        map[string]bool{},
+		results:       map[string]string{},
+	}
+}
+
+// enqueue schedules changed files for a background check. Paths are
+// workspace-relative from Result.ChangedFiles (absolute when under a granted
+// extra write root), mirroring SelfCorrector's resolution.
+func (diagnostics *asyncDiagnostics) enqueue(ctx context.Context, files []string) {
+	if diagnostics == nil || len(files) == 0 {
+		return
+	}
+	diagnostics.mu.Lock()
+	defer diagnostics.mu.Unlock()
+	for _, file := range files {
+		path := diagnostics.absPath(file)
+		if diagnostics.queued[path] {
+			continue
+		}
+		diagnostics.queued[path] = true
+		// A completed result for this file describes a pre-edit state now.
+		delete(diagnostics.results, path)
+		diagnostics.queue = append(diagnostics.queue, path)
+	}
+	if diagnostics.working == nil && len(diagnostics.queue) > 0 {
+		done := make(chan struct{})
+		diagnostics.working = done
+		go diagnostics.work(ctx, done)
+	}
+}
+
+func (diagnostics *asyncDiagnostics) work(ctx context.Context, done chan struct{}) {
+	defer close(done)
+	for {
+		diagnostics.mu.Lock()
+		if len(diagnostics.queue) == 0 || ctx.Err() != nil {
+			diagnostics.working = nil
+			diagnostics.mu.Unlock()
+			return
+		}
+		path := diagnostics.queue[0]
+		diagnostics.queue = diagnostics.queue[1:]
+		// Un-mark before checking so an edit that lands DURING the check
+		// re-enqueues the file and it is checked again in its final state.
+		delete(diagnostics.queued, path)
+		diagnostics.mu.Unlock()
+
+		block := diagnostics.check(ctx, path)
+
+		diagnostics.mu.Lock()
+		if block != "" {
+			diagnostics.results[path] = block
+		} else {
+			delete(diagnostics.results, path)
+		}
+		diagnostics.mu.Unlock()
+	}
+}
+
+// drain waits briefly for the worker to go quiet, then formats and consumes
+// all completed error blocks as one nudge. Returns "" when there is nothing
+// to report or the worker is still busy (results then surface next turn).
+func (diagnostics *asyncDiagnostics) drain(ctx context.Context) string {
+	return diagnostics.drainWithin(ctx, asyncDiagnosticsDrainTimeout)
+}
+
+// drainFinal is drain with the finalization budget: called before the run's
+// last model request, where an undelivered result would otherwise be lost.
+func (diagnostics *asyncDiagnostics) drainFinal(ctx context.Context) string {
+	return diagnostics.drainWithin(ctx, asyncDiagnosticsFinalDrainTimeout)
+}
+
+func (diagnostics *asyncDiagnostics) drainWithin(ctx context.Context, timeout time.Duration) string {
+	if diagnostics == nil {
+		return ""
+	}
+	diagnostics.mu.Lock()
+	busy := diagnostics.working
+	diagnostics.mu.Unlock()
+	if busy != nil {
+		timer := time.NewTimer(timeout)
+		defer timer.Stop()
+		select {
+		case <-busy:
+		case <-timer.C:
+			return ""
+		case <-ctx.Done():
+			return ""
+		}
+	}
+	diagnostics.mu.Lock()
+	defer diagnostics.mu.Unlock()
+	if len(diagnostics.results) == 0 {
+		return ""
+	}
+	paths := make([]string, 0, len(diagnostics.results))
+	for path := range diagnostics.results {
+		paths = append(paths, path)
+	}
+	sort.Strings(paths)
+	blocks := make([]string, 0, len(paths))
+	for _, path := range paths {
+		blocks = append(blocks, diagnostics.results[path])
+	}
+	diagnostics.results = map[string]string{}
+	return asyncDiagnosticsNudge + strings.Join(blocks, "\n")
+}
+
+func (diagnostics *asyncDiagnostics) absPath(path string) string {
+	if filepath.IsAbs(path) {
+		return path
+	}
+	return filepath.Join(diagnostics.workspaceRoot, path)
+}

internal/agent/async_diagnostics_test.go

@@ -0,0 +1,338 @@
+package agent
+
+import (
+	"context"
+	"path/filepath"
+	"strings"
+	"sync"
+	"testing"
+	"time"
+
+	"github.com/Gitlawb/zero/internal/tools"
+	"github.com/Gitlawb/zero/internal/zeroruntime"
+)
+
+func TestAsyncDiagnosticsNilCollectorNoOps(t *testing.T) {
+	var diagnostics *asyncDiagnostics
+	if diagnostics != newAsyncDiagnostics(nil, "/ws") {
+		t.Fatal("nil check must yield a nil collector")
+	}
+	diagnostics.enqueue(context.Background(), []string{"a.go"})
+	if nudge := diagnostics.drain(context.Background()); nudge != "" {
+		t.Fatalf("nil collector drain = %q, want empty", nudge)
+	}
+}
+
+func TestAsyncDiagnosticsCollectsAndDrainsOnce(t *testing.T) {
+	// A real absolute root: a literal "/ws" is not filepath.IsAbs on Windows
+	// (no drive letter), which is exactly what the assertions below check.
+	root := t.TempDir()
+	var mu sync.Mutex
+	var checked []string
+	check := func(_ context.Context, absPath string) string {
+		mu.Lock()
+		checked = append(checked, absPath)
+		mu.Unlock()
+		if filepath.Base(absPath) == "clean.go" {
+			return ""
+		}
+		return "ERR " + filepath.Base(absPath)
+	}
+	diagnostics := newAsyncDiagnostics(check, root)
+	diagnostics.enqueue(context.Background(), []string{"broken.go", "clean.go"})
+
+	nudge := diagnostics.drain(context.Background())
+	if !strings.HasPrefix(nudge, asyncDiagnosticsNudge) {
+		t.Fatalf("nudge missing prefix: %q", nudge)
+	}
+	if !strings.Contains(nudge, "ERR broken.go") || strings.Contains(nudge, "clean.go") {
+		t.Fatalf("nudge = %q, want broken.go errors only", nudge)
+	}
+	mu.Lock()
+	for _, path := range checked {
+		if !filepath.IsAbs(path) || !strings.HasPrefix(path, root+string(filepath.Separator)) {
+			t.Fatalf("check received %q, want path under workspace root %q", path, root)
+		}
+	}
+	mu.Unlock()
+	if second := diagnostics.drain(context.Background()); second != "" {
+		t.Fatalf("second drain = %q, want empty (results are consumed)", second)
+	}
+}
+
+// A check still running at drain time defers delivery to the next drain
+// instead of blocking the turn.
+func TestAsyncDiagnosticsSlowCheckDefersToNextDrain(t *testing.T) {
+	previous := asyncDiagnosticsDrainTimeout
+	asyncDiagnosticsDrainTimeout = 20 * time.Millisecond
+	defer func() { asyncDiagnosticsDrainTimeout = previous }()
+
+	release := make(chan struct{})
+	check := func(_ context.Context, absPath string) string {
+		<-release
+		return "ERR " + filepath.Base(absPath)
+	}
+	diagnostics := newAsyncDiagnostics(check, "/ws")
+	diagnostics.enqueue(context.Background(), []string{"slow.go"})
+
+	start := time.Now()
+	if nudge := diagnostics.drain(context.Background()); nudge != "" {
+		t.Fatalf("drain during in-flight check = %q, want empty", nudge)
+	}
+	if elapsed := time.Since(start); elapsed > time.Second {
+		t.Fatalf("drain blocked %v, want bounded by the (shortened) timeout", elapsed)
+	}
+
+	close(release)
+	deadline := time.Now().Add(2 * time.Second)
+	for {
+		if nudge := diagnostics.drain(context.Background()); nudge != "" {
+			if !strings.Contains(nudge, "ERR slow.go") {
+				t.Fatalf("deferred nudge = %q", nudge)
+			}
+			return
+		}
+		if time.Now().After(deadline) {
+			t.Fatal("deferred diagnostics never delivered")
+		}
+		time.Sleep(5 * time.Millisecond)
+	}
+}
+
+// Re-enqueueing a file after its check completed replaces the stale result.
+func TestAsyncDiagnosticsReEditReplacesResult(t *testing.T) {
+	var mu sync.Mutex
+	response := "ERR old"
+	check := func(context.Context, string) string {
+		mu.Lock()
+		defer mu.Unlock()
+		return response
+	}
+	diagnostics := newAsyncDiagnostics(check, "/ws")
+
+	diagnostics.enqueue(context.Background(), []string{"a.go"})
+	waitForIdle(t, diagnostics)
+	mu.Lock()
+	response = "ERR new"
+	mu.Unlock()
+	diagnostics.enqueue(context.Background(), []string{"a.go"})
+
+	nudge := diagnostics.drain(context.Background())
+	if !strings.Contains(nudge, "ERR new") || strings.Contains(nudge, "ERR old") {
+		t.Fatalf("nudge = %q, want only the re-check result", nudge)
+	}
+}
+
+func waitForIdle(t *testing.T, diagnostics *asyncDiagnostics) {
+	t.Helper()
+	deadline := time.Now().Add(2 * time.Second)
+	for {
+		diagnostics.mu.Lock()
+		busy := diagnostics.working
+		diagnostics.mu.Unlock()
+		if busy == nil {
+			return
+		}
+		if time.Now().After(deadline) {
+			t.Fatal("worker never went idle")
+		}
+		time.Sleep(2 * time.Millisecond)
+	}
+}
+
+// changedFilesTool is a fake mutating tool reporting a changed file without
+// touching the filesystem.
+type changedFilesTool struct{}
+
+func (changedFilesTool) Name() string        { return "fake_edit" }
+func (changedFilesTool) Description() string { return "test mutating tool" }
+func (changedFilesTool) Parameters() tools.Schema {
+	return tools.Schema{Type: "object", Properties: map[string]tools.PropertySchema{}}
+}
+func (changedFilesTool) Safety() tools.Safety {
+	return tools.Safety{SideEffect: tools.SideEffectWrite, Permission: tools.PermissionAllow, Reason: "test"}
+}
+func (changedFilesTool) Run(context.Context, map[string]any) tools.Result {
+	return tools.Result{Status: tools.StatusOK, Output: "edited", ChangedFiles: []string{"main.go"}}
+}
+
+// End-to-end through Run: an edit turn does NOT carry diagnostics in the tool
+// result (the old inline path), but the next request contains the background
+// diagnostics nudge, before the model finalizes.
+func TestRunDeliversAsyncDiagnosticsNudgeNextTurn(t *testing.T) {
+	registry := tools.NewRegistry()
+	registry.Register(changedFilesTool{})
+	provider := &mockProvider{
+		turns: [][]zeroruntime.StreamEvent{
+			{
+				{Type: zeroruntime.StreamEventToolCallStart, ToolCallID: "call-1", ToolName: "fake_edit"},
+				{Type: zeroruntime.StreamEventToolCallDelta, ToolCallID: "call-1", ArgumentsFragment: `{}`},
+				{Type: zeroruntime.StreamEventToolCallEnd, ToolCallID: "call-1"},
+				{Type: zeroruntime.StreamEventDone},
+			},
+			{
+				{Type: zeroruntime.StreamEventText, Content: "done"},
+				{Type: zeroruntime.StreamEventDone},
+			},
+		},
+	}
+	root := t.TempDir()
+	var checkedPath string
+	var mu sync.Mutex
+
+	result, err := Run(context.Background(), "fix main.go", provider, Options{
+		Registry: registry,
+		Cwd:      root,
+		FileDiagnostics: func(_ context.Context, absPath string) string {
+			mu.Lock()
+			checkedPath = absPath
+			mu.Unlock()
+			return "main.go:1:1 error: boom"
+		},
+	})
+	if err != nil {
+		t.Fatal(err)
+	}
+	if result.FinalAnswer != "done" {
+		t.Fatalf("final answer = %q", result.FinalAnswer)
+	}
+	if len(provider.requests) != 2 {
+		t.Fatalf("expected 2 provider turns, got %d", len(provider.requests))
+	}
+
+	// The tool result itself must NOT block on / embed diagnostics.
+	for _, message := range provider.requests[1].Messages {
+		if message.Role == zeroruntime.MessageRoleTool && strings.Contains(message.Content, "boom") {
+			t.Fatalf("diagnostics leaked into the tool result: %q", message.Content)
+		}
+	}
+	// The nudge must be present in the second request as a user message.
+	found := false
+	for _, message := range provider.requests[1].Messages {
+		if message.Role == zeroruntime.MessageRoleUser && strings.HasPrefix(message.Content, asyncDiagnosticsNudge) {
+			if !strings.Contains(message.Content, "boom") {
+				t.Fatalf("nudge missing diagnostics: %q", message.Content)
+			}
+			found = true
+		}
+	}
+	if !found {
+		t.Fatal("no async diagnostics nudge in the follow-up request")
+	}
+	mu.Lock()
+	defer mu.Unlock()
+	if want := filepath.Join(root, "main.go"); checkedPath != want {
+		t.Fatalf("checked path = %q, want %q", checkedPath, want)
+	}
+}
+
+// When the run hits the maxTurns ceiling right after an edit turn, the
+// final-answer request must still carry the diagnostics nudge — otherwise an
+// error introduced by the last edit would go unreported in the summary.
+func TestRunDrainsAsyncDiagnosticsBeforeMaxTurnsFinalAnswer(t *testing.T) {
+	registry := tools.NewRegistry()
+	registry.Register(changedFilesTool{})
+	provider := &mockProvider{
+		turns: [][]zeroruntime.StreamEvent{
+			{
+				{Type: zeroruntime.StreamEventToolCallStart, ToolCallID: "call-1", ToolName: "fake_edit"},
+				{Type: zeroruntime.StreamEventToolCallDelta, ToolCallID: "call-1", ArgumentsFragment: `{}`},
+				{Type: zeroruntime.StreamEventToolCallEnd, ToolCallID: "call-1"},
+				{Type: zeroruntime.StreamEventDone},
+			},
+			{
+				{Type: zeroruntime.StreamEventText, Content: "summary"},
+				{Type: zeroruntime.StreamEventDone},
+			},
+		},
+	}
+
+	result, err := Run(context.Background(), "fix main.go", provider, Options{
+		Registry: registry,
+		Cwd:      t.TempDir(),
+		MaxTurns: 1,
+		FileDiagnostics: func(context.Context, string) string {
+			return "main.go:1:1 error: boom"
+		},
+	})
+	if err != nil {
+		t.Fatal(err)
+	}
+	if result.FinalAnswer != "summary" {
+		t.Fatalf("final answer = %q", result.FinalAnswer)
+	}
+	if len(provider.requests) != 2 {
+		t.Fatalf("expected the max-turns final-answer request, got %d requests", len(provider.requests))
+	}
+	found := false
+	for _, message := range provider.requests[1].Messages {
+		if message.Role == zeroruntime.MessageRoleUser && strings.HasPrefix(message.Content, asyncDiagnosticsNudge) {
+			found = true
+		}
+	}
+	if !found {
+		t.Fatal("max-turns final-answer request missing the diagnostics nudge")
+	}
+}
+
+// When the model finalizes while a slow check is still in flight (the
+// per-turn drain missed it), the finalization gate must wait it out and give
+// the model one more turn with the nudge instead of dropping the errors.
+func TestRunFinalizationGateDeliversLateDiagnostics(t *testing.T) {
+	previous := asyncDiagnosticsDrainTimeout
+	asyncDiagnosticsDrainTimeout = time.Millisecond
+	defer func() { asyncDiagnosticsDrainTimeout = previous }()
+
+	registry := tools.NewRegistry()
+	registry.Register(changedFilesTool{})
+	provider := &mockProvider{
+		turns: [][]zeroruntime.StreamEvent{
+			{
+				{Type: zeroruntime.StreamEventToolCallStart, ToolCallID: "call-1", ToolName: "fake_edit"},
+				{Type: zeroruntime.StreamEventToolCallDelta, ToolCallID: "call-1", ArgumentsFragment: `{}`},
+				{Type: zeroruntime.StreamEventToolCallEnd, ToolCallID: "call-1"},
+				{Type: zeroruntime.StreamEventDone},
+			},
+			{
+				{Type: zeroruntime.StreamEventText, Content: "done"},
+				{Type: zeroruntime.StreamEventDone},
+			},
+			{
+				{Type: zeroruntime.StreamEventText, Content: "done after fix"},
+				{Type: zeroruntime.StreamEventDone},
+			},
+		},
+	}
+
+	result, err := Run(context.Background(), "fix main.go", provider, Options{
+		Registry: registry,
+		Cwd:      t.TempDir(),
+		FileDiagnostics: func(context.Context, string) string {
+			// Slower than the (shortened) per-turn drain, well under the
+			// finalization budget.
+			time.Sleep(100 * time.Millisecond)
+			return "main.go:1:1 error: boom"
+		},
+	})
+	if err != nil {
+		t.Fatal(err)
+	}
+	if result.FinalAnswer != "done after fix" {
+		t.Fatalf("final answer = %q, want the post-nudge answer", result.FinalAnswer)
+	}
+	if len(provider.requests) != 3 {
+		t.Fatalf("expected a third request carrying the late nudge, got %d requests", len(provider.requests))
+	}
+	found := false
+	for _, message := range provider.requests[2].Messages {
+		if message.Role == zeroruntime.MessageRoleUser && strings.HasPrefix(message.Content, asyncDiagnosticsNudge) {
+			if !strings.Contains(message.Content, "boom") {
+				t.Fatalf("nudge missing diagnostics: %q", message.Content)
+			}
+			found = true
+		}
+	}
+	if !found {
+		t.Fatal("finalization gate did not deliver the late diagnostics nudge")
+	}
+}

internal/agent/command_prefix.go

@@ -2,6 +2,7 @@ package agent
 
 import (
 	"fmt"
+	"runtime"
 	"strings"
 
 	"github.com/Gitlawb/zero/internal/sandbox"
@@ -27,10 +28,20 @@ func proposedCommandPrefix(toolName string, args map[string]any) []string {
 		}
 		return nil
 	}
-	for _, tokens := range segments {
-		if !knownSafeCommandSegment(tokens) && sandbox.ValidCommandPrefix(tokens) {
-			return append([]string(nil), tokens...)
+	// Only propose approving a prefix of one segment when every other segment
+	// in the command is independently known-safe. Once a prefix is approved,
+	// shellExecutionArgsForApproval escalates the whole command (every
+	// segment, not just the approved one) to bypass the sandbox, so offering a
+	// prefix that leaves an MSYS-prone (or otherwise unsafe) segment uncovered
+	// would let that segment run unsandboxed without ever being reviewed.
+	for index, tokens := range segments {
+		if knownSafeCommandSegment(tokens) {
+			continue
 		}
+		if !sandbox.ValidCommandPrefix(tokens) || !otherSegmentsKnownSafe(segments, index) {
+			return nil
+		}
+		return append([]string(nil), tokens...)
 	}
 	if len(segments) == 0 || !sandbox.ValidCommandPrefix(segments[0]) {
 		return nil
@@ -38,6 +49,20 @@ func proposedCommandPrefix(toolName string, args map[string]any) []string {
 	return append([]string(nil), segments[0]...)
 }
 
+// otherSegmentsKnownSafe reports whether every segment other than the one at
+// skip is known-safe on its own.
+func otherSegmentsKnownSafe(segments [][]string, skip int) bool {
+	for index, tokens := range segments {
+		if index == skip {
+			continue
+		}
+		if !knownSafeCommandSegment(tokens) {
+			return false
+		}
+	}
+	return true
+}
+
 func matchCommandPrefix(toolName string, args map[string]any, options Options) (sandbox.CommandPrefixGrant, bool, bool) {
 	if !isShellCommandTool(toolName) || options.Sandbox == nil {
 		return sandbox.CommandPrefixGrant{}, false, false
@@ -168,10 +193,6 @@ func cleanPrefixRule(prefix []string) []string {
 	return cleaned
 }
 
-func safeRequestedPrefix(prefix []string, command []string) bool {
-	return safeRequestedPrefixForSegments(prefix, [][]string{command})
-}
-
 func safeRequestedPrefixForSegments(prefix []string, segments [][]string) bool {
 	if len(prefix) == 0 || !sandbox.ValidCommandPrefix(prefix) {
 		return false
@@ -196,14 +217,6 @@ func safeRequestedPrefixForSegments(prefix []string, segments [][]string) bool {
 	return matched
 }
 
-func safeShellCommandTokens(command string) ([]string, bool) {
-	segments, ok := safeShellCommandSegments(command)
-	if !ok || len(segments) != 1 {
-		return nil, false
-	}
-	return segments[0], true
-}
-
 func safeShellCommandSegments(command string) ([][]string, bool) {
 	command = strings.TrimSpace(command)
 	if command == "" {
@@ -275,7 +288,11 @@ func knownSafeCommandSegment(command []string) bool {
 	if len(command) == 0 {
 		return false
 	}
-	switch commandName(command[0]) {
+	name := commandName(command[0])
+	if runtime.GOOS == "windows" && tools.MsysProneCommandName(name) {
+		return false
+	}
+	switch name {
 	case "cat", "cd", "cut", "echo", "expr", "false", "grep", "head", "id",
 		"ls", "nl", "paste", "pwd", "rev", "seq", "stat", "tail", "tr",
 		"true", "uname", "uniq", "wc", "which", "whoami":

internal/agent/command_prefix_test.go

@@ -1,9 +1,11 @@
 package agent
 
 import (
+	"runtime"
 	"testing"
 
 	"github.com/Gitlawb/zero/internal/sandbox"
+	"github.com/Gitlawb/zero/internal/tools"
 )
 
 func TestProposedCommandPrefixUsesSafeSimpleCommands(t *testing.T) {
@@ -35,12 +37,37 @@ func TestProposedCommandPrefixHonorsValidatedRequestedPrefix(t *testing.T) {
 
 func TestProposedCommandPrefixSupportsSegmentedCommands(t *testing.T) {
 	got := proposedCommandPrefix("bash", map[string]any{"command": "ps aux | head -5"})
+	if runtime.GOOS == "windows" {
+		// head is MSYS-prone on Windows (#458), so proposedCommandPrefix must
+		// not offer "ps aux" as a reusable prefix here: approving it would
+		// escalate the whole command, including the uncovered head segment,
+		// to bypass the sandbox unreviewed. See
+		// TestProposedCommandPrefixRejectsPrefixLeavingUnsafeTailUncovered for
+		// the platform-independent regression coverage of this behavior.
+		if got != nil {
+			t.Fatalf("expected no prefix on Windows because head is MSYS-prone, got %#v", got)
+		}
+		return
+	}
 	want := []string{"ps", "aux"}
 	if !equalStringSlices(got, want) {
 		t.Fatalf("prefix = %#v, want %#v", got, want)
 	}
 }
 
+// TestProposedCommandPrefixRejectsPrefixLeavingUnsafeTailUncovered guards
+// against proposedCommandPrefix offering to approve one segment of a
+// multi-segment command (e.g. "ps aux") while a different segment (e.g. "npm
+// install") is not known-safe. shellExecutionArgsForApproval escalates the
+// entire command once any prefix is approved, so an uncovered unsafe segment
+// would bypass the sandbox unreviewed. Uses npm, which is never known-safe on
+// any platform, so the assertion does not depend on runtime.GOOS.
+func TestProposedCommandPrefixRejectsPrefixLeavingUnsafeTailUncovered(t *testing.T) {
+	if got := proposedCommandPrefix("bash", map[string]any{"command": "ps aux && npm install"}); got != nil {
+		t.Fatalf("expected no prefix because npm segment is not known-safe, got %#v", got)
+	}
+}
+
 func TestProposedCommandPrefixHonorsRequestedPrefixAcrossSegments(t *testing.T) {
 	got := proposedCommandPrefix("bash", map[string]any{
 		"command":     "git status --short && git status --branch",
@@ -120,13 +147,35 @@ func TestProposedCommandPrefixRejectsUnsafeLaunchers(t *testing.T) {
 func TestMatchCommandPrefixCoversSegmentedCommandWithSafeTail(t *testing.T) {
 	engine := sandbox.NewEngine(sandbox.EngineOptions{WorkspaceRoot: t.TempDir()})
 	engine.GrantCommandPrefixForSession("bash", []string{"ps", "aux"})
+	// head is MSYS-prone on Windows (#458) and must not count as a known-safe tail.
+	command := "ps aux | head -5"
+	if runtime.GOOS == "windows" {
+		command = "ps aux | echo ok"
+	}
 
-	grant, ok, session := matchCommandPrefix("bash", map[string]any{"command": "ps aux | head -5"}, Options{Sandbox: engine})
+	grant, ok, session := matchCommandPrefix("bash", map[string]any{"command": command}, Options{Sandbox: engine})
 	if !ok || !session || !equalStringSlices(grant.Prefix, []string{"ps", "aux"}) {
 		t.Fatalf("match = %#v ok=%v session=%v, want session ps aux prefix", grant, ok, session)
 	}
 }
 
+func TestKnownSafeCommandSegmentRejectsMsysProneOnWindows(t *testing.T) {
+	if runtime.GOOS != "windows" {
+		t.Skip("windows-only known-safe MSYS guard")
+	}
+	for _, command := range [][]string{{"head", "-5"}, {"cat", "file.txt"}, {"grep", "pat"}} {
+		if knownSafeCommandSegment(command) {
+			t.Fatalf("expected %q to be unsafe on Windows, got known-safe", command)
+		}
+	}
+	if !knownSafeCommandSegment([]string{"echo", "ok"}) {
+		t.Fatal("expected echo to remain known-safe on Windows")
+	}
+	if !tools.MsysProneCommandName("head") {
+		t.Fatal("expected head to be MSYS-prone")
+	}
+}
+
 func TestMatchCommandPrefixRejectsUncoveredSegment(t *testing.T) {
 	engine := sandbox.NewEngine(sandbox.EngineOptions{WorkspaceRoot: t.TempDir()})
 	engine.GrantCommandPrefixForSession("bash", []string{"ps", "aux"})

internal/agent/compaction.go

@@ -21,12 +21,17 @@ import (
 // the agent loop owns the provider wiring.
 
 // defaultCompactionPreserveLast is how many trailing messages are kept verbatim
-// when the caller does not specify a preserve count.
-const defaultCompactionPreserveLast = 8
+// when the caller does not specify a preserve count. Kept in sync with the replay
+// reconstruction default in internal/sessions/replay.go.
+const defaultCompactionPreserveLast = 6
 
 // compactionTriggerRatio is the fraction of the context window at which
-// proactive compaction kicks in (top of each turn).
-const compactionTriggerRatio = 0.8
+// proactive compaction kicks in (top of each turn). NOTE: this is speculative
+// tuning — validate against the per-turn latency/compaction metrics (roadmap 5.2)
+// before trusting it. Measured fixed overhead is ~7.4k tokens (not the ~16k the
+// roadmap assumed), so context fills slower; if data shows 0.7 triggers redundant
+// summarization, raise it back toward 0.8.
+const compactionTriggerRatio = 0.7
 
 // summaryLabel prefixes the injected summary so it is unmistakable in the
 // transcript (and so tests can assert on it).
@@ -280,6 +285,25 @@ func isContextLimitError(message string) bool {
 	return false
 }
 
+// isStreamTimeoutError reports whether a streamed error is a provider stream
+// idle/stall timeout (providerio surfaces these as "idle timeout after …" /
+// "no output for …" / "stream stalled"). Such a turn produced nothing, so when
+// no content was forwarded yet it can be safely re-issued on a fresh connection
+// (see the stall-retry in the agent loop). Substring + case-insensitive to
+// tolerate the prefixed "provider stream error: …" wrapping.
+func isStreamTimeoutError(message string) bool {
+	lowered := strings.ToLower(strings.TrimSpace(message))
+	if lowered == "" {
+		return false
+	}
+	for _, needle := range []string{"idle timeout after", "no output for", "stream stalled"} {
+		if strings.Contains(lowered, needle) {
+			return true
+		}
+	}
+	return false
+}
+
 // compactionState carries the per-run state the agent loop needs to compact a
 // conversation safely. It is created once per Run and is a no-op whenever
 // options.ContextWindow <= 0.
@@ -301,6 +325,43 @@ type compactionState struct {
 	// OnText is deliberately NOT forwarded (compaction stays invisible to the user),
 	// but its token COST must still be counted so usage reports and budgets include it.
 	onUsage func(Usage)
+
+	// calibrationRatio scales the raw byte/4 token estimate toward the provider's
+	// real prompt-token count. ApproxTextTokens over-counts code-heavy content by
+	// ~15-20%, which would trip compaction early (at ~60% of true capacity). It
+	// starts at 1.0 and converges via an EMA as each turn reports actual usage, so
+	// later turns compact nearer to real capacity. Zero is treated as 1.0.
+	calibrationRatio float64
+}
+
+// calibrate folds one turn's (rawEstimate, actualPromptTokens) sample into the
+// running calibration ratio. A single sample is clamped to a sane band so an
+// outlier (a huge cache-read turn, a provider-overhead spike) can't skew the
+// estimate enough to disable or thrash compaction.
+func (state *compactionState) calibrate(rawEstimate int, actualPromptTokens int) {
+	if !state.enabled || rawEstimate <= 0 || actualPromptTokens <= 0 {
+		return
+	}
+	sample := float64(actualPromptTokens) / float64(rawEstimate)
+	if sample < 0.5 {
+		sample = 0.5
+	} else if sample > 2.0 {
+		sample = 2.0
+	}
+	if state.calibrationRatio <= 0 {
+		state.calibrationRatio = 1.0
+	}
+	const alpha = 0.3 // weight on the newest sample; smooths jitter across turns
+	state.calibrationRatio = state.calibrationRatio*(1-alpha) + sample*alpha
+}
+
+// calibratedTokens applies the learned ratio to a raw estimate. Before any sample
+// arrives (ratio unset) it returns the raw estimate unchanged.
+func (state *compactionState) calibratedTokens(raw int) int {
+	if state.calibrationRatio <= 0 {
+		return raw
+	}
+	return int(float64(raw) * state.calibrationRatio)
 }
 
 func newCompactionState(options Options) *compactionState {
@@ -329,7 +390,7 @@ func (state *compactionState) maybeCompact(
 	// the messages; both the threshold check and the shrink check below use the
 	// same term so they stay consistent.
 	toolTokens := estimateToolDefTokens(tools)
-	size := estimateTokens(messages) + toolTokens
+	size := state.calibratedTokens(estimateTokens(messages) + toolTokens)
 	if size <= state.threshold {
 		return messages
 	}
@@ -346,7 +407,7 @@ func (state *compactionState) maybeCompact(
 	// entirely and preserve recent turns verbatim.
 	if pruned, reclaimed := pruneStaleToolOutput(messages, state.preserveLast); reclaimed > 0 {
 		messages = pruned
-		size = estimateTokens(messages) + toolTokens
+		size = state.calibratedTokens(estimateTokens(messages) + toolTokens)
 		if size <= state.threshold {
 			state.lowWaterMark = size
 			return messages
@@ -362,7 +423,7 @@ func (state *compactionState) maybeCompact(
 		// later turn) can try again; we never drop messages on failure here.
 		return messages
 	}
-	newSize := estimateTokens(compacted) + toolTokens
+	newSize := state.calibratedTokens(estimateTokens(compacted) + toolTokens)
 	if newSize >= size {
 		// Compaction did not actually shrink anything (e.g. nothing to
 		// summarize). Leave the history untouched and don't churn next turn.
@@ -420,7 +481,7 @@ func (state *compactionState) recover(
 	// the SAME combined (messages + tool-defs) domain maybeCompact uses, so the
 	// proactive shrink-guard compares like with like.
 	state.reactiveAttempted = true
-	state.lowWaterMark = estimateTokens(result) + estimateToolDefTokens(tools)
+	state.lowWaterMark = state.calibratedTokens(estimateTokens(result) + estimateToolDefTokens(tools))
 	return result, true, nil
 }
 

internal/agent/compaction_calibrate_test.go

@@ -0,0 +1,50 @@
+package agent
+
+import "testing"
+
+func TestCalibrationConvergesTowardActual(t *testing.T) {
+	state := &compactionState{enabled: true}
+
+	// Before any sample, the estimate is unchanged.
+	if got := state.calibratedTokens(1000); got != 1000 {
+		t.Fatalf("uncalibrated tokens = %d, want 1000", got)
+	}
+
+	// The estimator over-counts: our raw estimate is 1000 but the provider reports
+	// 850 (a 0.85 ratio). Feeding that sample repeatedly should pull the ratio
+	// toward 0.85, so a 1000-token raw estimate calibrates downward.
+	for range 20 {
+		state.calibrate(1000, 850)
+	}
+	got := state.calibratedTokens(1000)
+	if got >= 1000 || got < 820 || got > 880 {
+		t.Fatalf("calibrated tokens = %d, want ~850 after convergence", got)
+	}
+}
+
+func TestCalibrateIgnoresDegenerateSamples(t *testing.T) {
+	state := &compactionState{enabled: true}
+	state.calibrate(0, 500) // zero estimate
+	state.calibrate(500, 0) // zero actual
+	if state.calibrationRatio != 0 {
+		t.Fatalf("degenerate samples must not move the ratio, got %v", state.calibrationRatio)
+	}
+	// Disabled state never calibrates.
+	disabled := &compactionState{enabled: false}
+	disabled.calibrate(1000, 850)
+	if disabled.calibrationRatio != 0 {
+		t.Fatalf("disabled compaction must not calibrate, got %v", disabled.calibrationRatio)
+	}
+}
+
+func TestCalibrateClampsOutliers(t *testing.T) {
+	state := &compactionState{enabled: true}
+	// A wild outlier (10x) is clamped to 2.0, so one bad sample can't blow up the
+	// ratio and disable compaction.
+	for range 50 {
+		state.calibrate(1000, 10000)
+	}
+	if state.calibrationRatio > 2.0 {
+		t.Fatalf("ratio should be clamped at 2.0, got %v", state.calibrationRatio)
+	}
+}

internal/agent/compaction_preserve.go

@@ -20,6 +20,16 @@ const (
 	toolNameUpdatePlan = "update_plan"
 	toolNameToolSearch = "tool_search"
 	toolNameSkill      = "skill"
+	toolNameWriteFile  = "write_file"
+	toolNameEditFile   = "edit_file"
+)
+
+// maxRecentEdits caps how many edited files are carried across a compaction, and
+// maxEditNoteBytes caps each edit's one-line note, so a long editing run can't
+// bloat the preserved-state block.
+const (
+	maxRecentEdits   = 20
+	maxEditNoteBytes = 160
 )
 
 const (
@@ -100,12 +110,123 @@ func formatPlanArguments(arguments string) string {
 }
 
 // skillEntry is a named preserved body. It began as loaded-skill state and is
-// reused for loaded tools and project instruction blocks.
+// reused for loaded tools, project instruction blocks, and recent file edits.
 type skillEntry struct {
 	name string
 	body string
 }
 
+// recentEdits returns the files mutated by write_file/edit_file calls in messages
+// — latest note per path, in last-seen order — as skillEntry{name: path, body:
+// note}. After compaction elides the editing turns, this tells the model WHAT it
+// changed in each file (from the tool's result) so it need not re-read to
+// rediscover its own footprint. Capped at maxRecentEdits paths.
+//
+// Ordering is by LAST edit, not first: re-editing a file moves it to the newest
+// position so the tail cap below keeps the file the model most recently touched
+// rather than an earlier, now-stale entry.
+func recentEdits(messages []zeroruntime.Message) []skillEntry {
+	pathByID := map[string]string{}
+	sequence := make([]string, 0)
+	for _, message := range messages {
+		for _, call := range message.ToolCalls {
+			if call.Name != toolNameWriteFile && call.Name != toolNameEditFile {
+				continue
+			}
+			path := editPathFromArguments(call.Arguments)
+			if path == "" {
+				continue
+			}
+			if call.ID != "" {
+				pathByID[call.ID] = path
+			}
+			sequence = append(sequence, path)
+		}
+	}
+	order := lastSeenOrder(sequence)
+	if len(order) == 0 {
+		return nil
+	}
+
+	noteByPath := map[string]string{}
+	for _, message := range messages {
+		if message.Role != zeroruntime.MessageRoleTool || message.ToolCallID == "" {
+			continue
+		}
+		if path, ok := pathByID[message.ToolCallID]; ok {
+			noteByPath[path] = editNote(message.Content)
+		}
+	}
+
+	// Keep the most recent maxRecentEdits paths (the tail of last-seen order).
+	if len(order) > maxRecentEdits {
+		order = order[len(order)-maxRecentEdits:]
+	}
+	entries := make([]skillEntry, 0, len(order))
+	for _, path := range order {
+		entries = append(entries, skillEntry{name: path, body: noteByPath[path]})
+	}
+	return entries
+}
+
+// lastSeenOrder dedupes paths keeping each at its LAST occurrence, preserving
+// chronological order otherwise. A re-edited path therefore lands at the newest
+// position rather than staying pinned to where it first appeared.
+func lastSeenOrder(paths []string) []string {
+	lastIdx := make(map[string]int, len(paths))
+	for i, p := range paths {
+		lastIdx[p] = i
+	}
+	order := make([]string, 0, len(lastIdx))
+	for i, p := range paths {
+		if lastIdx[p] == i {
+			order = append(order, p)
+		}
+	}
+	return order
+}
+
+// editPathFromArguments pulls the target file path from a write_file/edit_file
+// call's JSON arguments (path/file/file_path/filename aliases). Returns "" on
+// malformed arguments or no path.
+func editPathFromArguments(arguments string) string {
+	var parsed struct {
+		Path     string `json:"path"`
+		File     string `json:"file"`
+		FilePath string `json:"file_path"`
+		Filename string `json:"filename"`
+	}
+	if err := json.Unmarshal([]byte(strings.TrimSpace(arguments)), &parsed); err != nil {
+		return ""
+	}
+	for _, candidate := range []string{parsed.Path, parsed.File, parsed.FilePath, parsed.Filename} {
+		if s := strings.TrimSpace(candidate); s != "" {
+			return s
+		}
+	}
+	return ""
+}
+
+// editNote reduces an edit tool's result to a single short line (its first
+// non-empty line, byte-capped on a rune boundary) for the preserved summary.
+func editNote(content string) string {
+	for _, line := range strings.Split(content, "\n") {
+		s := strings.TrimSpace(line)
+		if s == "" {
+			continue
+		}
+		if len(s) > maxEditNoteBytes {
+			limit := maxEditNoteBytes
+			for limit > 0 && !utf8.RuneStart(s[limit]) {
+				limit--
+			}
+			s = s[:limit] + "…"
+		}
+		return s
+	}
+	return ""
+}
+
 // loadedSkills returns the skills loaded via the skill tool in messages — the
 // latest body per name, in first-seen order — matching each skill tool call to
 // its tool result by id.
@@ -310,11 +431,17 @@ func capBody(body string) string {
 // preservedState is the JSON shape of the carried-across-compaction block.
 type preservedState struct {
 	Plan                string                 `json:"plan,omitempty"`
+	RecentEdits         []preservedEdit        `json:"recent_edits,omitempty"`
 	Tools               []preservedTool        `json:"tools,omitempty"`
 	Skills              []preservedSkill       `json:"skills,omitempty"`
 	ProjectInstructions []preservedInstruction `json:"project_instructions,omitempty"`
 }
 
+type preservedEdit struct {
+	Path string `json:"path"`
+	Note string `json:"note,omitempty"`
+}
+
 type preservedTool struct {
 	Name string `json:"name"`
 	Body string `json:"body"`
@@ -347,6 +474,12 @@ func appendPreservedState(summary string, middle []zeroruntime.Message) string {
 		plan = priorState.Plan
 	}
 
+	// Recent edits: merge edits preserved earlier with fresh write_file/edit_file
+	// results in middle (newer note per path wins AND moves to the newest
+	// position), so the tail cap keeps the files the model most recently touched
+	// across repeated compactions rather than dropping a just-re-edited file.
+	edits := capRecentEdits(mergeRecentEdits(preservedEditsToEntries(priorState.RecentEdits), recentEdits(middle)))
+
 	// Tools: preserve deferred tool_search schemas from the transcript. Fresh
 	// loads override older carried copies by name.
 	tools := mergeSkillEntries(preservedToolsToEntries(priorState.Tools), loadedToolSchemas(middle))
@@ -360,19 +493,76 @@ func appendPreservedState(summary string, middle []zeroruntime.Message) string {
 		projectInstructionEntries(middle),
 	)
 
-	if block := formatPreservedState(plan, tools, skills, instructions); block != "" {
+	if block := formatPreservedState(plan, edits, tools, skills, instructions); block != "" {
 		summary += "\n\n" + block
 	}
 	return summary
 }
 
+// mergeRecentEdits overlays fresh edits onto edits preserved by an earlier
+// compaction. Unlike mergeSkillEntries (which keeps refreshed entries in their
+// original slot), a path touched again by a fresh edit MOVES to the newest
+// position and takes the fresh note, so capRecentEdits — which keeps only the
+// most-recent tail — never drops a file the model just re-edited. Paths present
+// only in the older set keep their relative order, ahead of the fresh ones.
+func mergeRecentEdits(older, newer []skillEntry) []skillEntry {
+	if len(newer) == 0 {
+		return older
+	}
+	freshBody := make(map[string]string, len(newer))
+	freshOrder := make([]string, 0, len(newer))
+	for _, e := range newer {
+		if _, ok := freshBody[e.name]; !ok {
+			freshOrder = append(freshOrder, e.name)
+		}
+		freshBody[e.name] = e.body
+	}
+	merged := make([]skillEntry, 0, len(older)+len(freshOrder))
+	// Older entries that were NOT re-edited keep their position, oldest first.
+	for _, e := range older {
+		if _, refreshed := freshBody[e.name]; refreshed {
+			continue // re-appended at its newest position below
+		}
+		merged = append(merged, e)
+	}
+	// Fresh edits follow, in last-seen order, each carrying the fresh note — so
+	// the most recently touched files sit at the tail the cap preserves.
+	for _, name := range freshOrder {
+		merged = append(merged, skillEntry{name: name, body: freshBody[name]})
+	}
+	return merged
+}
+
+// capRecentEdits bounds the preserved edit list to the most recent maxRecentEdits
+// entries after a merge, so repeated compactions can't grow it without limit.
+func capRecentEdits(entries []skillEntry) []skillEntry {
+	if len(entries) > maxRecentEdits {
+		return entries[len(entries)-maxRecentEdits:]
+	}
+	return entries
+}
+
+func preservedEditsToEntries(edits []preservedEdit) []skillEntry {
+	entries := make([]skillEntry, 0, len(edits))
+	for _, e := range edits {
+		if e.Path == "" {
+			continue
+		}
+		entries = append(entries, skillEntry{name: e.Path, body: e.Note})
+	}
+	return entries
+}
+
 // formatPreservedState renders state as the labelled, single-line
 // JSON block. Returns "" when there is nothing to preserve.
-func formatPreservedState(plan string, tools, skills, instructions []skillEntry) string {
-	if plan == "" && len(tools) == 0 && len(skills) == 0 && len(instructions) == 0 {
+func formatPreservedState(plan string, edits, tools, skills, instructions []skillEntry) string {
+	if plan == "" && len(edits) == 0 && len(tools) == 0 && len(skills) == 0 && len(instructions) == 0 {
 		return ""
 	}
 	state := preservedState{Plan: plan}
+	for _, e := range edits {
+		state.RecentEdits = append(state.RecentEdits, preservedEdit{Path: e.name, Note: e.body})
+	}
 	for _, t := range tools {
 		state.Tools = append(state.Tools, preservedTool{Name: t.name, Body: t.body})
 	}

internal/agent/compaction_recent_edits_test.go

@@ -0,0 +1,121 @@
+package agent
+
+import (
+	"fmt"
+	"strings"
+	"testing"
+
+	"github.com/Gitlawb/zero/internal/zeroruntime"
+)
+
+// recentEdits extracts each mutated file's path and a one-line note from the
+// matching tool result, latest note per path in last-seen order.
+func TestRecentEditsExtractsPathsAndNotes(t *testing.T) {
+	messages := []zeroruntime.Message{
+		{Role: zeroruntime.MessageRoleAssistant, ToolCalls: []zeroruntime.ToolCall{
+			{ID: "e1", Name: "write_file", Arguments: `{"path":"internal/foo.go","content":"package foo"}`},
+		}},
+		{Role: zeroruntime.MessageRoleTool, ToolCallID: "e1", Content: "Wrote internal/foo.go (12 lines)"},
+		{Role: zeroruntime.MessageRoleAssistant, ToolCalls: []zeroruntime.ToolCall{
+			{ID: "e2", Name: "edit_file", Arguments: `{"path":"internal/bar.go","old_string":"a","new_string":"b"}`},
+		}},
+		{Role: zeroruntime.MessageRoleTool, ToolCallID: "e2", Content: "Applied edit to internal/bar.go"},
+	}
+
+	edits := recentEdits(messages)
+	if len(edits) != 2 {
+		t.Fatalf("expected 2 edited files, got %d: %#v", len(edits), edits)
+	}
+	if edits[0].name != "internal/foo.go" || !strings.Contains(edits[0].body, "12 lines") {
+		t.Fatalf("first edit = %#v, want foo.go with its note", edits[0])
+	}
+	if edits[1].name != "internal/bar.go" || !strings.Contains(edits[1].body, "Applied edit") {
+		t.Fatalf("second edit = %#v, want bar.go with its note", edits[1])
+	}
+}
+
+// After compaction elides the editing turns, the preserved-state block still
+// names the edited files and what changed, so the model needn't re-read them.
+func TestCompactionPreservesRecentEdits(t *testing.T) {
+	messages := []zeroruntime.Message{
+		{Role: zeroruntime.MessageRoleSystem, Content: "system"},
+		{Role: zeroruntime.MessageRoleUser, Content: "add a flag"},
+		{Role: zeroruntime.MessageRoleAssistant, Content: "editing", ToolCalls: []zeroruntime.ToolCall{
+			{ID: "e1", Name: "write_file", Arguments: `{"path":"cmd/main.go","content":"..."}`},
+		}},
+		{Role: zeroruntime.MessageRoleTool, ToolCallID: "e1", Content: "Wrote cmd/main.go (adds --version flag)"},
+		{Role: zeroruntime.MessageRoleAssistant, Content: "done"},
+		{Role: zeroruntime.MessageRoleUser, Content: "continue"},
+		{Role: zeroruntime.MessageRoleAssistant, Content: "continuing"},
+	}
+	summary := compactStateConversation(t, messages)
+
+	state := parsePreservedStateBlock(summary)
+	if len(state.RecentEdits) != 1 {
+		t.Fatalf("expected 1 preserved edit, got %#v", state.RecentEdits)
+	}
+	if state.RecentEdits[0].Path != "cmd/main.go" || !strings.Contains(state.RecentEdits[0].Note, "--version") {
+		t.Fatalf("preserved edit = %#v, want cmd/main.go + its note", state.RecentEdits[0])
+	}
+}
+
+// A fresh edit note for a path overrides the one carried from an earlier
+// compaction (newer wins) and moves the path to the newest position rather than
+// duplicating it or leaving it pinned to its old slot.
+func TestRecentEditsMergeNewerWins(t *testing.T) {
+	prior := preservedState{RecentEdits: []preservedEdit{{Path: "a.go", Note: "old note"}, {Path: "z.go", Note: "z"}}}
+	older := preservedEditsToEntries(prior.RecentEdits)
+	newer := []skillEntry{{name: "a.go", body: "new note"}, {name: "b.go", body: "added"}}
+
+	merged := mergeRecentEdits(older, newer)
+	// z.go (untouched) keeps its lead; a.go is re-edited so it takes the new note
+	// and moves behind z.go, ahead of the brand-new b.go: [z.go, a.go, b.go].
+	if len(merged) != 3 {
+		t.Fatalf("expected z.go + refreshed a.go + b.go, got %#v", merged)
+	}
+	if merged[0].name != "z.go" {
+		t.Fatalf("untouched z.go should stay first, got %#v", merged)
+	}
+	if merged[1].name != "a.go" || merged[1].body != "new note" {
+		t.Fatalf("re-edited a.go should move to the newest edits with the new note, got %#v", merged[1])
+	}
+	if merged[2].name != "b.go" {
+		t.Fatalf("brand-new b.go should be last, got %#v", merged[2])
+	}
+}
+
+// Regression: once more than maxRecentEdits distinct files are tracked, re-editing
+// an early file must move it into the capped tail rather than leaving it to be
+// dropped — otherwise the next compaction omits exactly the file the model most
+// recently touched.
+func TestRecentEditsCapKeepsReeditedFile(t *testing.T) {
+	// Earlier compaction preserved exactly maxRecentEdits files, f0 the oldest.
+	older := make([]skillEntry, 0, maxRecentEdits)
+	for i := 0; i < maxRecentEdits; i++ {
+		older = append(older, skillEntry{name: fmt.Sprintf("f%d.go", i), body: "old"})
+	}
+	// Fresh window re-edits the OLDEST file (f0) and adds a brand-new file, so the
+	// merged list exceeds the cap and something must be dropped.
+	newer := []skillEntry{{name: "f0.go", body: "re-edited"}, {name: "fresh.go", body: "new"}}
+
+	merged := capRecentEdits(mergeRecentEdits(older, newer))
+	if len(merged) != maxRecentEdits {
+		t.Fatalf("cap should keep %d entries, got %d", maxRecentEdits, len(merged))
+	}
+
+	var f0 *skillEntry
+	for i := range merged {
+		if merged[i].name == "f0.go" {
+			f0 = &merged[i]
+		}
+		if merged[i].name == "f1.go" {
+			t.Fatalf("f1.go (now the least-recently touched) should have been dropped, got %#v", merged)
+		}
+	}
+	if f0 == nil {
+		t.Fatalf("re-edited f0.go must survive the cap, got %#v", merged)
+	}
+	if f0.body != "re-edited" {
+		t.Fatalf("surviving f0.go should carry the fresh note, got %q", f0.body)
+	}
+}

internal/agent/decode_args_test.go

@@ -0,0 +1,99 @@
+package agent
+
+import "testing"
+
+// decodeToolArguments must (a) decode a normal single-object payload unchanged,
+// (b) tolerate a weak model packing multiple concatenated top-level JSON objects
+// into one arguments string by decoding the FIRST object and ignoring the rest
+// (the minimax-m3 "invalid character '{' after top-level value" failure), and
+// (c) still reject a genuinely malformed/truncated first object.
+func TestDecodeToolArguments(t *testing.T) {
+	t.Run("single object decodes normally", func(t *testing.T) {
+		var a map[string]any
+		if err := decodeToolArguments(`{"path":"x"}`, &a); err != nil {
+			t.Fatalf("err = %v, want nil", err)
+		}
+		if a["path"] != "x" {
+			t.Fatalf("decoded %v, want path=x", a)
+		}
+	})
+
+	t.Run("concatenated objects decode the first, ignore trailing", func(t *testing.T) {
+		var b map[string]any
+		if err := decodeToolArguments(`{"path":"first"}{"path":"second"}`, &b); err != nil {
+			t.Fatalf("multi-object should decode the first, got err = %v", err)
+		}
+		if b["path"] != "first" {
+			t.Fatalf("decoded %v, want the first object (path=first)", b)
+		}
+	})
+
+	t.Run("distinct tools' args concatenated (real minimax case): first object used", func(t *testing.T) {
+		var c map[string]any
+		if err := decodeToolArguments(`{"plan":[1,2]}{"cmd":"go version"}`, &c); err != nil {
+			t.Fatalf("err = %v, want nil", err)
+		}
+		if _, ok := c["plan"]; !ok || c["cmd"] != nil {
+			t.Fatalf("decoded %v, want only the first object's keys", c)
+		}
+	})
+
+	t.Run("truncated first object still errors", func(t *testing.T) {
+		var d map[string]any
+		if err := decodeToolArguments(`{"path":`, &d); err == nil {
+			t.Fatal("truncated JSON must still error")
+		}
+	})
+
+	t.Run("non-JSON trailing garbage still errors (not masked)", func(t *testing.T) {
+		var g map[string]any
+		if err := decodeToolArguments(`{"x":1}xyz`, &g); err == nil {
+			t.Fatal("a valid first object followed by non-JSON garbage must still error")
+		}
+	})
+
+	t.Run("recoverableToolArguments: only whole-JSON trailing is tolerated", func(t *testing.T) {
+		cases := []struct {
+			in    string
+			ok    bool
+			first string
+		}{
+			{`{"a":1}`, true, `{"a":1}`},
+			{`{"a":1}{"b":2}`, true, `{"a":1}`},
+			{`{"a":1}  {"b":2}  `, true, `{"a":1}`},
+			{`{"a":1}xyz`, false, ""},
+			{`{"a":`, false, ""},
+			{`   `, false, ""},
+		}
+		for _, c := range cases {
+			first, ok := recoverableToolArguments(c.in)
+			if ok != c.ok || (ok && first != c.first) {
+				t.Fatalf("recoverableToolArguments(%q) = (%q,%v), want (%q,%v)", c.in, first, ok, c.first, c.ok)
+			}
+		}
+	})
+
+	t.Run("historySafeToolCalls recovers the first object, not {}", func(t *testing.T) {
+		out := historySafeToolCalls([]ToolCall{
+			{ID: "c1", Name: "read_file", Arguments: `{"path":"README.md"}{"path":"AGENTS.md"}`}, // concatenated -> first
+			{ID: "c2", Name: "read_file", Arguments: `{"path":"x"}xyz`},                          // garbage -> {}
+			{ID: "c3", Name: "read_file", Arguments: `{"path":"ok"}`},                            // valid -> unchanged
+		})
+		if out[0].Arguments != `{"path":"README.md"}` {
+			t.Fatalf("concatenated args should record the first object, got %q", out[0].Arguments)
+		}
+		if out[1].Arguments != "{}" {
+			t.Fatalf("garbage-trailing args should record {}, got %q", out[1].Arguments)
+		}
+		if out[2].Arguments != `{"path":"ok"}` {
+			t.Fatalf("valid args should be unchanged, got %q", out[2].Arguments)
+		}
+	})
+
+	t.Run("whitespace-only errors (caller guards empty)", func(t *testing.T) {
+		var e map[string]any
+		if err := decodeToolArguments("   ", &e); err == nil {
+			t.Fatal("whitespace-only must error")
+		}
+	})
+}

internal/agent/file_diagnostics.go

@@ -0,0 +1,62 @@
+package agent
+
+import (
+	"context"
+	"os"
+	"path/filepath"
+	"strings"
+	"time"
+
+	"github.com/Gitlawb/zero/internal/lsp"
+)
+
+// fileDiagnosticsTimeout bounds one inline post-edit diagnostics check so a
+// slow or wedged language server can never hang a tool call; on timeout the
+// edit simply reports without a diagnostics block.
+const fileDiagnosticsTimeout = 10 * time.Second
+
+// NewFileDiagnostics adapts an *lsp.Manager to the per-edit inline diagnostics
+// callback (tools.RunOptions.Diagnostics): it reads the just-written file,
+// checks it against the file's language server, and formats error-severity
+// diagnostics for the model. Warnings and hints are excluded — nagging about
+// style on every edit is noise, while a type error the edit just introduced is
+// exactly what the model should see before its next step. Diagnostics are
+// rendered with workspace-relative paths: the absolute path would put the
+// local username/home directory into the model prompt and session transcript
+// on every edit. Returns nil when manager is nil, disabling inline diagnostics
+// entirely.
+func NewFileDiagnostics(manager *lsp.Manager, workspaceRoot string) func(context.Context, string) string {
+	if manager == nil {
+		return nil
+	}
+	return func(ctx context.Context, absPath string) string {
+		text, err := os.ReadFile(absPath)
+		if err != nil {
+			return ""
+		}
+		checkCtx, cancel := context.WithTimeout(ctx, fileDiagnosticsTimeout)
+		defer cancel()
+		diagnostics, err := manager.Check(checkCtx, absPath, string(text))
+		if err != nil {
+			return ""
+		}
+		errors := lsp.FilterBySeverity(diagnostics, lsp.SeverityError)
+		if len(errors) == 0 {
+			return ""
+		}
+		return lsp.FormatDiagnostics(diagnosticsDisplayPath(workspaceRoot, absPath), errors)
+	}
+}
+
+// diagnosticsDisplayPath renders absPath relative to the workspace root for
+// model-facing output, falling back to the file's base name when the path is
+// outside the workspace (a bare name still identifies the file without
+// exposing the directory layout).
+func diagnosticsDisplayPath(workspaceRoot, absPath string) string {
+	if workspaceRoot != "" {
+		if rel, err := filepath.Rel(workspaceRoot, absPath); err == nil && !strings.HasPrefix(rel, "..") {
+			return rel
+		}
+	}
+	return filepath.Base(absPath)
+}

internal/agent/file_diagnostics_test.go

@@ -0,0 +1,24 @@
+package agent
+
+import (
+	"path/filepath"
+	"testing"
+)
+
+// Diagnostics are model-facing: absolute paths would leak the local username
+// and directory layout into the prompt and session transcript on every edit.
+func TestDiagnosticsDisplayPath(t *testing.T) {
+	root := filepath.Join("/Users", "someone", "project")
+	cases := []struct {
+		root, abs, want string
+	}{
+		{root, filepath.Join(root, "internal", "a.go"), filepath.Join("internal", "a.go")},
+		{root, filepath.Join("/etc", "other.go"), "other.go"}, // outside root -> base name only
+		{"", filepath.Join("/home", "user", "x.go"), "x.go"},  // no root -> base name only
+	}
+	for _, c := range cases {
+		if got := diagnosticsDisplayPath(c.root, c.abs); got != c.want {
+			t.Errorf("diagnosticsDisplayPath(%q, %q) = %q, want %q", c.root, c.abs, got, c.want)
+		}
+	}
+}

internal/agent/guardrails.go

@@ -22,6 +22,12 @@ const (
 	// calls have executed since the last update_plan call.
 	staleToolCallThreshold = 10
 
+	// stalePlanTurnThreshold injects the same one-shot reminder once this many
+	// turns have passed since the last update_plan while plan items are still
+	// pending — the turn-based complement to staleToolCallThreshold, catching a
+	// plan that drifts stale across many turns that each make few tool calls.
+	stalePlanTurnThreshold = 8
+
 	// toolOnlyProgressReminderAt injects a one-shot progress nudge after this
 	// many consecutive turns contain tool calls but no visible assistant text.
 	// It does not stop the run; it tells the model to synthesize what it already
@@ -43,8 +49,12 @@ const (
 	toolFailureHintAt = 2
 	// toolFailureStopAt halts the run after a tool fails this many times in a row
 	// with the same error, so NO model (weak or strong) burns turns looping on a
-	// bad call.
-	toolFailureStopAt = 4
+	// bad call. Set to 6 (not 4): a corrective hint fires at toolFailureHintAt (2),
+	// and a model iterating on a genuinely tricky edit can legitimately fail a few
+	// times after the hint while converging — stopping at 4 cut those runs short.
+	// The streak still resets the moment the tool succeeds or hits a different
+	// error, so this only affects true same-error loops.
+	toolFailureStopAt = 6
 
 	// maxContinueNudges bounds how many times the headless completion gate
 	// (Options.RequireCompletionSignal) re-prompts a model that stopped without a
@@ -354,8 +364,13 @@ type guardState struct {
 	emptyTurns               int
 	totalToolCalls           int
 	toolCallsSincePlanUpdate int
-	planEverCalled           bool
-	notCalledReminderSent    bool
+	// turnsSincePlanUpdate counts turns (not individual tool calls) since the last
+	// update_plan, so a plan that goes stale across many low-tool-call turns is
+	// still caught — the tool-call counter alone can take many turns to trip when
+	// the model makes only one call per turn.
+	turnsSincePlanUpdate  int
+	planEverCalled        bool
+	notCalledReminderSent bool
 	// staleReminderSent records whether the stale reminder has already fired for
 	// the current stale interval. It is cleared when a plan update opens a new
 	// interval, making the reminder one-shot per interval rather than per turn.
@@ -414,8 +429,9 @@ func (state *guardState) observeToolResult(name string, failed bool, output stri
 func (state *guardState) observeTurn(collected zeroruntime.CollectedStream) (stop bool) {
 	hasToolCalls := len(collected.ToolCalls) > 0
 	hasVisibleText := strings.TrimSpace(collected.Text) != ""
+	hasReasoning := collected.HasReasoning || len(collected.ReasoningBlocks) > 0
 
-	if hasToolCalls || hasVisibleText {
+	if hasToolCalls || hasVisibleText || hasReasoning {
 		state.emptyTurns = 0
 	} else {
 		state.emptyTurns++
@@ -427,11 +443,16 @@ func (state *guardState) observeTurn(collected zeroruntime.CollectedStream) (sto
 		state.toolOnlyReminderSent = false
 	}
 
+	// One turn has passed; the plan-update below resets this to 0 when the model
+	// refreshes the plan this turn.
+	state.turnsSincePlanUpdate++
+
 	for _, call := range collected.ToolCalls {
 		state.totalToolCalls++
 		if call.Name == planToolName {
 			state.planEverCalled = true
 			state.toolCallsSincePlanUpdate = 0
+			state.turnsSincePlanUpdate = 0
 			// A fresh plan update opens a new stale interval.
 			state.staleReminderSent = false
 			// Record how many items remain so the completion gate knows whether
@@ -485,10 +506,12 @@ func (state *guardState) progressReminder() string {
 // number of turns completed so far).
 func (state *guardState) planReminder(turn int) string {
 	// STALE reminder takes priority: a long run without a plan update is the
-	// stronger signal. One-shot per stale interval.
-	if state.planEverCalled &&
-		!state.staleReminderSent &&
-		state.toolCallsSincePlanUpdate >= staleToolCallThreshold {
+	// stronger signal. Fires on either the tool-call streak OR a turn streak with
+	// pending items (so a plan drifting stale across many low-call turns is caught,
+	// while a fully-completed plan is left alone). One-shot per stale interval.
+	if state.planEverCalled && !state.staleReminderSent &&
+		(state.toolCallsSincePlanUpdate >= staleToolCallThreshold ||
+			(state.turnsSincePlanUpdate >= stalePlanTurnThreshold && state.planItemsPending > 0)) {
 		state.staleReminderSent = true
 		return planStaleReminder(state.toolCallsSincePlanUpdate)
 	}

internal/agent/guardrails_test.go

@@ -22,6 +22,14 @@ func textTurn(content string) []zeroruntime.StreamEvent {
 	}
 }
 
+// reasoningTurn produces live reasoning without visible assistant text.
+func reasoningTurn(content string) []zeroruntime.StreamEvent {
+	return []zeroruntime.StreamEvent{
+		{Type: zeroruntime.StreamEventReasoning, Content: content},
+		{Type: zeroruntime.StreamEventDone},
+	}
+}
+
 // toolTurn produces a turn that calls a named tool with the given args JSON.
 func toolTurn(callID string, toolName string, args string) []zeroruntime.StreamEvent {
 	return []zeroruntime.StreamEvent{
@@ -101,6 +109,31 @@ func TestRunResetsEmptyTurnCounterOnVisibleOutput(t *testing.T) {
 	}
 }
 
+func TestRunResetsEmptyTurnCounterOnReasoning(t *testing.T) {
+	provider := &mockProvider{
+		turns: [][]zeroruntime.StreamEvent{
+			reasoningTurn("thinking 1"),
+			reasoningTurn("thinking 2"),
+			reasoningTurn("thinking 3"),
+			textTurn("done"),
+		},
+	}
+
+	result, err := Run(context.Background(), "go", provider, Options{
+		Registry: tools.NewRegistry(),
+		MaxTurns: 12,
+	})
+	if err != nil {
+		t.Fatal(err)
+	}
+	if result.FinalAnswer != "done" {
+		t.Fatalf("expected reasoning-only turns to keep the run live until final answer, got %q", result.FinalAnswer)
+	}
+	if len(provider.requests) != 4 {
+		t.Fatalf("expected 4 turns, got %d", len(provider.requests))
+	}
+}
+
 func TestRunResetsEmptyTurnCounterOnToolCall(t *testing.T) {
 	root := t.TempDir()
 	writeAgentTestFile(t, root+"/notes.txt", "alpha")

internal/agent/loop.go

@@ -5,6 +5,7 @@ import (
 	"encoding/json"
 	"errors"
 	"fmt"
+	"io"
 	"sort"
 	"strings"
 	"sync"
@@ -20,6 +21,17 @@ import (
 const maxTurnsAnswer = "Agent reached maximum number of turns without a final answer."
 const maxTurnsFinalAnswerPrompt = "You have reached the tool-turn limit. Do not call tools. Give a concise final answer now: summarize what you completed, what you found, and any remaining blockers."
 
+// maxStreamStallRetries bounds how many times a turn that timed out (idle/stall)
+// WITH NO OUTPUT yet is re-issued on a fresh connection before giving up. Only
+// the no-output case is retried (a partial turn would duplicate), so this is a
+// safe recovery for a stalled/dead pooled connection.
+//
+// Set to 1 (not 2): each attempt can itself idle for the full stream timeout
+// (~5min) before the stall is even detected, so 2 retries left an interactive
+// session frozen for ~15min. One retry keeps the common single-hiccup recovery
+// while bounding the worst case to ~2× the idle timeout.
+const maxStreamStallRetries = 1
+
 const (
 	toolResultMetaControl       = "control"
 	toolResultControlSpecReview = "spec_review_required"
@@ -125,6 +137,12 @@ func Run(ctx context.Context, prompt string, provider Provider, options Options)
 	guards := newGuardState()
 	compactor := newCompactionState(options)
 
+	// Background post-edit diagnostics: files changed by mutating tools are
+	// checked off the tool-call critical path and any errors are appended as a
+	// nudge before the NEXT request (see async_diagnostics.go). nil when no
+	// FileDiagnostics callback is wired; every method no-ops on nil.
+	postEditDiagnostics := newAsyncDiagnostics(options.FileDiagnostics, options.Cwd)
+
 	// loaded tracks deferred-eligible tools the model has pulled via tool_search
 	// during THIS run. It is consulted by partitionTools each turn to expose a
 	// loaded tool's full schema; it lives only for the run (v1 within-run scope).
@@ -138,15 +156,31 @@ func Run(ctx context.Context, prompt string, provider Provider, options Options)
 	// has already been demanded this run, so it fires at most once.
 	acceptanceRequested := false
 
+	// toolDefCache memoizes each tool's rendered JSON-schema definition across
+	// turns (a tool's advertised schema is stable for the run), so partitionTools
+	// doesn't re-run the recursive schema→map conversion for every tool every turn.
+	toolDefCache := map[string]zeroruntime.ToolDefinition{}
+
 	result := Result{Messages: copyMessages(messages)}
 	for turn := 0; turn < maxTurns; turn++ {
 		result.Turns = turn + 1
 
+		// Deliver background post-edit diagnostics from the previous turn's edits
+		// BEFORE compaction so the nudge is part of the request being budgeted.
+		// A brief wait at most (asyncDiagnosticsDrainTimeout); an unfinished check
+		// simply delivers on a later turn.
+		if nudge := postEditDiagnostics.drain(ctx); nudge != "" {
+			messages = append(messages, zeroruntime.Message{
+				Role:    zeroruntime.MessageRoleUser,
+				Content: nudge,
+			})
+		}
+
 		// Build the per-turn tool list first so proactive compaction can include
 		// the tool-definition tokens (they ride on every request) in its estimate.
 		// partitionTools depends only on registry/permissions/options/loaded, not on
 		// the messages, so computing it before compaction is safe.
-		exposed, _ := partitionTools(registry, permissionMode, options, loaded)
+		exposed, _ := partitionToolsCached(registry, permissionMode, options, loaded, toolDefCache)
 
 		// PROACTIVE compaction: if the history is approaching the model's
 		// context window, summarize the oldest middle before building the
@@ -156,6 +190,7 @@ func Run(ctx context.Context, prompt string, provider Provider, options Options)
 			Messages:        copyMessages(messages),
 			Tools:           exposed,
 			ReasoningEffort: options.ReasoningEffort,
+			PromptCacheKey:  options.SessionID,
 		}
 
 		// Report the per-category context budget for this turn so a surface can
@@ -191,6 +226,7 @@ func Run(ctx context.Context, prompt string, provider Provider, options Options)
 					Messages:        copyMessages(messages),
 					Tools:           exposed,
 					ReasoningEffort: options.ReasoningEffort,
+					PromptCacheKey:  options.SessionID,
 				}
 				// Pre-content connect after a context-limit compaction: route through the
 				// reconnect helper so a transient upstream hiccup here doesn't fail the
@@ -203,55 +239,75 @@ func Run(ctx context.Context, prompt string, provider Provider, options Options)
 			}
 		}
 
-		collected := zeroruntime.CollectStreamWithOptions(ctx, stream, zeroruntime.CollectOptions{
-			OnText:          options.OnText,
-			OnReasoning:     options.OnReasoning,
-			OnUsage:         options.OnUsage,
-			OnToolCallStart: options.OnToolCallStart,
-			OnToolCallDelta: options.OnToolCallDelta,
-		})
-		if collected.Error != "" {
-			collectedErr := errors.New(collected.Error)
-			if isImageRejectionError(collectedErr) {
-				result.Messages = copyMessages(messages)
-				return result, fmt.Errorf("model %s rejected the image: %s. The model may not support image input — try switching to a vision-capable model (claude, gpt-4o, gemini)", options.Model, collected.Error)
+		// forwardedVisibleText flags forwarded final PROSE (OnText) specifically —
+		// the only output whose re-stream on a stall retry would visibly DUPLICATE
+		// answer text the user already read. Transient working indicators
+		// (reasoning previews, tool-call "writing…" previews) are deliberately NOT
+		// counted: a turn that streamed only those then stalled — the common
+		// gpt-5.x / ollama "froze mid-write_file" case — is safe to re-issue, since
+		// the TUI resets the tool-call preview on the next tool-call-start and a
+		// stalled turn is never appended to `messages` (it returns on the error
+		// before the append), so the retry re-sends clean context with no
+		// conversation-state duplication.
+		forwardedVisibleText := false
+		forwardingOpts := zeroruntime.CollectOptions{OnUsage: options.OnUsage}
+		if options.OnText != nil {
+			forwardingOpts.OnText = func(s string) { forwardedVisibleText = true; options.OnText(s) }
+		}
+		if options.OnReasoning != nil {
+			forwardingOpts.OnReasoning = func(s string) { options.OnReasoning(s) }
+		}
+		if options.OnToolCallStart != nil {
+			forwardingOpts.OnToolCallStart = func(id, name string) { options.OnToolCallStart(id, name) }
+		}
+		if options.OnToolCallDelta != nil {
+			forwardingOpts.OnToolCallDelta = func(id, fragment string) { options.OnToolCallDelta(id, fragment) }
+		}
+		// recoverStreamError applies the same non-stall recovery the initial stream
+		// gets to ANY collected error — including one from a reissued (stall-retry)
+		// stream: an image rejection gets the friendly wrapping, and a context limit
+		// gets one compaction + reactive reissue (omitting visible callbacks, since
+		// any pre-error output was already forwarded). It returns the possibly-updated
+		// collected and a non-nil stop error when the run must end now.
+		recoverStreamError := func(collected zeroruntime.CollectedStream) (zeroruntime.CollectedStream, error) {
+			if isImageRejectionError(errors.New(collected.Error)) {
+				return collected, fmt.Errorf("model %s rejected the image: %s. The model may not support image input — try switching to a vision-capable model (claude, gpt-4o, gemini)", options.Model, collected.Error)
 			}
-			// REACTIVE compaction: the streamed error may also be a context
-			// limit (some providers surface it mid-stream). Compact and retry
-			// the same turn once before giving up.
+			// REACTIVE compaction: the streamed error may also be a context limit
+			// (some providers surface it mid-stream). Compact and retry once.
 			if compacted, retried, retryErr := compactor.recover(ctx, provider, messages, request.Tools, collected.Error); retried {
 				messages = compacted
 				if retryErr != nil {
-					result.Messages = copyMessages(messages)
-					return result, retryErr
+					return collected, retryErr
 				}
-				// Reuse the SAME active-mode partition (exposed) from this
-				// turn rather than the bare toolDefinitions: exposed depends on
-				// registry+loaded (not the messages), so they stay valid after compaction.
-				// Routing through an empty-loaded partition here would re-hide every
-				// already-loaded deferred tool.
+				// Reuse the SAME active-mode partition (exposed) from this turn rather
+				// than the bare toolDefinitions: exposed depends on registry+loaded (not
+				// the messages), so it stays valid after compaction.
 				retryRequest := zeroruntime.CompletionRequest{
 					Messages:        copyMessages(messages),
 					Tools:           exposed,
 					ReasoningEffort: options.ReasoningEffort,
+					PromptCacheKey:  options.SessionID,
 				}
-				// Pre-content reconnect of a mid-stream disconnect: route the connect
-				// through the reconnect helper too (AUDIT-L1).
 				retryStream, retryStreamErr := streamWithReconnect(ctx, provider, retryRequest, reconnectNoticeFor(options))
 				if retryStreamErr != nil {
-					result.Messages = copyMessages(messages)
-					return result, retryStreamErr
+					return collected, retryStreamErr
 				}
-				// Omit OnText on the reactive retry: when the original error
-				// surfaced MID-stream, partial text was already forwarded to the
-				// user. Re-streaming the retried response on top of it would
-				// duplicate output. OnUsage IS kept so token telemetry/budgeting
-				// still counts the successful retry. The retried text is captured
-				// in collected.Text and becomes the turn's assistant message.
 				collected = zeroruntime.CollectStreamWithOptions(ctx, retryStream, zeroruntime.CollectOptions{
 					OnUsage: options.OnUsage,
 				})
 			}
+			return collected, nil
+		}
+
+		collected := zeroruntime.CollectStreamWithOptions(ctx, stream, forwardingOpts)
+		if collected.Error != "" {
+			updated, stop := recoverStreamError(collected)
+			collected = updated
+			if stop != nil {
+				result.Messages = copyMessages(messages)
+				return result, stop
+			}
 		}
 		// Check ctx first: on cancellation helpers.go sets collected.Error to
 		// ctx.Err().Error(), so returning errors.New(collected.Error) would lose
@@ -260,10 +316,70 @@ func Run(ctx context.Context, prompt string, provider Provider, options Options)
 			result.Messages = copyMessages(messages)
 			return result, ctx.Err()
 		}
-		if collected.Error != "" {
-			result.Messages = copyMessages(messages)
-			return result, errors.New(collected.Error)
+		// A stream idle/stall timeout is safely re-issued when the turn committed NO
+		// answer text — no forwarded visible prose (forwardedVisibleText) and no
+		// collected final text (collected.Text). This covers two cases:
+		//   1. Nothing streamed at all before the connection died (the original
+		//      macOS stale-pooled-connection hang past the response-header timeout).
+		//   2. The model streamed transient reasoning and began a tool call (e.g. a
+		//      large write_file) then froze mid-arguments — the common gpt-5.x /
+		//      ollama heartbeat-pause stall. That partial tool call is NEVER executed
+		//      (a turn with collected.Error returns before dispatch below) and NEVER
+		//      appended to `messages`, so a retry re-issues clean context; the only
+		//      re-render is transient previews, not duplicated answer text. This is
+		//      why the gate no longer excludes collected.ToolCalls: an incomplete
+		//      tool call from a timed-out stream is discard-and-retry, not output.
+		// A turn that forwarded real prose is NOT retried (it would duplicate visible
+		// answer text) and falls through to the error return below. Capped +
+		// exponential backoff, with a user-visible notice per attempt.
+		for attempt := 1; attempt <= maxStreamStallRetries &&
+			isStreamTimeoutError(collected.Error) && !forwardedVisibleText &&
+			collected.Text == ""; attempt++ {
+			if notify := stallRetryNoticeFor(options); notify != nil {
+				notify(attempt, maxStreamStallRetries)
+			}
+			if err := sleepWithContext(ctx, backoffFor(attempt)); err != nil {
+				result.Messages = copyMessages(messages)
+				return result, err
+			}
+			retryRequest := zeroruntime.CompletionRequest{
+				Messages:        copyMessages(messages),
+				Tools:           exposed,
+				ReasoningEffort: options.ReasoningEffort,
+				PromptCacheKey:  options.SessionID,
+			}
+			retryStream, retryErr := streamWithReconnect(ctx, provider, retryRequest, reconnectNoticeFor(options))
+			if retryErr != nil {
+				result.Messages = copyMessages(messages)
+				return result, retryErr
+			}
+			collected = zeroruntime.CollectStreamWithOptions(ctx, retryStream, forwardingOpts)
 		}
+		if collected.Error != "" {
+			// Route a reissued stream's non-stall error through the SAME recovery as
+			// the initial stream (image-rejection wrapping / context-limit compaction)
+			// rather than returning it raw.
+			updated, stop := recoverStreamError(collected)
+			collected = updated
+			if stop != nil {
+				result.Messages = copyMessages(messages)
+				return result, stop
+			}
+			if collected.Error != "" {
+				result.Messages = copyMessages(messages)
+				return result, errors.New(collected.Error)
+			}
+		}
+
+		// Calibrate the compaction token estimator against the provider's real
+		// prompt-token count for the request we just sent, so later turns trigger
+		// compaction near true capacity instead of ~15% early on code-heavy history.
+		// Recompute the estimate HERE (not at request-build time): a reactive
+		// compaction may have replaced `messages` with a smaller set and re-sent, so
+		// this reflects the request that actually produced collected.Usage. The
+		// assistant reply is appended below, after this, so `messages` is still the
+		// sent request.
+		compactor.calibrate(estimateTokens(messages)+estimateToolDefTokens(exposed), collected.Usage.InputTokens)
 
 		// Carry the turn's terminal stop reason so a final answer cut off at the
 		// output token cap (or by a content filter) is reported as truncated. A
@@ -375,6 +491,19 @@ func Run(ctx context.Context, prompt string, provider Provider, options Options)
 					continue
 				}
 			}
+			// Finalization diagnostics gate: edits from this run may still have
+			// checks in flight — the per-turn drain waits only briefly and defers
+			// to "a later turn", but a final answer means there is no later turn.
+			// Wait out the full inline-era budget once; an introduced error gives
+			// the model one more turn to see (and fix) it instead of being lost.
+			// Free for runs that never edited: an idle collector returns "".
+			if nudge := postEditDiagnostics.drainFinal(ctx); nudge != "" {
+				messages = append(messages, zeroruntime.Message{
+					Role:    zeroruntime.MessageRoleUser,
+					Content: nudge,
+				})
+				continue
+			}
 			result.FinalAnswer = collected.Text
 			result.Messages = copyMessages(messages)
 			return result, nil
@@ -400,11 +529,35 @@ func Run(ctx context.Context, prompt string, provider Provider, options Options)
 		// between tool_results breaks strict provider replay) — same after-batch
 		// rationale as turnRequestedModel above.
 		var changedFilesThisBatch []string
+		// Parallel read-ahead state: results for calls[precomputedStart:precomputedEnd]
+		// executed concurrently, consumed strictly in order below.
+		var precomputed []precomputedToolResult
+		precomputedStart, precomputedEnd := 0, 0
 		for index, call := range collected.ToolCalls {
+			// When this call starts a consecutive run of >= 2 auto-allowed read-only
+			// calls, execute the whole run concurrently now (see parallel_tools.go).
+			// The scan happens lazily at the run's first index — never ahead of a
+			// pending mutating call — so read-after-write ordering is preserved.
+			if index >= precomputedEnd {
+				runEnd := index
+				for runEnd < len(collected.ToolCalls) && parallelSafeToolCall(registry, collected.ToolCalls[runEnd], options) {
+					runEnd++
+				}
+				if runEnd-index >= 2 {
+					precomputed = executeParallelReadBatch(ctx, registry, collected.ToolCalls, index, runEnd, permissionMode, options)
+					precomputedStart, precomputedEnd = index, runEnd
+				}
+			}
 			if options.OnToolCall != nil {
 				options.OnToolCall(call)
 			}
-			toolResult, abortErr := executeToolCall(ctx, registry, call, permissionMode, options)
+			var toolResult ToolResult
+			var abortErr error
+			if index >= precomputedStart && index < precomputedEnd {
+				toolResult, abortErr = precomputed[index-precomputedStart].result, precomputed[index-precomputedStart].abortErr
+			} else {
+				toolResult, abortErr = executeToolCall(ctx, registry, call, permissionMode, options)
+			}
 			if options.OnToolResult != nil {
 				options.OnToolResult(toolResult)
 			}
@@ -467,11 +620,14 @@ func Run(ctx context.Context, prompt string, provider Provider, options Options)
 				failureHint = toolFailureHint(call.Name, toolSchemaJSON(registry, call.Name), toolResult.Output)
 			}
 
-			// Post-edit self-correction: collect the files this successful mutating
-			// tool changed; verification runs once over the union after the batch.
-			// A read-only tool (no ChangedFiles) never contributes.
-			if options.SelfCorrect != nil && toolResult.Status == tools.StatusOK && len(toolResult.ChangedFiles) > 0 {
+			// Collect the files this successful mutating tool changed. Self-correct
+			// verification runs once over the union after the batch; background
+			// diagnostics start NOW so the language server analyzes while the rest
+			// of the batch executes. A read-only tool (no ChangedFiles) never
+			// contributes.
+			if toolResult.Status == tools.StatusOK && len(toolResult.ChangedFiles) > 0 {
 				changedFilesThisBatch = append(changedFilesThisBatch, toolResult.ChangedFiles...)
+				postEditDiagnostics.enqueue(ctx, toolResult.ChangedFiles)
 			}
 		}
 
@@ -554,6 +710,17 @@ func Run(ctx context.Context, prompt string, provider Provider, options Options)
 	// mid-run, not a natural completion (a finished run returns via the no-tool-call
 	// success path before this). Under the headless completion gate that is INCOMPLETE,
 	// not success, so a run that loops to the turn limit isn't reported as done.
+	//
+	// The final-answer call is one more model request, so diagnostics from the
+	// LAST turn's edits get a drain too — with the finalization budget, since
+	// there is no later turn to defer to — otherwise an error introduced by
+	// the final edit would go unreported in the summary.
+	if nudge := postEditDiagnostics.drainFinal(ctx); nudge != "" {
+		messages = append(messages, zeroruntime.Message{
+			Role:    zeroruntime.MessageRoleUser,
+			Content: nudge,
+		})
+	}
 	if answer, finalMessages, finishReason := finalAnswerAfterMaxTurns(ctx, provider, messages, options); strings.TrimSpace(answer) != "" {
 		result.FinalAnswer = answer
 		result.FinishReason = finishReason
@@ -586,6 +753,7 @@ func finalAnswerAfterMaxTurns(ctx context.Context, provider Provider, messages [
 	stream, err := streamWithReconnect(ctx, provider, zeroruntime.CompletionRequest{
 		Messages:        copyMessages(finalMessages),
 		ReasoningEffort: options.ReasoningEffort,
+		PromptCacheKey:  options.SessionID,
 	}, reconnectNoticeFor(options))
 	if err != nil {
 		return "", messages, ""
@@ -615,8 +783,20 @@ func historySafeToolCalls(calls []ToolCall) []ToolCall {
 	for i, call := range calls {
 		safe[i] = call
 		args := strings.TrimSpace(call.Arguments)
-		if args == "" || !json.Valid([]byte(args)) {
+		switch {
+		case args == "":
 			safe[i].Arguments = "{}"
+		case json.Valid([]byte(args)):
+			// already a single valid JSON value — keep as-is
+		default:
+			// Recover the first object for the concatenated case so the replayed
+			// transcript matches what executeToolCall actually ran; otherwise (real
+			// corruption) fall back to an empty object.
+			if first, ok := recoverableToolArguments(args); ok {
+				safe[i].Arguments = first
+			} else {
+				safe[i].Arguments = "{}"
+			}
 		}
 	}
 	return safe
@@ -641,10 +821,53 @@ func toolSchemaJSON(registry *tools.Registry, name string) string {
 // is a non-nil ABORT error only when the call demands the whole run stop (a
 // canceled/timed-out ask_user prompt) rather than continuing — every ordinary
 // success or tool error returns a nil abort error.
+// recoverableToolArguments inspects a tool call's raw JSON arguments. It returns
+// the raw text of the FIRST JSON value when the payload is one value optionally
+// followed by additional WHOLE JSON values (the weak-model concatenated case, e.g.
+// `{A}{B}`). ok is false when the payload is genuinely malformed: a bad/truncated
+// first value, OR trailing NON-JSON garbage after a valid first value (e.g.
+// `{"x":1}xyz`) — that case must still error, not be silently accepted. Sharing
+// this between dispatch and replay-history keeps them consistent: only a cleanly
+// recoverable first object is ever used.
+func recoverableToolArguments(arguments string) (first string, ok bool) {
+	dec := json.NewDecoder(strings.NewReader(arguments))
+	var head json.RawMessage
+	if err := dec.Decode(&head); err != nil {
+		return "", false
+	}
+	// The remainder must be only whole JSON values (or nothing); any decode error
+	// other than EOF means trailing garbage — reject so corruption still surfaces.
+	for {
+		var rest json.RawMessage
+		if err := dec.Decode(&rest); err != nil {
+			if errors.Is(err, io.EOF) {
+				return strings.TrimSpace(string(head)), true
+			}
+			return "", false
+		}
+	}
+}
+
+// decodeToolArguments decodes a tool call's JSON arguments into v. It tolerates a
+// weaker model that packs MULTIPLE concatenated top-level JSON objects into one
+// arguments string (`{A}{B}`) — the cause of "invalid character '{' after top-level
+// value" failures with small models like minimax-m3 — by decoding the FIRST object
+// and ignoring the trailing WHOLE JSON values, so the primary intended call runs.
+// A genuinely malformed payload (bad/truncated first object, or non-JSON trailing
+// garbage) still returns the real parse error; a standard single-object payload
+// decodes exactly as before.
+func decodeToolArguments(arguments string, v any) error {
+	if first, ok := recoverableToolArguments(arguments); ok {
+		return json.Unmarshal([]byte(first), v)
+	}
+	// Not cleanly recoverable — surface the genuine parse error.
+	return json.Unmarshal([]byte(arguments), v)
+}
+
 func executeToolCall(ctx context.Context, registry *tools.Registry, call ToolCall, permissionMode PermissionMode, options Options) (ToolResult, error) {
 	args := map[string]any{}
 	if call.Arguments != "" {
-		if err := json.Unmarshal([]byte(call.Arguments), &args); err != nil {
+		if err := decodeToolArguments(call.Arguments, &args); err != nil {
 			return ToolResult{
 				ToolCallID: call.ID,
 				Name:       call.Name,
@@ -686,6 +909,25 @@ func executeToolCall(ctx context.Context, registry *tools.Registry, call ToolCal
 		}
 	}
 
+	// A shell command that sets sandbox_permissions: with_additional_permissions
+	// must carry a valid additional_permissions payload; inlineAdditionalPermissionsProfile
+	// is the single source of truth for that shape, and buildPermissionEvent (below)
+	// calls it again to render the prompt's scope text. Check it here, before any
+	// permission prompt is shown: a malformed payload can never be satisfied no
+	// matter what the user decides, so surfacing it as a plain tool error (which the
+	// model can see and retry) is both clearer and avoids presenting a normal-looking
+	// prompt whose "allow" path is guaranteed to fail with a confusing denial.
+	if toolFound && isShellCommandTool(call.Name) {
+		if _, _, err := inlineAdditionalPermissionsProfile(args, options.Cwd); err != nil {
+			return ToolResult{
+				ToolCallID: call.ID,
+				Name:       call.Name,
+				Status:     tools.StatusError,
+				Output:     "Error: " + err.Error(),
+			}, nil
+		}
+	}
+
 	// ask_user is intercepted in the loop (like permissions) so the question can
 	// be routed to an interactive front-end instead of blocking inside the tool.
 	// When no front-end is wired up it degrades to the tool's own graceful Run().
@@ -902,6 +1144,12 @@ func executeToolCall(ctx context.Context, registry *tools.Registry, call ToolCal
 		// Per-session file version tracker so write_file/edit_file refuse to clobber
 		// a file that changed on disk outside Zero since it was last read.
 		FileTracker: options.FileTracker,
+		// Post-edit diagnostics are NOT forwarded to the tools: they used to run
+		// synchronously inside edit_file/write_file and block every mutating tool
+		// call on the language server (≥300ms debounce floor, 10s cap). The loop
+		// now checks changed files in the background and nudges the model before
+		// its next request instead (see async_diagnostics.go). The tools' inline
+		// mechanism (tools.RunOptions.Diagnostics) stays for direct API callers.
 		// Forward the run's operator tool filters so a filter-aware tool
 		// (tool_search) never discloses or loads an operator-hidden deferred tool.
 		EnabledTools:  options.EnabledTools,
@@ -2029,6 +2277,18 @@ func buildPermissionEvent(call ToolCall, tool tools.Tool, args map[string]any, p
 		}
 	}
 
+	// A command that explicitly requests additional sandbox permissions
+	// (sandbox_permissions: with_additional_permissions) is an ELEVATION the user
+	// must consent to, even when the base command's sandbox decision was Allow.
+	// shouldRequestPermission blocks on OnPermissionRequest for exactly this case,
+	// so the event MUST be a prompt — otherwise it carried Action=allow while the
+	// loop waited on a decision, and a UI that renders only prompts (the TUI)
+	// dropped it, deadlocking the run. Keep this consistent with
+	// shouldRequestPermission's own condition.
+	if action == PermissionActionAllow && !permissionGranted && shellCommandAdditionalPermissionsRequested(args) {
+		action = PermissionActionPrompt
+	}
+
 	if safety.Permission == tools.PermissionAllow && action == PermissionActionAllow && !grantMatched && block == nil {
 		return PermissionEvent{}, false
 	}
@@ -2180,7 +2440,12 @@ func inlineAdditionalPermissionsProfile(args map[string]any, basePath string) (s
 	}
 	raw, ok := args["additional_permissions"]
 	if !ok || raw == nil {
-		return sandbox.RequestPermissionProfile{}, true, fmt.Errorf("missing additional_permissions; provide at least one of network or file_system")
+		return sandbox.RequestPermissionProfile{}, true, fmt.Errorf(
+			"sandbox_permissions was set to %q but no additional_permissions object was provided. "+
+				`Include one, for example additional_permissions: {"network": {"enabled": true}} or `+
+				`{"file_system": {"write": ["/path"]}}. If this command does not need elevated permissions, `+
+				"omit sandbox_permissions entirely and retry",
+			tools.SandboxPermissionsWithAdditionalPermissions)
 	}
 	data, err := json.Marshal(raw)
 	if err != nil {
@@ -2195,7 +2460,9 @@ func inlineAdditionalPermissionsProfile(args map[string]any, basePath string) (s
 		return sandbox.RequestPermissionProfile{}, true, err
 	}
 	if normalized.Empty() {
-		return sandbox.RequestPermissionProfile{}, true, fmt.Errorf("additional_permissions must include at least one requested permission in network or file_system")
+		return sandbox.RequestPermissionProfile{}, true, fmt.Errorf(
+			"additional_permissions must include at least one of network or file_system, for example " +
+				`{"network": {"enabled": true}} or {"file_system": {"write": ["/path"]}}`)
 	}
 	grantProfile, err := sandbox.RequestPermissionGrantProfile(normalized)
 	if err != nil {
@@ -2294,6 +2561,17 @@ func permissionActionFromSandbox(action sandbox.Action) PermissionAction {
 // exposed. The exposed slice is alpha-sorted by name, matching the legacy order
 // so the inactive path is stable.
 func partitionTools(registry *tools.Registry, permissionMode PermissionMode, options Options, loaded map[string]bool) ([]zeroruntime.ToolDefinition, string) {
+	return partitionToolsCached(registry, permissionMode, options, loaded, nil)
+}
+
+// partitionToolsCached is partitionTools with an optional per-tool definition
+// cache. The partitioning itself (visibility, deferral, ordering) is recomputed
+// every call — it must be, because a tool's deferred state can flip mid-run (e.g.
+// swarm tools un-defer once a swarm is active). Only the expensive part —
+// rendering each tool's JSON-schema parameters — is memoized by tool name, since a
+// tool's advertised name/description/schema is stable for the run. defCache nil
+// disables caching (used by tests and the plain partitionTools entrypoint).
+func partitionToolsCached(registry *tools.Registry, permissionMode PermissionMode, options Options, loaded map[string]bool, defCache map[string]zeroruntime.ToolDefinition) ([]zeroruntime.ToolDefinition, string) {
 	registeredTools := registry.All()
 
 	visible := make([]tools.Tool, 0, len(registeredTools))
@@ -2328,76 +2606,115 @@ func partitionTools(registry *tools.Registry, permissionMode PermissionMode, opt
 
 	active := options.DeferThreshold > 0 && eligible >= options.DeferThreshold && loaderUsable
 
-	definitions := make([]zeroruntime.ToolDefinition, 0, len(visible))
-	exposedNames := make(map[string]bool, len(visible))
+	// INACTIVE: every visible tool eager (except tool_search), alpha-sorted. With no
+	// deferral there is no mid-session loading, so this is byte-stable across turns
+	// and byte-identical to the pre-deferral output.
+	if !active {
+		definitions := make([]zeroruntime.ToolDefinition, 0, len(visible))
+		for _, tool := range visible {
+			if tool.Name() == tools.ToolSearchToolName {
+				continue
+			}
+			definitions = append(definitions, cachedRuntimeToolDefinition(defCache, tool))
+		}
+		sort.Slice(definitions, func(left int, right int) bool {
+			return definitions[left].Name < definitions[right].Name
+		})
+		return definitions, ""
+	}
+
+	// ACTIVE: lay the tools array out so its cacheable prefix does NOT shift when a
+	// deferred tool loads mid-session. The tools block is part of the provider's
+	// cached prefix; if a load reorders it, the cache invalidates from that point
+	// through the system prompt and messages — on EVERY load. Previously the whole
+	// array was alpha-sorted each turn, so a newly loaded tool was INSERTED into the
+	// middle and shifted every later definition. Instead we build three append-only
+	// regions:
+	//   1. non-deferred tools, alpha-sorted — always present, identical every turn;
+	//   2. tool_search — always present, right after the stable block;
+	//   3. loaded deferred tools, alpha-sorted — APPENDED, so an unloaded->loaded
+	//      transition grows the tail instead of inserting into the eager block.
+	// (tool_search's discovery text still shrinks as tools load, so keeping it and
+	// the loaded tail AFTER the eager block preserves the eager tools' cache across a
+	// load; fully stabilizing the loader's own description is a scoped follow-up.)
+	eager := make([]zeroruntime.ToolDefinition, 0, len(visible))
+	loadedTail := make([]zeroruntime.ToolDefinition, 0)
 	var hiddenTools []tools.Tool
 	for _, tool := range visible {
 		name := tool.Name()
-		deferred := tools.IsDeferred(tool)
-
-		if !active {
-			// Inactive: byte-identical to legacy, but tool_search is never advertised.
-			if name == tools.ToolSearchToolName {
-				continue
+		if name == tools.ToolSearchToolName {
+			continue // added explicitly between the eager block and the loaded tail
+		}
+		if tools.IsDeferred(tool) {
+			if loaded[name] {
+				loadedTail = append(loadedTail, cachedRuntimeToolDefinition(defCache, tool))
+			} else {
+				hiddenTools = append(hiddenTools, tool)
 			}
-			definitions = append(definitions, zeroruntime.ToolDefinition{
-				Name:        name,
-				Description: tool.Description(),
-				Parameters:  schemaToRuntimeMap(tool.Parameters()),
-			})
-			exposedNames[name] = true
 			continue
 		}
-
-		// Active path.
-		if deferred && !loaded[name] {
-			hiddenTools = append(hiddenTools, tool)
-			continue
-		}
-		definitions = append(definitions, zeroruntime.ToolDefinition{
-			Name:        name,
-			Description: tool.Description(),
-			Parameters:  schemaToRuntimeMap(tool.Parameters()),
-		})
-		exposedNames[name] = true
+		eager = append(eager, cachedRuntimeToolDefinition(defCache, tool))
 	}
-
-	// On the ACTIVE path tool_search is guaranteed runnable (active implies
-	// loaderUsable), so it must ALWAYS be reachable — even when a non-empty
-	// EnabledTools allowlist omits it (the operator allowlisted the deferred
-	// tools, not the loader). Expose its full definition whenever it is not
-	// already in the exposed set. This never runs on the inactive path, so the
-	// byte-identical below-threshold output is preserved.
-	discovery := ""
-	if active && len(hiddenTools) > 0 {
-		discovery = tools.BuildToolSearchDescription(hiddenTools)
-	}
-	if active && !exposedNames[tools.ToolSearchToolName] {
-		description := loader.Description()
-		if discovery != "" {
-			description = discovery
-		}
-		definitions = append(definitions, zeroruntime.ToolDefinition{
-			Name:        loader.Name(),
-			Description: description,
-			Parameters:  schemaToRuntimeMap(loader.Parameters()),
-		})
-	} else if active && discovery != "" {
-		for i := range definitions {
-			if definitions[i].Name == tools.ToolSearchToolName {
-				definitions[i].Description = discovery
-				break
-			}
-		}
-	}
-
-	sort.Slice(definitions, func(left int, right int) bool {
-		return definitions[left].Name < definitions[right].Name
+	sort.Slice(eager, func(left int, right int) bool {
+		return eager[left].Name < eager[right].Name
+	})
+	sort.Slice(loadedTail, func(left int, right int) bool {
+		return loadedTail[left].Name < loadedTail[right].Name
 	})
 
+	discovery := ""
+	if len(hiddenTools) > 0 {
+		discovery = tools.BuildToolSearchDescription(hiddenTools)
+	}
+
+	// tool_search is guaranteed runnable on the active path, so it is ALWAYS exposed
+	// — even when a non-empty EnabledTools allowlist omits it (the operator
+	// allowlisted the deferred tools, not the loader). It sits right after the stable
+	// eager block and carries the discovery list for still-hidden tools.
+	description := loader.Description()
+	if discovery != "" {
+		description = discovery
+	}
+	definitions := make([]zeroruntime.ToolDefinition, 0, len(eager)+1+len(loadedTail))
+	definitions = append(definitions, eager...)
+	definitions = append(definitions, zeroruntime.ToolDefinition{
+		Name:        loader.Name(),
+		Description: description,
+		Parameters:  schemaToRuntimeMap(loader.Parameters()),
+	})
+	definitions = append(definitions, loadedTail...)
+
 	return definitions, discovery
 }
 
+// cachedRuntimeToolDefinition returns the tool's rendered definition, reusing a
+// cached render when defCache holds one for this tool name. A tool's advertised
+// definition is stable across a run, so caching skips the recursive schema→map
+// conversion (schemaToRuntimeMap) that would otherwise run for every tool on every
+// turn. tool_search is excluded by its callers (its description is dynamic), so it
+// never poisons the cache. A nil cache computes fresh.
+func cachedRuntimeToolDefinition(defCache map[string]zeroruntime.ToolDefinition, tool tools.Tool) zeroruntime.ToolDefinition {
+	if defCache == nil {
+		return runtimeToolDefinition(tool)
+	}
+	if def, ok := defCache[tool.Name()]; ok {
+		return def
+	}
+	def := runtimeToolDefinition(tool)
+	defCache[tool.Name()] = def
+	return def
+}
+
+// runtimeToolDefinition renders a tool's advertised definition (name, description,
+// JSON-schema parameters) as sent to the provider.
+func runtimeToolDefinition(tool tools.Tool) zeroruntime.ToolDefinition {
+	return zeroruntime.ToolDefinition{
+		Name:        tool.Name(),
+		Description: tool.Description(),
+		Parameters:  schemaToRuntimeMap(tool.Parameters()),
+	}
+}
+
 func ToolVisible(tool tools.Tool, permissionMode PermissionMode, enabledTools []string, disabledTools []string) bool {
 	return ToolAllowedByFilters(tool.Name(), enabledTools, disabledTools) && ToolAdvertised(tool, permissionMode)
 }

internal/agent/loop_test.go

@@ -4,6 +4,7 @@ import (
 	"context"
 	"encoding/json"
 	"errors"
+	"fmt"
 	"os"
 	"path/filepath"
 	"runtime"
@@ -37,6 +38,35 @@ func (provider *mockProvider) StreamCompletion(ctx context.Context, request zero
 	return ch, nil
 }
 
+type recordingWebSearchTool struct {
+	calls []map[string]any
+}
+
+func (tool *recordingWebSearchTool) Name() string        { return "web_search" }
+func (tool *recordingWebSearchTool) Description() string { return "test web search tool" }
+func (tool *recordingWebSearchTool) Parameters() tools.Schema {
+	return tools.Schema{
+		Type: "object",
+		Properties: map[string]tools.PropertySchema{
+			"query": {Type: "string"},
+		},
+		Required:             []string{"query"},
+		AdditionalProperties: false,
+	}
+}
+func (tool *recordingWebSearchTool) Safety() tools.Safety {
+	return tools.Safety{
+		SideEffect:      tools.SideEffectNetwork,
+		Permission:      tools.PermissionPrompt,
+		Reason:          "Sends model-provided search query text to the configured web search backend.",
+		AdvertiseInAuto: true,
+	}
+}
+func (tool *recordingWebSearchTool) Run(_ context.Context, args map[string]any) tools.Result {
+	tool.calls = append(tool.calls, cloneArgs(args))
+	return tools.Result{Status: tools.StatusOK, Output: "1. T — https://x.test"}
+}
+
 type sandboxDeniedRetryTool struct {
 	calls []map[string]any
 }
@@ -865,7 +895,7 @@ func TestRunRejectsLocalWebFetchBeforePermissionRequest(t *testing.T) {
 	}
 }
 
-func TestRunAdvertisesAllowedWebSearchInAutoMode(t *testing.T) {
+func TestRunAdvertisesPromptedWebSearchInAutoMode(t *testing.T) {
 	t.Setenv("ZERO_WEBSEARCH_BASE_URL", "https://search.example/api")
 	registry := tools.NewRegistry()
 	for _, tool := range tools.CoreNetworkTools() {
@@ -896,6 +926,52 @@ func TestRunAdvertisesAllowedWebSearchInAutoMode(t *testing.T) {
 	}
 }
 
+func TestRunRequestsPermissionBeforeWebSearchExecution(t *testing.T) {
+	search := &recordingWebSearchTool{}
+	registry := tools.NewRegistry()
+	registry.Register(search)
+	provider := &mockProvider{
+		turns: [][]zeroruntime.StreamEvent{
+			{
+				{Type: zeroruntime.StreamEventToolCallStart, ToolCallID: "call-1", ToolName: "web_search"},
+				{Type: zeroruntime.StreamEventToolCallDelta, ToolCallID: "call-1", ArgumentsFragment: `{"query":"private workspace detail"}`},
+				{Type: zeroruntime.StreamEventToolCallEnd, ToolCallID: "call-1"},
+				{Type: zeroruntime.StreamEventDone},
+			},
+			{
+				{Type: zeroruntime.StreamEventText, Content: "done"},
+				{Type: zeroruntime.StreamEventDone},
+			},
+		},
+	}
+	var requests []PermissionRequest
+
+	result, err := Run(context.Background(), "search", provider, Options{
+		Registry:       registry,
+		PermissionMode: PermissionModeAsk,
+		OnPermissionRequest: func(_ context.Context, request PermissionRequest) (PermissionDecision, error) {
+			requests = append(requests, request)
+			return PermissionDecision{Action: PermissionDecisionDeny, Reason: "network not approved"}, nil
+		},
+	})
+
+	if err != nil {
+		t.Fatal(err)
+	}
+	if result.FinalAnswer != "done" {
+		t.Fatalf("expected final answer after denied tool call, got %q", result.FinalAnswer)
+	}
+	if len(requests) != 1 {
+		t.Fatalf("expected one permission request, got %#v", requests)
+	}
+	if requests[0].ToolName != "web_search" || requests[0].Permission != string(tools.PermissionPrompt) || requests[0].SideEffect != string(tools.SideEffectNetwork) {
+		t.Fatalf("unexpected permission request: %#v", requests[0])
+	}
+	if len(search.calls) != 0 {
+		t.Fatalf("web_search backend must not run when permission is denied, got calls %#v", search.calls)
+	}
+}
+
 func TestRunFiltersAdvertisedTools(t *testing.T) {
 	root := t.TempDir()
 	registry := tools.NewRegistry()
@@ -1053,7 +1129,10 @@ func TestRunSanitizesMalformedToolCallArgumentsBeforeRetry(t *testing.T) {
 		turns: [][]zeroruntime.StreamEvent{
 			{
 				{Type: zeroruntime.StreamEventToolCallStart, ToolCallID: "call-1", ToolName: "read_file"},
-				{Type: zeroruntime.StreamEventToolCallDelta, ToolCallID: "call-1", ArgumentsFragment: `{"path":"README.md"}{"path":"AGENTS.md"}`},
+				// Genuinely malformed (truncated) args: still error -> sanitize -> retry.
+				// (Concatenated multi-object args are now tolerated; see
+				// TestRunRecoversFirstObjectFromConcatenatedToolArgs.)
+				{Type: zeroruntime.StreamEventToolCallDelta, ToolCallID: "call-1", ArgumentsFragment: `{"path":"README.md"`},
 				{Type: zeroruntime.StreamEventToolCallEnd, ToolCallID: "call-1"},
 				{Type: zeroruntime.StreamEventDone},
 			},
@@ -1103,6 +1182,53 @@ func TestRunSanitizesMalformedToolCallArgumentsBeforeRetry(t *testing.T) {
 	}
 }
 
+// A weak model (e.g. minimax-m3) that packs MULTIPLE concatenated JSON objects into
+// one tool-call slot must RECOVER the first object and run the call, not fail with
+// "invalid character '{' after top-level value". The trailing object(s) are dropped
+// (the model re-issues them on a later turn if still needed).
+func TestRunRecoversFirstObjectFromConcatenatedToolArgs(t *testing.T) {
+	root := t.TempDir()
+	if err := os.WriteFile(filepath.Join(root, "README.md"), []byte("hello readme"), 0o644); err != nil {
+		t.Fatal(err)
+	}
+	registry := tools.NewRegistry()
+	registry.Register(tools.NewReadFileTool(root))
+	provider := &mockProvider{
+		turns: [][]zeroruntime.StreamEvent{
+			{
+				{Type: zeroruntime.StreamEventToolCallStart, ToolCallID: "call-1", ToolName: "read_file"},
+				{Type: zeroruntime.StreamEventToolCallDelta, ToolCallID: "call-1", ArgumentsFragment: `{"path":"README.md"}{"path":"AGENTS.md"}`},
+				{Type: zeroruntime.StreamEventToolCallEnd, ToolCallID: "call-1"},
+				{Type: zeroruntime.StreamEventDone},
+			},
+			{
+				{Type: zeroruntime.StreamEventText, Content: "done"},
+				{Type: zeroruntime.StreamEventDone},
+			},
+		},
+	}
+
+	result, err := Run(context.Background(), "read", provider, Options{Registry: registry, MaxTurns: 2})
+	if err != nil {
+		t.Fatal(err)
+	}
+	if result.FinalAnswer != "done" {
+		t.Fatalf("final answer = %q, want done", result.FinalAnswer)
+	}
+	var toolResult string
+	for _, message := range provider.requests[1].Messages {
+		if message.Role == zeroruntime.MessageRoleTool {
+			toolResult = message.Content
+		}
+	}
+	if strings.Contains(toolResult, "Failed to parse arguments") {
+		t.Fatalf("concatenated args should recover the first object, not fail to parse: %q", toolResult)
+	}
+	if !strings.Contains(toolResult, "hello readme") {
+		t.Fatalf("expected README.md contents (the first object's path), got %q", toolResult)
+	}
+}
+
 func TestRunDefersSelfCorrectFeedbackUntilAfterToolBatch(t *testing.T) {
 	// A single assistant turn with two tool calls where the first mutates and
 	// self-correct fails must keep both tool_results contiguous, with the feedback
@@ -1820,6 +1946,11 @@ func TestRunCommandPrefixApprovalBypassesSandboxForMatchingShellCalls(t *testing
 
 func TestRunCommandPrefixApprovalCoversSegmentedShellWithSafeTail(t *testing.T) {
 	root := t.TempDir()
+	segmentedCommand := `ps aux | head -5`
+	if runtime.GOOS == "windows" {
+		// head is MSYS-prone on Windows (#458) and no longer counts as a known-safe tail.
+		segmentedCommand = `ps aux | echo ok`
+	}
 	retryTool := &sandboxDeniedRetryTool{}
 	registry := tools.NewRegistry()
 	registry.Register(retryTool)
@@ -1833,7 +1964,7 @@ func TestRunCommandPrefixApprovalCoversSegmentedShellWithSafeTail(t *testing.T)
 			},
 			{
 				{Type: zeroruntime.StreamEventToolCallStart, ToolCallID: "call-2", ToolName: "bash"},
-				{Type: zeroruntime.StreamEventToolCallDelta, ToolCallID: "call-2", ArgumentsFragment: `{"command":"ps aux | head -5"}`},
+				{Type: zeroruntime.StreamEventToolCallDelta, ToolCallID: "call-2", ArgumentsFragment: fmt.Sprintf(`{"command":%q}`, segmentedCommand)},
 				{Type: zeroruntime.StreamEventToolCallEnd, ToolCallID: "call-2"},
 				{Type: zeroruntime.StreamEventDone},
 			},
@@ -2782,8 +2913,10 @@ func TestBuildSystemPromptInjectsHostShellContext(t *testing.T) {
 		t.Fatalf("expected operating system in environment block, got %q", prompt)
 	}
 	if runtime.GOOS == "windows" {
-		if !strings.Contains(prompt, "Windows cmd.exe syntax") || !strings.Contains(prompt, "cwd argument") {
-			t.Fatalf("expected Windows cmd.exe shell guidance in prompt, got %q", prompt)
+		for _, want := range []string{"Windows cmd.exe syntax", "cwd argument", "MSYS binaries", "grep", "require_escalated"} {
+			if !strings.Contains(prompt, want) {
+				t.Fatalf("expected Windows shell guidance to mention %q, got %q", want, prompt)
+			}
 		}
 	} else if !strings.Contains(prompt, "/bin/sh syntax") {
 		t.Fatalf("expected POSIX shell guidance in prompt, got %q", prompt)

internal/agent/parallel_tools.go

@@ -0,0 +1,97 @@
+package agent
+
+import (
+	"context"
+	"sync"
+
+	"github.com/Gitlawb/zero/internal/tools"
+)
+
+// Parallel read-ahead for tool batches. When a turn requests several
+// independent lookups (read_file + grep + glob is the common shape), executing
+// them one after another serializes pure I/O waits. A consecutive run of
+// auto-allowed read-only calls is executed concurrently instead; results are
+// then consumed in the original call order, so guard counters, message
+// ordering, abort semantics, and the surface's call/result event pairing are
+// byte-identical to sequential execution. Runs never span a mutating call: a
+// read that follows a write must observe the write, so eligibility is decided
+// per consecutive run, not per batch.
+
+// maxParallelReadTools bounds concurrent read-only tool executions in a turn.
+const maxParallelReadTools = 8
+
+// precomputedToolResult is one parallel read-ahead execution, keyed back to
+// its batch index by the caller.
+type precomputedToolResult struct {
+	result   ToolResult
+	abortErr error
+}
+
+// parallelSafeToolCall reports whether call may run concurrently with its
+// neighbors: the tool must exist, be side-effect-free (SideEffectRead), and be
+// auto-allowed for these args, so no interactive prompt or workspace mutation
+// is on the hot path. Loop-intercepted tools (ask_user, request_permissions)
+// and tool_search (mutates the deferred-tool set) stay sequential.
+func parallelSafeToolCall(registry *tools.Registry, call ToolCall, options Options) bool {
+	switch call.Name {
+	case "ask_user", tools.RequestPermissionsToolName, tools.ToolSearchToolName:
+		return false
+	}
+	tool, found := registry.Get(call.Name)
+	if !found || tool.Safety().SideEffect != tools.SideEffectRead {
+		return false
+	}
+	args := map[string]any{}
+	if call.Arguments != "" {
+		if err := decodeToolArguments(call.Arguments, &args); err != nil {
+			return false
+		}
+	}
+	return effectivePermission(tool, args) == tools.PermissionAllow
+}
+
+// executeParallelReadBatch runs calls[start:end] concurrently (bounded by
+// maxParallelReadTools) and returns results indexed relative to start. All
+// execution-side callbacks that can fire inside executeToolCall are serialized
+// behind one mutex: a permission prompt (a sandbox preflight can demand one
+// even for an auto-allowed read) must never appear twice at once on an
+// interactive front-end, and OnPermission event handlers append to shared
+// session-recording state without their own locking — two batched reads under
+// a granted extra root would otherwise race (the pre-batch serial loop never
+// had two callbacks in flight at once).
+func executeParallelReadBatch(ctx context.Context, registry *tools.Registry, calls []ToolCall, start, end int, permissionMode PermissionMode, options Options) []precomputedToolResult {
+	batchOptions := options
+	var callbackMutex sync.Mutex
+	if options.OnPermissionRequest != nil {
+		inner := options.OnPermissionRequest
+		batchOptions.OnPermissionRequest = func(ctx context.Context, request PermissionRequest) (PermissionDecision, error) {
+			callbackMutex.Lock()
+			defer callbackMutex.Unlock()
+			return inner(ctx, request)
+		}
+	}
+	if options.OnPermission != nil {
+		inner := options.OnPermission
+		batchOptions.OnPermission = func(event PermissionEvent) {
+			callbackMutex.Lock()
+			defer callbackMutex.Unlock()
+			inner(event)
+		}
+	}
+
+	results := make([]precomputedToolResult, end-start)
+	semaphore := make(chan struct{}, maxParallelReadTools)
+	var waitGroup sync.WaitGroup
+	for index := start; index < end; index++ {
+		waitGroup.Add(1)
+		go func(index int) {
+			defer waitGroup.Done()
+			semaphore <- struct{}{}
+			defer func() { <-semaphore }()
+			result, abortErr := executeToolCall(ctx, registry, calls[index], permissionMode, batchOptions)
+			results[index-start] = precomputedToolResult{result: result, abortErr: abortErr}
+		}(index)
+	}
+	waitGroup.Wait()
+	return results
+}

internal/agent/parallel_tools_test.go

@@ -0,0 +1,204 @@
+package agent
+
+import (
+	"context"
+	"fmt"
+	"sync"
+	"testing"
+	"time"
+
+	"github.com/Gitlawb/zero/internal/tools"
+	"github.com/Gitlawb/zero/internal/zeroruntime"
+)
+
+// probeTool records execution overlap and ordering so tests can assert what
+// actually ran concurrently.
+type probeTool struct {
+	name       string
+	sideEffect tools.SideEffect
+	delay      time.Duration
+	// shared, when set, receives this probe's start/end entries too, so a test
+	// can observe ordering ACROSS probes (e.g. reads vs a write barrier).
+	shared *probeLog
+
+	mu        sync.Mutex
+	active    int
+	maxActive int
+	log       []string
+}
+
+// probeLog is a mutex-guarded event log shared between probes.
+type probeLog struct {
+	mu      sync.Mutex
+	entries []string
+}
+
+func (log *probeLog) append(entry string) {
+	log.mu.Lock()
+	defer log.mu.Unlock()
+	log.entries = append(log.entries, entry)
+}
+
+func (log *probeLog) snapshot() []string {
+	log.mu.Lock()
+	defer log.mu.Unlock()
+	return append([]string(nil), log.entries...)
+}
+
+func (tool *probeTool) Name() string        { return tool.name }
+func (tool *probeTool) Description() string { return "test probe tool" }
+func (tool *probeTool) Parameters() tools.Schema {
+	return tools.Schema{
+		Type:                 "object",
+		Properties:           map[string]tools.PropertySchema{"id": {Type: "string"}},
+		AdditionalProperties: false,
+	}
+}
+func (tool *probeTool) Safety() tools.Safety {
+	return tools.Safety{SideEffect: tool.sideEffect, Permission: tools.PermissionAllow, Reason: "test"}
+}
+func (tool *probeTool) Run(_ context.Context, args map[string]any) tools.Result {
+	id, _ := args["id"].(string)
+	tool.mu.Lock()
+	tool.active++
+	if tool.active > tool.maxActive {
+		tool.maxActive = tool.active
+	}
+	tool.log = append(tool.log, "start:"+id)
+	tool.mu.Unlock()
+	if tool.shared != nil {
+		tool.shared.append("start:" + id)
+	}
+	time.Sleep(tool.delay)
+	if tool.shared != nil {
+		tool.shared.append("end:" + id)
+	}
+	tool.mu.Lock()
+	tool.active--
+	tool.log = append(tool.log, "end:"+id)
+	tool.mu.Unlock()
+	return tools.Result{Status: tools.StatusOK, Output: "probe " + id}
+}
+
+func probeCallEvents(callID, toolName, id string) []zeroruntime.StreamEvent {
+	return []zeroruntime.StreamEvent{
+		{Type: zeroruntime.StreamEventToolCallStart, ToolCallID: callID, ToolName: toolName},
+		{Type: zeroruntime.StreamEventToolCallDelta, ToolCallID: callID, ArgumentsFragment: fmt.Sprintf(`{"id":%q}`, id)},
+		{Type: zeroruntime.StreamEventToolCallEnd, ToolCallID: callID},
+	}
+}
+
+func TestParallelSafeToolCall(t *testing.T) {
+	registry := tools.NewRegistry()
+	registry.Register(&probeTool{name: "probe_read", sideEffect: tools.SideEffectRead})
+	registry.Register(&probeTool{name: "probe_write", sideEffect: tools.SideEffectWrite})
+
+	call := func(name, args string) ToolCall { return ToolCall{ID: "c", Name: name, Arguments: args} }
+	if !parallelSafeToolCall(registry, call("probe_read", `{"id":"a"}`), Options{}) {
+		t.Fatal("auto-allowed read tool must be parallel-safe")
+	}
+	if parallelSafeToolCall(registry, call("probe_write", `{"id":"a"}`), Options{}) {
+		t.Fatal("mutating tool must not be parallel-safe")
+	}
+	if parallelSafeToolCall(registry, call("unknown_tool", `{}`), Options{}) {
+		t.Fatal("unknown tool must not be parallel-safe")
+	}
+	if parallelSafeToolCall(registry, call("probe_read", `{"id":`), Options{}) {
+		t.Fatal("undecodable arguments must not be parallel-safe")
+	}
+	if parallelSafeToolCall(registry, call("ask_user", `{}`), Options{}) {
+		t.Fatal("loop-intercepted tools must stay sequential")
+	}
+}
+
+func TestRunExecutesConsecutiveReadsConcurrently(t *testing.T) {
+	probe := &probeTool{name: "probe_read", sideEffect: tools.SideEffectRead, delay: 60 * time.Millisecond}
+	registry := tools.NewRegistry()
+	registry.Register(probe)
+
+	turnOne := append(probeCallEvents("call-1", "probe_read", "a"), probeCallEvents("call-2", "probe_read", "b")...)
+	turnOne = append(turnOne, probeCallEvents("call-3", "probe_read", "c")...)
+	turnOne = append(turnOne, zeroruntime.StreamEvent{Type: zeroruntime.StreamEventDone})
+	provider := &mockProvider{
+		turns: [][]zeroruntime.StreamEvent{
+			turnOne,
+			{{Type: zeroruntime.StreamEventText, Content: "done"}, {Type: zeroruntime.StreamEventDone}},
+		},
+	}
+
+	var results []ToolResult
+	_, err := Run(context.Background(), "probe", provider, Options{
+		Registry:     registry,
+		OnToolResult: func(result ToolResult) { results = append(results, result) },
+	})
+	if err != nil {
+		t.Fatal(err)
+	}
+	if probe.maxActive < 2 {
+		t.Fatalf("consecutive read-only calls must overlap, max concurrency was %d", probe.maxActive)
+	}
+	// Results must still be recorded in original call order.
+	if len(results) != 3 || results[0].ToolCallID != "call-1" || results[1].ToolCallID != "call-2" || results[2].ToolCallID != "call-3" {
+		t.Fatalf("tool results out of order: %#v", results)
+	}
+	messages := provider.requests[1].Messages
+	var toolOrder []string
+	for _, message := range messages {
+		if message.Role == zeroruntime.MessageRoleTool {
+			toolOrder = append(toolOrder, message.ToolCallID)
+		}
+	}
+	if len(toolOrder) != 3 || toolOrder[0] != "call-1" || toolOrder[1] != "call-2" || toolOrder[2] != "call-3" {
+		t.Fatalf("recorded tool messages out of order: %v", toolOrder)
+	}
+}
+
+func TestRunParallelReadsNeverSpanMutatingCall(t *testing.T) {
+	shared := &probeLog{}
+	read := &probeTool{name: "probe_read", sideEffect: tools.SideEffectRead, delay: 30 * time.Millisecond, shared: shared}
+	write := &probeTool{name: "probe_write", sideEffect: tools.SideEffectWrite, shared: shared}
+	registry := tools.NewRegistry()
+	registry.Register(read)
+	registry.Register(write)
+
+	turnOne := append(probeCallEvents("call-1", "probe_read", "r1"), probeCallEvents("call-2", "probe_read", "r2")...)
+	turnOne = append(turnOne, probeCallEvents("call-3", "probe_write", "w")...)
+	turnOne = append(turnOne, probeCallEvents("call-4", "probe_read", "r3")...)
+	turnOne = append(turnOne, probeCallEvents("call-5", "probe_read", "r4")...)
+	turnOne = append(turnOne, zeroruntime.StreamEvent{Type: zeroruntime.StreamEventDone})
+	provider := &mockProvider{
+		turns: [][]zeroruntime.StreamEvent{
+			turnOne,
+			{{Type: zeroruntime.StreamEventText, Content: "done"}, {Type: zeroruntime.StreamEventDone}},
+		},
+	}
+
+	_, err := Run(context.Background(), "probe", provider, Options{Registry: registry})
+	if err != nil {
+		t.Fatal(err)
+	}
+
+	// Cross-probe ordering on the SHARED log: the write must start only after
+	// both first-batch reads finished, and both second-batch reads must start
+	// only after the write finished — batches never cross a mutating call.
+	log := shared.snapshot()
+	index := func(entry string) int {
+		for i, e := range log {
+			if e == entry {
+				return i
+			}
+		}
+		t.Fatalf("entry %q missing from shared log: %v", entry, log)
+		return -1
+	}
+	writeStart, writeEnd := index("start:w"), index("end:w")
+	if firstBatchMaxEnd := max(index("end:r1"), index("end:r2")); writeStart < firstBatchMaxEnd {
+		t.Fatalf("write started before the first read batch finished: %v", log)
+	}
+	if secondBatchMinStart := min(index("start:r3"), index("start:r4")); secondBatchMinStart < writeEnd {
+		t.Fatalf("second read batch started before the write finished: %v", log)
+	}
+	if read.maxActive < 2 {
+		t.Fatalf("reads within a batch must overlap, max concurrency was %d", read.maxActive)
+	}
+}

internal/agent/partition_cache_stable_test.go

@@ -0,0 +1,44 @@
+package agent
+
+import (
+	"testing"
+
+	"github.com/Gitlawb/zero/internal/tools"
+)
+
+func TestPartitionToolsActiveAppendsLoadedToolAfterEagerBlock(t *testing.T) {
+	// Cache-stability property: loading a deferred tool must APPEND it after the
+	// always-eager tools, not sort it into the middle of them — otherwise the tool
+	// definitions (the provider's cacheable prefix) shift on every load. The old
+	// behaviour alpha-sorted the whole array, so "mcp__srv__alpha" jumped ahead of
+	// "read_file".
+	root := t.TempDir()
+	registry := tools.NewRegistry()
+	registry.Register(tools.NewReadFileTool(root)) // non-deferred, "read_file"
+	registry.Register(fakeToolSearchTool{})
+	registry.Register(fakeDeferredTool{name: "mcp__srv__alpha", desc: "alpha"})
+	registry.Register(fakeDeferredTool{name: "mcp__srv__beta", desc: "beta"})
+
+	exposed, _ := partitionTools(registry, PermissionModeAuto, Options{DeferThreshold: 2}, map[string]bool{"mcp__srv__alpha": true})
+
+	pos := func(name string) int {
+		for i := range exposed {
+			if exposed[i].Name == name {
+				return i
+			}
+		}
+		return -1
+	}
+	readPos, searchPos, alphaPos := pos("read_file"), pos("tool_search"), pos("mcp__srv__alpha")
+	if readPos < 0 || searchPos < 0 || alphaPos < 0 {
+		t.Fatalf("expected read_file, tool_search and loaded alpha all exposed, got %#v", exposed)
+	}
+	// The eager tools must precede the loaded deferred tool (appended, not inserted).
+	if readPos >= alphaPos || searchPos >= alphaPos {
+		t.Fatalf("loaded deferred tool must be appended after the eager block; got read=%d search=%d alpha=%d", readPos, searchPos, alphaPos)
+	}
+	// beta was never loaded — it stays hidden.
+	if pos("mcp__srv__beta") != -1 {
+		t.Fatalf("unloaded deferred tool must stay hidden, got %#v", exposed)
+	}
+}

internal/agent/partition_cache_test.go

@@ -0,0 +1,63 @@
+package agent
+
+import (
+	"context"
+	"reflect"
+	"testing"
+
+	"github.com/Gitlawb/zero/internal/tools"
+	"github.com/Gitlawb/zero/internal/zeroruntime"
+)
+
+// countingSchemaTool records how many times its Parameters() schema is read, so a
+// test can prove the definition cache avoids re-rendering across turns.
+type countingSchemaTool struct {
+	name  string
+	calls *int
+}
+
+func (t countingSchemaTool) Name() string        { return t.name }
+func (t countingSchemaTool) Description() string { return "counts schema reads" }
+func (t countingSchemaTool) Parameters() tools.Schema {
+	*t.calls++
+	return tools.Schema{Type: "object", AdditionalProperties: false, Properties: map[string]tools.PropertySchema{
+		"x": {Type: "string"},
+	}}
+}
+func (t countingSchemaTool) Safety() tools.Safety {
+	return tools.Safety{SideEffect: tools.SideEffectRead, Permission: tools.PermissionAllow}
+}
+func (t countingSchemaTool) Run(_ context.Context, _ map[string]any) tools.Result {
+	return tools.Result{Status: tools.StatusOK}
+}
+
+// The cache renders a tool's schema once and reuses it across calls, and its
+// output is identical to the uncached path.
+func TestPartitionToolsCacheRendersOnceAndMatchesUncached(t *testing.T) {
+	calls := 0
+	registry := tools.NewRegistry()
+	registry.Register(countingSchemaTool{name: "alpha", calls: &calls})
+	registry.Register(countingSchemaTool{name: "beta", calls: &calls})
+
+	uncached, _ := partitionTools(registry, PermissionModeAuto, Options{}, map[string]bool{})
+
+	base := calls
+	cache := map[string]zeroruntime.ToolDefinition{}
+	first, _ := partitionToolsCached(registry, PermissionModeAuto, Options{}, map[string]bool{}, cache)
+	rendersAfterFirst := calls - base
+	second, _ := partitionToolsCached(registry, PermissionModeAuto, Options{}, map[string]bool{}, cache)
+	rendersAfterSecond := calls - base
+
+	if rendersAfterFirst != 2 {
+		t.Fatalf("first cached call should render 2 tools once each, rendered %d times", rendersAfterFirst)
+	}
+	if rendersAfterSecond != 2 {
+		t.Fatalf("second cached call must reuse the cache (no new schema reads), total renders = %d", rendersAfterSecond)
+	}
+	if !reflect.DeepEqual(first, second) {
+		t.Fatal("cached calls must return identical definitions")
+	}
+	if !reflect.DeepEqual(first, uncached) {
+		t.Fatal("cached output must match the uncached partitionTools output")
+	}
+}

internal/agent/permission_additional_perms_test.go

@@ -0,0 +1,70 @@
+package agent
+
+import (
+	"context"
+	"testing"
+
+	"github.com/Gitlawb/zero/internal/sandbox"
+	"github.com/Gitlawb/zero/internal/tools"
+)
+
+// promptShellTool is a minimal shell-style tool whose Safety requires a prompt,
+// standing in for bash so buildPermissionEvent can be exercised directly.
+type promptShellTool struct{}
+
+func (promptShellTool) Name() string        { return "bash" }
+func (promptShellTool) Description() string { return "run a shell command" }
+func (promptShellTool) Parameters() tools.Schema {
+	return tools.Schema{Type: "object", AdditionalProperties: false}
+}
+func (promptShellTool) Safety() tools.Safety {
+	return tools.Safety{SideEffect: tools.SideEffectShell, Permission: tools.PermissionPrompt, Reason: "runs a shell command"}
+}
+func (promptShellTool) Run(context.Context, map[string]any) tools.Result {
+	return tools.Result{Status: tools.StatusOK}
+}
+
+// A command whose sandbox decision is Allow but that explicitly requests
+// additional sandbox permissions is an ELEVATION: it must be surfaced as a
+// prompt, never Action=allow. Previously it carried allow while the loop still
+// blocked on OnPermissionRequest, and the TUI (which renders only prompts)
+// dropped it → the run deadlocked forever. This is the exact gpt-5.5
+// folder-creation shape captured in the reported hang.
+func TestBuildPermissionEventPromptsForAdditionalPermissions(t *testing.T) {
+	call := ToolCall{ID: "call_1", Name: "bash", Arguments: `{}`}
+	args := map[string]any{
+		"command":             "mkdir foo && cd foo && go mod init foo",
+		"sandbox_permissions": string(tools.SandboxPermissionsWithAdditionalPermissions),
+	}
+	// Base command allowed by the sandbox…
+	decision := &sandbox.Decision{Action: sandbox.ActionAllow}
+
+	event, ok := buildPermissionEvent(call, promptShellTool{}, args, false, PermissionModeAsk, Options{}, decision)
+	if !ok {
+		t.Fatal("buildPermissionEvent returned ok=false; the elevation request must produce an event")
+	}
+	if event.Action != PermissionActionPrompt {
+		t.Fatalf("Action = %q, want prompt — an additional-permissions elevation must ask the user, not auto-allow", event.Action)
+	}
+	// Sanity: shouldRequestPermission agrees the loop blocks on this.
+	if !shouldRequestPermission(promptShellTool{}, args, false, decision) {
+		t.Fatal("shouldRequestPermission must be true for an additional-permissions request")
+	}
+}
+
+// Without the additional-permissions flag, a sandbox-allowed command is NOT a
+// prompt (shouldRequestPermission is false; it just runs) — the override must
+// not over-reach and prompt for ordinary allowed commands.
+func TestBuildPermissionEventKeepsAllowForOrdinaryAllowedCommand(t *testing.T) {
+	call := ToolCall{ID: "call_2", Name: "bash", Arguments: `{}`}
+	args := map[string]any{"command": "ls"}
+	decision := &sandbox.Decision{Action: sandbox.ActionAllow}
+
+	event, _ := buildPermissionEvent(call, promptShellTool{}, args, false, PermissionModeAsk, Options{}, decision)
+	if event.Action == PermissionActionPrompt {
+		t.Fatalf("Action = prompt for an ordinary allowed command; want allow (no spurious prompt)")
+	}
+	if shouldRequestPermission(promptShellTool{}, args, false, decision) {
+		t.Fatal("shouldRequestPermission must be false for an ordinary sandbox-allowed command")
+	}
+}

internal/agent/plan_staleness_test.go

@@ -0,0 +1,64 @@
+package agent
+
+import (
+	"strings"
+	"testing"
+
+	"github.com/Gitlawb/zero/internal/zeroruntime"
+)
+
+func planUpdateTurn(status string) zeroruntime.CollectedStream {
+	return zeroruntime.CollectedStream{ToolCalls: []zeroruntime.ToolCall{
+		{Name: "update_plan", Arguments: `{"plan":[{"content":"step one","status":"` + status + `"}]}`},
+	}}
+}
+
+// Turn-based staleness: a plan that goes many turns without an update — while
+// items are still pending — draws the stale reminder even though few tool calls
+// have run (the tool-call streak alone would not have tripped).
+func TestPlanReminderFiresOnTurnStaleness(t *testing.T) {
+	state := newGuardState()
+	state.observeTurn(planUpdateTurn("in_progress"))
+
+	// Text-only turns advance the turn counter without adding tool calls.
+	for range stalePlanTurnThreshold {
+		state.observeTurn(zeroruntime.CollectedStream{Text: "still working…"})
+	}
+
+	if state.toolCallsSincePlanUpdate >= staleToolCallThreshold {
+		t.Fatalf("precondition: the tool-call trigger must not fire (%d calls)", state.toolCallsSincePlanUpdate)
+	}
+	if got := state.planReminder(stalePlanTurnThreshold + 2); !strings.Contains(got, planStaleReminderMarker) {
+		t.Fatalf("expected a turn-based stale reminder, got %q", got)
+	}
+}
+
+// A fully-completed plan is NOT stale: no pending items means the turn-based
+// trigger stays quiet even after many turns, so a finished plan isn't nagged.
+func TestPlanReminderSkipsTurnStalenessWhenComplete(t *testing.T) {
+	state := newGuardState()
+	state.observeTurn(planUpdateTurn("completed"))
+
+	for range stalePlanTurnThreshold + 3 {
+		state.observeTurn(zeroruntime.CollectedStream{Text: "wrapping up…"})
+	}
+
+	if got := state.planReminder(stalePlanTurnThreshold + 5); got != "" {
+		t.Fatalf("a completed plan must not draw a stale reminder, got %q", got)
+	}
+}
+
+// A plan update resets the turn counter, so the stale reminder does not fire
+// right after a fresh update.
+func TestPlanUpdateResetsTurnStaleness(t *testing.T) {
+	state := newGuardState()
+	state.observeTurn(planUpdateTurn("in_progress"))
+	for range stalePlanTurnThreshold - 1 {
+		state.observeTurn(zeroruntime.CollectedStream{Text: "working…"})
+	}
+	// Refresh the plan — resets the turn streak.
+	state.observeTurn(planUpdateTurn("in_progress"))
+	if got := state.planReminder(stalePlanTurnThreshold + 1); got != "" {
+		t.Fatalf("a just-refreshed plan must not be stale, got %q", got)
+	}
+}

internal/agent/prompt_budget_test.go

@@ -0,0 +1,51 @@
+package agent
+
+import (
+	"testing"
+
+	"github.com/Gitlawb/zero/internal/tools"
+)
+
+// These ceilings are a deliberate ratchet on Zero's fixed per-turn overhead — the
+// system prompt and the eager tool schemas that ride on EVERY request. They are
+// set ~10% above the current measured cost, using ApproxTextTokens (the same
+// estimate the compaction loop and /context use, ~non-whitespace-bytes/4). A change
+// that pushes either past its ceiling must be justified (and the ceiling raised
+// deliberately) or trimmed — the per-turn floor should not creep up silently.
+//
+// Measured baselines (2026-07): base system prompt ~3160 tokens; the tools a normal
+// (auto permission-mode) interactive turn sends ~2430 tokens. The tool figure is
+// the auto-mode advertised set, not the full core registry — higher-risk tools that
+// only advertise in other modes are excluded, matching what a real turn actually
+// pays.
+const (
+	maxBaseSystemPromptTokens = 3500
+	maxEagerToolSchemaTokens  = 2700
+)
+
+func TestSystemPromptTokenBudget(t *testing.T) {
+	// Minimal render: a model is set (so the session block renders) but no Cwd, so
+	// the workspace map, project guidelines, and repo map are excluded — this is the
+	// fixed base every session pays regardless of the workspace it runs in.
+	prompt := buildSystemPrompt(Options{Model: "claude-opus-4-8"})
+	got := ApproxTextTokens(prompt)
+	t.Logf("base system prompt: %d tokens (%d bytes)", got, len(prompt))
+	if got > maxBaseSystemPromptTokens {
+		t.Fatalf("base system prompt is %d tokens, over the %d ceiling — trim it or raise the ceiling deliberately", got, maxBaseSystemPromptTokens)
+	}
+}
+
+func TestEagerToolSchemaTokenBudget(t *testing.T) {
+	registry := tools.NewRegistry()
+	for _, tool := range tools.CoreToolsScoped(t.TempDir(), nil) {
+		registry.Register(tool)
+	}
+	// Options{} keeps DeferThreshold at 0, so deferral is inactive and every core
+	// tool is exposed eagerly — exactly what a plugin-free session sends each turn.
+	exposed, _ := partitionTools(registry, PermissionModeAuto, Options{}, map[string]bool{})
+	got := estimateToolDefTokens(exposed)
+	t.Logf("eager core tool schemas: %d tokens across %d tools", got, len(exposed))
+	if got > maxEagerToolSchemaTokens {
+		t.Fatalf("eager tool schemas are %d tokens, over the %d ceiling — defer a tool or raise the ceiling deliberately", got, maxEagerToolSchemaTokens)
+	}
+}

internal/agent/reconnect.go

@@ -3,9 +3,11 @@ package agent
 import (
 	"context"
 	"fmt"
+	"math/rand"
 	"strings"
 	"time"
 
+	"github.com/Gitlawb/zero/internal/errhint"
 	"github.com/Gitlawb/zero/internal/zeroruntime"
 )
 
@@ -17,10 +19,20 @@ import (
 // times. We retry ONLY the connect (not a partially-consumed stream), so no
 // already-forwarded OnText is ever duplicated.
 const (
-	maxStreamReconnects = 2
-	streamReconnectBase = 500 * time.Millisecond
+	// maxStreamReconnects is how many times the connect is re-issued after a
+	// transient disconnect. 4 (not 2): with jittered exponential backoff this rides
+	// out a multi-second network blip (~0.5s + 1s + 2s + 4s worst case) instead of
+	// dying on a 2s hiccup and re-burning every token on a restart.
+	maxStreamReconnects = 4
+	// streamReconnectMax caps a single backoff so the tail attempts don't wait
+	// minutes on a long outage.
+	streamReconnectMax = 8 * time.Second
 )
 
+// streamReconnectBase is the first-retry delay (doubled each subsequent attempt).
+// A var, not a const, so tests can shrink it to keep the exhaustion path fast.
+var streamReconnectBase = 500 * time.Millisecond
+
 // reconnectNotifier is called before each retry with the 1-based attempt number
 // and the max, so the caller can surface a "Reconnecting N/max" notice. Nil is
 // fine.
@@ -40,6 +52,21 @@ func reconnectNoticeFor(options Options) reconnectNotifier {
 	}
 }
 
+// stallRetryNoticeFor builds a notifier for the loop's content-stall retry (a
+// stream that connected and produced only transient output, then went silent).
+// Distinct wording from reconnectNoticeFor's "connection lost" because the
+// connection was fine — the model stalled. Surfaced through OnReasoning, the
+// non-content channel, so it never folds into the answer text. Nil when there
+// is no reasoning sink (the retry still happens silently).
+func stallRetryNoticeFor(options Options) reconnectNotifier {
+	if options.OnReasoning == nil {
+		return nil
+	}
+	return func(attempt, max int) {
+		options.OnReasoning(fmt.Sprintf("\n[no output — model stalled; retrying %d/%d…]\n", attempt, max))
+	}
+}
+
 // streamWithReconnect issues request via provider.StreamCompletion and, on a
 // transient disconnect error, retries the connect up to maxStreamReconnects
 // times with exponential backoff. It returns the live stream on success, or the
@@ -58,7 +85,7 @@ func streamWithReconnect(ctx context.Context, provider Provider, request zerorun
 		if notify != nil {
 			notify(attempt, maxStreamReconnects)
 		}
-		if waitErr := sleepWithContext(ctx, backoffFor(attempt)); waitErr != nil {
+		if waitErr := sleepWithContext(ctx, jitteredBackoff(attempt)); waitErr != nil {
 			return nil, err // ctx cancelled while waiting; surface the original error
 		}
 		stream, err = provider.StreamCompletion(ctx, request)
@@ -81,6 +108,19 @@ func shouldReconnect(ctx context.Context, err error) bool {
 	if isContextLimitError(msg) || isImageRejectionError(err) {
 		return false
 	}
+	// HTTP 5xx statuses are non-reconnectable and must be excluded BEFORE the
+	// transport substring match below — otherwise "504 Gateway Timeout" would slip
+	// through on the generic "timeout" needle. 503 already exhausted
+	// providerio.SendWithRetry (retrying is a redundant double-retry); 500/502/504
+	// are non-idempotent by providerio's rule — the completion POST may already
+	// have reached the model before the upstream/gateway gave up, so replaying the
+	// connect risks duplicate billable work. Digit-boundary matched so an
+	// incidental "504" inside a latency/id number is not mistaken for a status.
+	if errhint.HasStatusCode(msg, "500", "502", "503", "504") {
+		return false
+	}
+	// Transport-level disconnects only. A genuine transport failure (EOF, reset,
+	// refused, timeout) means no response was received, which is safe to reconnect.
 	for _, needle := range []string{
 		"eof",
 		"connection reset",
@@ -91,8 +131,6 @@ func shouldReconnect(ctx context.Context, err error) bool {
 		"timed out",
 		"temporarily unavailable",
 		"i/o timeout",
-		"503",
-		"502",
 		"server closed",
 		"unexpected end",
 	} {
@@ -103,14 +141,31 @@ func shouldReconnect(ctx context.Context, err error) bool {
 	return false
 }
 
+// backoffFor is the deterministic exponential base delay for a 1-based attempt,
+// capped at streamReconnectMax. Jitter is layered on separately (jitteredBackoff).
 func backoffFor(attempt int) time.Duration {
 	d := streamReconnectBase
 	for i := 1; i < attempt; i++ {
+		if d >= streamReconnectMax {
+			return streamReconnectMax
+		}
 		d *= 2
 	}
+	if d > streamReconnectMax {
+		d = streamReconnectMax
+	}
 	return d
 }
 
+// jitteredBackoff adds up to 50% random jitter on top of backoffFor so concurrent
+// runs (swarm members, a cron fleet) that all trip on the same outage don't
+// reconnect in lockstep and hammer a recovering endpoint. Never shorter than the
+// deterministic base, so backoff still grows attempt over attempt.
+func jitteredBackoff(attempt int) time.Duration {
+	base := backoffFor(attempt)
+	return base + time.Duration(rand.Int63n(int64(base/2)+1))
+}
+
 func sleepWithContext(ctx context.Context, d time.Duration) error {
 	timer := time.NewTimer(d)
 	defer timer.Stop()

internal/agent/reconnect_test.go

@@ -44,6 +44,10 @@ func TestStreamWithReconnectRecoversFromTransientDisconnect(t *testing.T) {
 }
 
 func TestStreamWithReconnectGivesUpAfterMax(t *testing.T) {
+	// Shrink the backoff so exhausting all retries stays fast (real base would
+	// sleep ~7.5s across 4 attempts).
+	defer func(orig time.Duration) { streamReconnectBase = orig }(streamReconnectBase)
+	streamReconnectBase = time.Millisecond
 	// Always fails with a disconnect error → exhausts retries and returns it.
 	p := &flakyProvider{failBefore: 99, failErr: errors.New("connection reset by peer")}
 	_, err := streamWithReconnect(context.Background(), p, zeroruntime.CompletionRequest{}, nil)
@@ -87,7 +91,7 @@ func TestShouldReconnectClassification(t *testing.T) {
 	ctx := context.Background()
 	disconnects := []string{
 		"unexpected EOF", "connection reset by peer", "broken pipe",
-		"i/o timeout", "503 Service Unavailable", "server closed the connection",
+		"i/o timeout", "server closed the connection", "connection refused",
 	}
 	for _, m := range disconnects {
 		if !shouldReconnect(ctx, errors.New(m)) {
@@ -97,6 +101,13 @@ func TestShouldReconnectClassification(t *testing.T) {
 	notDisconnects := []string{
 		"context length exceeded", "invalid api key", "model not found",
 		"400 bad request: unsupported parameter",
+		// HTTP 5xx statuses are handled by providerio.SendWithRetry (503) or are
+		// non-idempotent (500/502/504); the reconnect path must not double-retry
+		// them. "504 Gateway Timeout" must NOT slip through on the generic "timeout"
+		// transport signal — a gateway timeout is non-idempotent, since the upstream
+		// may have already processed the completion POST before giving up.
+		"503 Service Unavailable", "502 Bad Gateway",
+		"504 Gateway Timeout", "500 Internal Server Error",
 	}
 	for _, m := range notDisconnects {
 		if shouldReconnect(ctx, errors.New(m)) {
@@ -112,6 +123,24 @@ func TestBackoffGrows(t *testing.T) {
 	if backoffFor(2) != 2*streamReconnectBase {
 		t.Fatalf("attempt 2 backoff = %v, want %v", backoffFor(2), 2*streamReconnectBase)
 	}
+	// The exponential base is capped so late attempts don't wait minutes.
+	if got := backoffFor(20); got != streamReconnectMax {
+		t.Fatalf("attempt 20 backoff = %v, want cap %v", got, streamReconnectMax)
+	}
+}
+
+func TestJitteredBackoffStaysInBounds(t *testing.T) {
+	// Jitter never drops below the deterministic base and never exceeds base*1.5,
+	// so backoff still grows attempt over attempt while decorrelating retries.
+	for attempt := 1; attempt <= 5; attempt++ {
+		base := backoffFor(attempt)
+		for i := 0; i < 200; i++ {
+			got := jitteredBackoff(attempt)
+			if got < base || got > base+base/2 {
+				t.Fatalf("attempt %d jittered backoff %v out of [%v, %v]", attempt, got, base, base+base/2)
+			}
+		}
+	}
 }
 
 func TestReconnectNoticeRoutesThroughReasoning(t *testing.T) {

internal/agent/skills_context_test.go

@@ -0,0 +1,88 @@
+package agent
+
+import (
+	"strconv"
+	"strings"
+	"testing"
+)
+
+func TestSkillsContextCapsLongList(t *testing.T) {
+	// 60 skills with realistic trigger-rich descriptions (~180 chars each, under
+	// the 200-rune truncation) total well past the 4096-byte list budget, so the
+	// overflow summary must kick in while the block stays bounded.
+	longDesc := strings.Repeat("use when the request touches deployments, release notes, or verification; ", 2) + "number "
+	skills := make([]SkillInfo, 0, 60)
+	for i := 0; i < 60; i++ {
+		n := strconv.Itoa(i)
+		skills = append(skills, SkillInfo{Name: "skill-" + n, Description: longDesc + n})
+	}
+	got := skillsContext(Options{Skills: skills})
+	if len(got) > skillsContextListBudget+1000 {
+		t.Fatalf("skills block should stay bounded near the %d-byte budget, got %d bytes:\n%s", skillsContextListBudget, len(got), got)
+	}
+	if !strings.Contains(got, "more (call skill") {
+		t.Fatalf("expected an overflow summary line, got:\n%s", got)
+	}
+	// The first skill is always listed regardless of budget.
+	if !strings.Contains(got, "- skill-0:") {
+		t.Fatalf("expected the first skill to always be listed, got:\n%s", got)
+	}
+}
+
+// A realistic mid-size skill set (20 skills, trigger-rich descriptions) must be
+// listed in FULL — no overflow summary. This pins the fix for the old 640-byte
+// budget, under which skills past ~#6 were invisible to the model and therefore
+// never triggered.
+func TestSkillsContextListsRealisticSetInFull(t *testing.T) {
+	desc := "Use when the user asks about deployments, release notes, or pre-merge verification runs."
+	skills := make([]SkillInfo, 0, 20)
+	for i := 0; i < 20; i++ {
+		skills = append(skills, SkillInfo{Name: "skill-" + strconv.Itoa(i), Description: desc})
+	}
+	got := skillsContext(Options{Skills: skills})
+	if strings.Contains(got, "more (call skill") {
+		t.Fatalf("20 described skills must all be listed without overflow, got:\n%s", got)
+	}
+	for i := 0; i < 20; i++ {
+		if !strings.Contains(got, "- skill-"+strconv.Itoa(i)+":") {
+			t.Fatalf("skill-%d missing from the list:\n%s", i, got)
+		}
+	}
+}
+
+func TestSkillsContext(t *testing.T) {
+	if got := skillsContext(Options{}); got != "" {
+		t.Fatalf("no skills should yield an empty section, got %q", got)
+	}
+	got := skillsContext(Options{Skills: []SkillInfo{
+		{Name: "commit-writer", Description: "Write a conventional-commit message."},
+		{Name: "  ", Description: "nameless, should be skipped"},
+		{Name: "reviewer"},
+	}})
+	if !strings.Contains(got, "<available_skills>") || !strings.Contains(got, "</available_skills>") {
+		t.Fatalf("missing available_skills block: %q", got)
+	}
+	if !strings.Contains(got, "- commit-writer: Write a conventional-commit message.") {
+		t.Fatalf("missing commit-writer line: %q", got)
+	}
+	if !strings.Contains(got, "- reviewer\n") {
+		t.Fatalf("reviewer (no description) line missing: %q", got)
+	}
+	if strings.Contains(got, "nameless") {
+		t.Fatalf("nameless entry should be skipped: %q", got)
+	}
+}
+
+func TestSystemPromptIncludesSkillsOnlyWhenInstalled(t *testing.T) {
+	with := buildSystemPrompt(Options{Skills: []SkillInfo{
+		{Name: "commit-writer", Description: "Write a commit message."},
+	}})
+	if !strings.Contains(with, "<available_skills>") || !strings.Contains(with, "skill tool") {
+		t.Fatalf("expected available_skills guidance in system prompt: %q", with)
+	}
+	// Default (no skills) must reproduce the prior prompt: no skills block.
+	without := buildSystemPrompt(Options{})
+	if strings.Contains(without, "<available_skills>") {
+		t.Fatalf("available_skills block must not appear without skills")
+	}
+}

internal/agent/stall_retry_test.go

@@ -0,0 +1,168 @@
+package agent
+
+import (
+	"context"
+	"sync/atomic"
+	"testing"
+
+	"github.com/Gitlawb/zero/internal/tools"
+	"github.com/Gitlawb/zero/internal/zeroruntime"
+)
+
+// stallProvider connects successfully (HTTP 200) but the stream emits a stall/idle
+// timeout error for the first stallBefore calls, then succeeds with "done". The
+// pre-stall emission simulates different partial-output shapes:
+//   - partialText: forwarded PROSE (StreamEventText) — must NOT be retried
+//     (re-issuing would duplicate visible answer text).
+//   - reasoningText: transient reasoning preview — IS retried.
+//   - partialToolCall: a tool call started (+ arg delta) but never ended, the
+//     "froze mid-write_file" case — IS retried (the incomplete call is never
+//     executed or committed).
+type stallProvider struct {
+	calls           int32
+	stallBefore     int32
+	partialText     string
+	reasoningText   string
+	partialToolCall string
+}
+
+func (p *stallProvider) StreamCompletion(_ context.Context, _ zeroruntime.CompletionRequest) (<-chan zeroruntime.StreamEvent, error) {
+	n := atomic.AddInt32(&p.calls, 1)
+	ch := make(chan zeroruntime.StreamEvent, 5)
+	if n <= p.stallBefore {
+		if p.partialText != "" {
+			ch <- zeroruntime.StreamEvent{Type: zeroruntime.StreamEventText, Content: p.partialText}
+		}
+		if p.reasoningText != "" {
+			ch <- zeroruntime.StreamEvent{Type: zeroruntime.StreamEventReasoning, Content: p.reasoningText}
+		}
+		if p.partialToolCall != "" {
+			// A tool call that starts and streams a partial argument fragment but
+			// never gets StreamEventToolCallEnd, then the stream errors — exactly
+			// the gpt-5.x "began the big write_file then froze" shape.
+			ch <- zeroruntime.StreamEvent{Type: zeroruntime.StreamEventToolCallStart, ToolCallID: "tc_1", ToolName: p.partialToolCall}
+			ch <- zeroruntime.StreamEvent{Type: zeroruntime.StreamEventToolCallDelta, ToolCallID: "tc_1", ArgumentsFragment: `{"path":"x.html","content":"<!doctype`}
+		}
+		ch <- zeroruntime.StreamEvent{
+			Type:  zeroruntime.StreamEventError,
+			Error: "provider stream error: no output for 6m (the model produced nothing)",
+		}
+		close(ch)
+		return ch, nil
+	}
+	ch <- zeroruntime.StreamEvent{Type: zeroruntime.StreamEventText, Content: "done"}
+	ch <- zeroruntime.StreamEvent{Type: zeroruntime.StreamEventDone}
+	close(ch)
+	return ch, nil
+}
+
+// A no-output stall is re-issued on a fresh connection and recovers — this is the
+// macOS stale-pooled-connection hang turned into an automatic recovery.
+func TestRunRetriesStreamStallWithNoOutput(t *testing.T) {
+	p := &stallProvider{stallBefore: 1}
+	result, err := Run(context.Background(), "go", p, Options{Registry: tools.NewRegistry()})
+	if err != nil {
+		t.Fatalf("a no-output stall should retry to success, got %v", err)
+	}
+	if result.FinalAnswer != "done" {
+		t.Fatalf("final answer = %q, want %q", result.FinalAnswer, "done")
+	}
+	if got := atomic.LoadInt32(&p.calls); got != 2 {
+		t.Fatalf("want 2 calls (1 stall + 1 retry), got %d", got)
+	}
+}
+
+// A stall AFTER partial output must NOT be retried — re-issuing would duplicate
+// the already-streamed text. It surfaces the error instead.
+func TestRunDoesNotRetryStallAfterPartialOutput(t *testing.T) {
+	p := &stallProvider{stallBefore: 1, partialText: "partial"}
+	_, err := Run(context.Background(), "go", p, Options{Registry: tools.NewRegistry()})
+	if err == nil {
+		t.Fatal("a stall after partial output must NOT be retried; want an error")
+	}
+	if got := atomic.LoadInt32(&p.calls); got != 1 {
+		t.Fatalf("partial-then-stall must not retry, got %d calls", got)
+	}
+}
+
+// A stall after only transient REASONING was forwarded (no prose text, no
+// completed tool call) IS retried: reasoning is a "thinking" preview, not
+// committed answer text, so re-streaming it on a clearly-signalled retry is
+// acceptable and recovers the common "reasoned then froze" stall. Only
+// forwarded final prose (OnText) blocks the retry (collected.Text is empty
+// here, so this exercises the transient-vs-prose gate specifically).
+func TestRunRetriesStallAfterOnlyReasoning(t *testing.T) {
+	p := &stallProvider{stallBefore: 1, reasoningText: "thinking hard"}
+	result, err := Run(context.Background(), "go", p, Options{
+		Registry:    tools.NewRegistry(),
+		OnReasoning: func(string) {},
+	})
+	if err != nil {
+		t.Fatalf("a stall after only reasoning should retry to success, got %v", err)
+	}
+	if result.FinalAnswer != "done" {
+		t.Fatalf("final answer = %q, want %q", result.FinalAnswer, "done")
+	}
+	if got := atomic.LoadInt32(&p.calls); got != 2 {
+		t.Fatalf("want 2 calls (1 stall + 1 retry), got %d", got)
+	}
+}
+
+// A stall mid-way through an INCOMPLETE tool call (started + partial args, never
+// ended) IS retried — the exact gpt-5.x / ollama "began the big write_file then
+// froze" case. The incomplete call is never executed (a turn returns on the
+// stall error before dispatch) and never committed to history, so re-issuing is
+// safe and recovers to success. This is the primary scenario this change fixes.
+func TestRunRetriesStallAfterIncompleteToolCall(t *testing.T) {
+	starts := 0
+	p := &stallProvider{stallBefore: 1, reasoningText: "planning", partialToolCall: "write_file"}
+	result, err := Run(context.Background(), "go", p, Options{
+		Registry:        tools.NewRegistry(),
+		OnReasoning:     func(string) {},
+		OnToolCallStart: func(string, string) { starts++ },
+		OnToolCallDelta: func(string, string) {},
+	})
+	if err != nil {
+		t.Fatalf("a stall mid-incomplete-tool-call should retry to success, got %v", err)
+	}
+	if result.FinalAnswer != "done" {
+		t.Fatalf("final answer = %q, want %q", result.FinalAnswer, "done")
+	}
+	if got := atomic.LoadInt32(&p.calls); got != 2 {
+		t.Fatalf("want 2 calls (1 stall + 1 retry), got %d", got)
+	}
+	if starts != 1 {
+		t.Fatalf("the incomplete tool call should have been forwarded once before the stall, got %d starts", starts)
+	}
+}
+
+// A persistent stall surfaces an error after exhausting the capped retries.
+func TestRunGivesUpAfterMaxStallRetries(t *testing.T) {
+	p := &stallProvider{stallBefore: 99}
+	_, err := Run(context.Background(), "go", p, Options{Registry: tools.NewRegistry()})
+	if err == nil {
+		t.Fatal("a persistent stall must surface an error after exhausting retries")
+	}
+	if got := atomic.LoadInt32(&p.calls); got != int32(1+maxStreamStallRetries) {
+		t.Fatalf("want %d calls (1 + %d retries), got %d", 1+maxStreamStallRetries, maxStreamStallRetries, got)
+	}
+}
+
+func TestIsStreamTimeoutError(t *testing.T) {
+	timeouts := []string{
+		"provider stream error: no output for 10m (the model produced nothing)",
+		"provider stream error: idle timeout after 5m0s (upstream stopped sending data)",
+		"stream stalled (upstream kept the connection alive but produced no output)",
+	}
+	for _, m := range timeouts {
+		if !isStreamTimeoutError(m) {
+			t.Fatalf("want timeout-classified: %q", m)
+		}
+	}
+	notTimeouts := []string{"", "context length exceeded", "rate limit error: slow down", "model not found"}
+	for _, m := range notTimeouts {
+		if isStreamTimeoutError(m) {
+			t.Fatalf("must NOT classify as timeout: %q", m)
+		}
+	}
+}

internal/agent/system_prompt.go

@@ -5,9 +5,11 @@ import (
 	"os"
 	"path/filepath"
 	"runtime"
+	"strconv"
 	"strings"
 	"unicode/utf8"
 
+	"github.com/Gitlawb/zero/internal/config"
 	"github.com/Gitlawb/zero/internal/repomap"
 	"github.com/Gitlawb/zero/internal/workspaceseed"
 )
@@ -37,6 +39,14 @@ const fallbackSystemPrompt = "You are Zero, a terminal coding agent. Help with t
 // general-to-specific order.
 var projectContextFiles = []string{"AGENTS.md", "ZERO.md", ".zero/AGENTS.md"}
 
+// userContextFile is the per-user instruction file, resolved under
+// config.UserConfigDir()/zero/ alongside the rest of Zero's per-user config
+// (config.json, commands, specialists) so users can keep personal guidance out
+// of individual repositories.
+const userContextFile = "ZERO.md"
+
+var userConfigDirForPrompt = config.UserConfigDir
+
 // maxProjectContextBytes caps how much of a single project doc is injected so
 // a large guidelines file can't blow the context budget.
 const maxProjectContextBytes = 8 << 10 // 8 KiB per file
@@ -80,12 +90,22 @@ func buildSystemPrompt(options Options) string {
 	if seed := workspaceSeedContext(options.Cwd); seed != "" {
 		sections = append(sections, seed)
 	}
+	// User guidelines are injected before workspace/project guidelines so the
+	// project's AGENTS.md/ZERO.md is the later, more specific instruction
+	// block. See userGuidelines for the explicit precedence note carried in
+	// the section text itself.
+	if user := userGuidelines(); user != "" {
+		sections = append(sections, user)
+	}
 	if ws := workspaceContext(options.Cwd); ws != "" {
 		sections = append(sections, ws)
 	}
 	if delegation := specialistDelegationContext(options); delegation != "" {
 		sections = append(sections, delegation)
 	}
+	if skillsBlock := skillsContext(options); skillsBlock != "" {
+		sections = append(sections, skillsBlock)
+	}
 	if style := responseStyleContext(options); style != "" {
 		sections = append(sections, style)
 	}
@@ -164,6 +184,75 @@ func specialistDelegationContext(options Options) string {
 	return b.String()
 }
 
+// skillsContext lists the reusable skills the model can pull in on demand via the
+// skill tool, so it invokes the right one on the first try instead of guessing a
+// name, failing, and reading the error. It mirrors specialistDelegationContext:
+// names + one-line descriptions only (never skill bodies), and it renders nothing
+// when no skills are installed, so a skill-less run reproduces the previous prompt
+// byte-for-byte.
+// skillsContextListBudget bounds the bytes spent listing individual skills so a
+// workspace with an extreme number of them can't bloat every turn's prompt.
+// Skills past the budget are summarized as a count rather than dropped — the
+// model can still load any of them by name (an unknown name returns the full
+// list from the skill tool).
+//
+// The budget is deliberately generous — roughly 18 skills with maximally long
+// (200-rune, ASCII) descriptions, or 40+ with typical terser ones; non-ASCII
+// descriptions consume it faster since the budget counts bytes. This list is
+// the model's ONLY discovery surface, so a skill squeezed out of it effectively
+// never triggers. The previous 640-byte cap silently broke invocation for
+// anyone with more than ~6 skills — the model cannot match a request against a
+// skill it cannot see.
+const skillsContextListBudget = 4096
+
+func skillsContext(options Options) string {
+	if len(options.Skills) == 0 {
+		return ""
+	}
+	var b strings.Builder
+	b.WriteString("<available_skills>\n")
+	b.WriteString("Reusable, on-demand instruction sets you can load with the skill tool. Before acting on a request, scan this list; when the request matches a skill's name or description, call skill with its exact name FIRST and follow the loaded guidance — do not guess names, do not skip a matching skill, and do not substitute your own approach for its instructions.\n")
+	listed, spent, omitted := 0, 0, 0
+	for _, info := range options.Skills {
+		name := strings.TrimSpace(info.Name)
+		if name == "" {
+			continue
+		}
+		line := "- " + name
+		if desc := strings.TrimSpace(info.Description); desc != "" {
+			line += ": " + truncateForSkillLine(desc)
+		}
+		line += "\n"
+		// Always list at least one skill; past the budget, summarize the remainder
+		// as a count instead of silently dropping it.
+		if listed > 0 && spent+len(line) > skillsContextListBudget {
+			omitted++
+			continue
+		}
+		b.WriteString(line)
+		spent += len(line)
+		listed++
+	}
+	if omitted > 0 {
+		b.WriteString("- …and " + strconv.Itoa(omitted) + " more (call skill with a name; an unknown name lists them all)\n")
+	}
+	b.WriteString("</available_skills>")
+	return b.String()
+}
+
+// truncateForSkillLine keeps a skill's one-line description short so a single
+// verbose description can't dominate the skills-list budget. The cap is roomy on
+// purpose: the description carries the skill's trigger conditions ("use when…"),
+// and truncating those is what stops the model from ever invoking the skill.
+func truncateForSkillLine(desc string) string {
+	const maxDescRunes = 200
+	runes := []rune(desc)
+	if len(runes) <= maxDescRunes {
+		return desc
+	}
+	return strings.TrimSpace(string(runes[:maxDescRunes])) + "…"
+}
+
 func sessionRuntimeContext(options Options) string {
 	provider := strings.TrimSpace(options.ProviderName)
 	model := strings.TrimSpace(options.Model)
@@ -197,7 +286,7 @@ func workspaceContext(cwd string) string {
 	b.WriteString("Working directory: " + cwd + "\n")
 	b.WriteString("Operating system: " + runtime.GOOS + "\n")
 	if runtime.GOOS == "windows" {
-		b.WriteString("Shell syntax: Windows cmd.exe syntax for exec_command/bash tools; prefer the workdir/cwd argument instead of cd when changing directories.\n")
+		b.WriteString("Shell syntax: Windows cmd.exe syntax for exec_command/bash tools; prefer the workdir/cwd argument instead of cd when changing directories. Do not pipe to or invoke POSIX coreutils from Git for Windows (usr\\bin head/grep/tail/cat/...): they are MSYS binaries and fail under the write-restricted sandbox; use native Zero tools (grep, read_file, list_directory, glob) or cmd.exe findstr/more instead, or sandbox_permissions require_escalated only when host-level execution is truly required.\n")
 	} else {
 		b.WriteString("Shell syntax: /bin/sh syntax for exec_command/bash tools; prefer the workdir/cwd argument instead of cd when changing directories.\n")
 	}
@@ -250,13 +339,7 @@ func projectGuidelines(cwd, gitRoot string) string {
 		if remaining := maxProjectContextTotalBytes - totalUsed; remaining < limit {
 			limit = remaining
 		}
-		if len(content) > limit {
-			cut := limit
-			for cut > 0 && !utf8.RuneStart(content[cut]) {
-				cut--
-			}
-			content = content[:cut] + "\n… (truncated)"
-		}
+		content = truncateGuidelineContent(content, limit)
 		label := projectGuidelineLabel(match, gitRoot)
 		b.WriteString("\n\n## Project guidelines (" + label + ")\n\n" + content)
 		totalUsed += len(content)
@@ -264,6 +347,92 @@ func projectGuidelines(cwd, gitRoot string) string {
 	return b.String()
 }
 
+// userGuidelines returns the per-user ZERO.md instructions block, if present.
+// The file lives in config.UserConfigDir()/zero/ next to Zero's other
+// per-user config; the basename match is case-insensitive so a file saved as
+// zero.md still resolves on case-sensitive filesystems, mirroring the project
+// guideline loader. The section carries an explicit precedence note because
+// this is a global, personal preferences file: it is injected earlier in the
+// prompt than the project's AGENTS.md/ZERO.md (see buildSystemPrompt), and
+// the note keeps that precedence unambiguous even if a model weighs later
+// context more heavily than section order alone implies.
+func userGuidelines() string {
+	configDir, err := userConfigDirForPrompt()
+	if err != nil {
+		return ""
+	}
+	configDir = strings.TrimSpace(configDir)
+	if configDir == "" {
+		return ""
+	}
+	match := findCaseInsensitiveFile(filepath.Join(configDir, "zero"), userContextFile)
+	if match == "" {
+		return ""
+	}
+	data, err := os.ReadFile(match)
+	if err != nil {
+		return ""
+	}
+	content := strings.TrimSpace(string(data))
+	if content == "" {
+		return ""
+	}
+	content = truncateGuidelineContent(content, maxProjectContextBytes)
+	return "## User guidelines (" + filepath.Base(match) + ")\n\n" +
+		"These are the operator's personal preferences, not project policy. " +
+		"Where they conflict with a repository's project guidelines below (AGENTS.md/ZERO.md), the project guidelines take precedence.\n\n" +
+		content
+}
+
+// truncationMarker is appended to guideline content that was cut short.
+const truncationMarker = "\n… (truncated)"
+
+// truncateGuidelineContent caps content at limit bytes without splitting a
+// UTF-8 rune, appending a truncation marker when anything was cut. Space for
+// the marker is reserved before choosing the cut point. When limit is too
+// small to fit the marker itself, the marker is dropped instead, so the
+// returned string never exceeds limit for any non-negative limit.
+func truncateGuidelineContent(content string, limit int) string {
+	if len(content) <= limit {
+		return content
+	}
+	if limit <= 0 {
+		return ""
+	}
+	cut := limit - len(truncationMarker)
+	if cut < 0 {
+		cut = 0
+	}
+	for cut > 0 && !utf8.RuneStart(content[cut]) {
+		cut--
+	}
+	if truncated := content[:cut] + truncationMarker; len(truncated) <= limit {
+		return truncated
+	}
+	// limit is smaller than the marker itself: fall back to a hard,
+	// rune-safe cut at limit bytes with no marker rather than exceeding it.
+	end := limit
+	for end > 0 && !utf8.RuneStart(content[end]) {
+		end--
+	}
+	return content[:end]
+}
+
+// findCaseInsensitiveFile returns the on-disk path of the regular file in dir
+// whose name matches basename case-insensitively, or "" when absent.
+func findCaseInsensitiveFile(dir, basename string) string {
+	entries, err := os.ReadDir(dir)
+	if err != nil {
+		return ""
+	}
+	for _, e := range entries {
+		if !e.IsDir() && strings.EqualFold(e.Name(), basename) {
+			return filepath.Join(dir, e.Name())
+		}
+	}
+	return ""
+}
+
 // projectGuidelineDirs returns the directory chain from gitRoot to cwd
 // (inclusive), in root-to-leaf order. If gitRoot is empty or unreachable
 // from cwd, the chain collapses to [cwd].
@@ -310,29 +479,22 @@ func projectGuidelineLabel(match, gitRoot string) string {
 
 // findProjectContextFile returns the first matching project guideline file
 // in dir. Lookup is case-insensitive on the basename and on the actual
-// filename returned, so a git-tracked file like AGENTS.MD (uppercase MD)
-// still resolves to its true filename on case-sensitive filesystems — the
-// returned path is what should appear in the project guidelines label.
+// filename returned, so a git-tracked file like AGENTS.md still resolves to
+// its true filename on case-sensitive filesystems — the returned path is what
+// should appear in the project guidelines label.
 // Returns "" when nothing matches.
 func findProjectContextFile(dir string) string {
 	for _, name := range projectContextFiles {
-		baseLower := strings.ToLower(filepath.Base(name))
 		// Walk to the file's parent through dir with case-insensitive segment
-		// matching, then find a regular file with the same lowercased
-		// basename. This works on both case-sensitive and case-insensitive
-		// filesystems and always returns the on-disk filename.
+		// matching, then find a regular file with the same basename. This works
+		// on both case-sensitive and case-insensitive filesystems and always
+		// returns the on-disk filename.
 		parent, ok := resolveDirCaseInsensitive(filepath.Dir(filepath.Join(dir, name)), dir)
 		if !ok {
 			continue
 		}
-		entries, err := os.ReadDir(parent)
-		if err != nil {
-			continue
-		}
-		for _, e := range entries {
-			if !e.IsDir() && strings.ToLower(e.Name()) == baseLower {
-				return filepath.Join(parent, e.Name())
-			}
+		if match := findCaseInsensitiveFile(parent, filepath.Base(name)); match != "" {
+			return match
 		}
 	}
 	return ""

internal/agent/system_prompt.md

@@ -46,14 +46,10 @@ work.
    formatting-only edits unless the user asked for them.
 4. **Verify.** Verify after edits; see the testing gate below. This is
    mandatory.
-5. **Summarize.** For a substantial task, close with an ELABORATE, well-structured
-   summary — never a single throwaway line. Use short headings and bullets: a
-   one-line "what I built/changed"; **Where it lives** — the key files and how they
-   fit together; **How each requirement is met** — walk through each requirement
-   from the request and name the specifics that satisfy it; what validation ran and
-   what was NOT verified; and a brief "Next steps" or "Want me to…?" offer. Scale
-   the depth to the work — a big build earns several sections, a trivial fix earns
-   a line or two.
+5. **Summarize.** Close with a summary scaled to the work: a trivial fix earns one
+   line; a substantial change earns a few short bullets covering what changed, where
+   it lives, and what you did and did not verify. Lead with the outcome, and never
+   pad — do not restate the task or narrate steps the user already watched you take.
 
 ## Editing discipline
 
@@ -64,7 +60,13 @@ work.
 - Make one tool call per file. Do not batch multi-file writes into a single
   shell or script invocation.
 - For edits to existing files, prefer edit_file or apply_patch with minimal,
-  targeted diffs. Match the existing indentation, imports, and idioms.
+  targeted diffs. Match the existing indentation, imports, and idioms. Match the
+  file's comment density: do not add explanatory comments unless the user asks or
+  the code is already comment-dense.
+- Solve the problem as posed, not a more general version of it. Add no
+  speculative abstraction, configurability, or handling for cases that cannot
+  occur, and nothing the user did not ask for. A small diff can still be
+  over-built; if a 200-line solution could be 50, rewrite it.
 - Preserve behavior you were not asked to change. Do not delete or rewrite code
   you did not author unless the task requires it; if you must, say so.
 

internal/agent/system_prompt_models.go

@@ -6,11 +6,13 @@ import "strings"
 //
 // Different model families respond to different framing: the GPT family benefits
 // from an explicit markdown/output spec and a strong "prefer native tools" nudge;
-// Gemini benefits from explicit tool-preference and conciseness guidance; the
-// Claude family is already well-aligned with the core prompt and only needs a
-// light comment-discipline nudge. modelPromptAddendum returns the family-specific
-// block appended after the core prompt. Classification is by model id (the agent
-// only has the model string — no registry dependency).
+// Gemini benefits from explicit tool-preference and conciseness guidance. The
+// Claude family is already aligned with the core prompt and needs no addendum —
+// comment discipline is universal in the core prompt now, so every family gets it.
+// modelPromptAddendum returns the family-specific block appended after the core
+// prompt, or "" when the family is unknown or needs no specialization.
+// Classification is by model id (the agent only has the model string — no registry
+// dependency).
 
 const (
 	familyOpenAI    = "openai"
@@ -45,9 +47,10 @@ func modelPromptAddendum(model string) string {
 		return openAIPromptAddendum
 	case familyGemini:
 		return geminiPromptAddendum
-	case familyAnthropic:
-		return anthropicPromptAddendum
 	default:
+		// Anthropic (aligned with the core prompt) and unknown families get no
+		// family-specific block; comment discipline now lives in the core prompt
+		// for every model.
 		return ""
 	}
 }
@@ -70,8 +73,3 @@ const geminiPromptAddendum = `<model_guidance>
   in one short clause why it is needed.
 - Use update_plan for any multi-step task and keep it current.
 </model_guidance>`
-
-const anthropicPromptAddendum = `<model_guidance>
-- Do not add code comments unless the user asks or the surrounding code is
-  comment-dense; match the file's existing comment density.
-</model_guidance>`

internal/agent/system_prompt_models_test.go

@@ -29,9 +29,11 @@ func TestBuildSystemPromptAppendsModelAddendum(t *testing.T) {
 	if got := buildSystemPrompt(Options{Model: "gpt-5"}); !strings.Contains(got, openAIPromptAddendum) {
 		t.Fatalf("expected the OpenAI addendum in the gpt-5 prompt")
 	}
+	// Claude is aligned with the core prompt and gets no family addendum now that
+	// comment discipline is universal; it must not pick up another family's block.
 	claude := buildSystemPrompt(Options{Model: "claude-opus-4-6"})
-	if !strings.Contains(claude, anthropicPromptAddendum) {
-		t.Fatalf("expected the Anthropic addendum in the claude prompt")
+	if strings.Contains(claude, "<model_guidance>") {
+		t.Fatalf("expected no model_guidance block for Claude (aligned with core prompt)")
 	}
 	if strings.Contains(claude, openAIPromptAddendum) {
 		t.Fatalf("the claude prompt must not contain the OpenAI addendum")

internal/agent/system_prompt_test.go

@@ -9,6 +9,25 @@ import (
 	"github.com/Gitlawb/zero/internal/sandbox"
 )
 
+// TestMain points userConfigDirForPrompt at an empty, package-wide temp
+// directory for every test in this package. Without this, buildSystemPrompt
+// falls back to the real config.UserConfigDir, so any developer with a
+// personal ~/.config/zero/ZERO.md would get its content folded into
+// otherwise-unrelated prompt assertions, making test runs non-deterministic
+// across contributor machines. Tests that specifically exercise user
+// guidelines stub userConfigDirForPrompt themselves via
+// withSystemPromptTestUserConfigDir(Func) and restore this default on cleanup.
+func TestMain(m *testing.M) {
+	dir, err := os.MkdirTemp("", "zero-agent-tests-*")
+	if err != nil {
+		panic(err)
+	}
+	userConfigDirForPrompt = func() (string, error) { return dir, nil }
+	code := m.Run()
+	os.RemoveAll(dir)
+	os.Exit(code)
+}
+
 func TestCoreSystemPromptIncludesCodingQualityRules(t *testing.T) {
 	prompt := strings.ToLower(buildSystemPrompt(Options{}))
 
@@ -23,7 +42,8 @@ func TestCoreSystemPromptIncludesCodingQualityRules(t *testing.T) {
 		"avoid broad refactors",
 		"search the web before answering",
 		"do not recognize",
-		"how each requirement is met",
+		"scaled to the work",
+		"comment density",
 	} {
 		if !strings.Contains(prompt, want) {
 			t.Fatalf("expected core system prompt to include %q, got:\n%s", want, buildSystemPrompt(Options{}))
@@ -127,8 +147,90 @@ func systemPromptTestBlock(t *testing.T, prompt, start, end string) string {
 	return body
 }
 
+func TestBuildSystemPromptIncludesUserGuidelines(t *testing.T) {
+	configDir := t.TempDir()
+	writeSystemPromptTestFile(t, configDir, "zero/ZERO.md", "  Prefer concise summaries.  \n")
+	t.Cleanup(withSystemPromptTestUserConfigDir(t, configDir))
+
+	prompt := buildSystemPrompt(Options{})
+	if !strings.Contains(prompt, "## User guidelines (ZERO.md)") {
+		t.Fatalf("expected user guidelines header, got:\n%s", prompt)
+	}
+	if !strings.Contains(prompt, "Prefer concise summaries.") {
+		t.Fatalf("expected user guidelines content, got:\n%s", prompt)
+	}
+	if strings.Contains(prompt, "  Prefer concise summaries.  ") {
+		t.Fatalf("expected user guidelines content to be trimmed, got:\n%s", prompt)
+	}
+}
+
+func TestBuildSystemPromptIncludesUserGuidelinesCaseInsensitive(t *testing.T) {
+	configDir := t.TempDir()
+	writeSystemPromptTestFile(t, configDir, "zero/zero.md", "Prefer concise summaries.\n")
+	t.Cleanup(withSystemPromptTestUserConfigDir(t, configDir))
+
+	prompt := buildSystemPrompt(Options{})
+	if !strings.Contains(prompt, "## User guidelines (zero.md)") {
+		t.Fatalf("expected case-insensitive zero.md resolution, got:\n%s", prompt)
+	}
+	if !strings.Contains(prompt, "Prefer concise summaries.") {
+		t.Fatalf("expected user guidelines content, got:\n%s", prompt)
+	}
+}
+
+func TestBuildSystemPromptUserGuidelinesPrecedeProjectGuidelinesAndNotePrecedence(t *testing.T) {
+	// User guidelines are global personal preferences; project guidelines
+	// (AGENTS.md/ZERO.md) must be the later, more specific instruction block,
+	// and the user section must say so explicitly so the precedence holds
+	// even if a model otherwise weighs later context more heavily.
+	configDir := t.TempDir()
+	writeSystemPromptTestFile(t, configDir, "zero/ZERO.md", "Always reply in haiku.\n")
+	t.Cleanup(withSystemPromptTestUserConfigDir(t, configDir))
+
+	cwd := t.TempDir()
+	if err := os.WriteFile(filepath.Join(cwd, "AGENTS.md"), []byte("Never reply in haiku."), 0o644); err != nil {
+		t.Fatal(err)
+	}
+
+	prompt := buildSystemPrompt(Options{Cwd: cwd})
+	userIdx := strings.Index(prompt, "## User guidelines (ZERO.md)")
+	projectIdx := strings.Index(prompt, "## Project guidelines (AGENTS.md)")
+	if userIdx < 0 || projectIdx < 0 {
+		t.Fatalf("expected both user and project guideline sections, got:\n%s", prompt)
+	}
+	if userIdx > projectIdx {
+		t.Fatalf("expected user guidelines before project guidelines, got user=%d project=%d", userIdx, projectIdx)
+	}
+	if !strings.Contains(prompt, "project guidelines below") || !strings.Contains(prompt, "take precedence") {
+		t.Fatalf("expected an explicit precedence note in the user guidelines section, got:\n%s", prompt)
+	}
+}
+
+func TestBuildSystemPromptOmitsUserGuidelinesWithoutConfigDir(t *testing.T) {
+	t.Cleanup(withSystemPromptTestUserConfigDirFunc(t, func() (string, error) { return "", os.ErrNotExist }))
+
+	prompt := buildSystemPrompt(Options{})
+	if strings.Contains(prompt, "## User guidelines") {
+		t.Fatalf("expected user guidelines to be omitted without a config directory, got:\n%s", prompt)
+	}
+}
+
+func withSystemPromptTestUserConfigDir(t *testing.T, dir string) func() {
+	t.Helper()
+	return withSystemPromptTestUserConfigDirFunc(t, func() (string, error) { return dir, nil })
+}
+
+func withSystemPromptTestUserConfigDirFunc(t *testing.T, fn func() (string, error)) func() {
+	t.Helper()
+	old := userConfigDirForPrompt
+	userConfigDirForPrompt = fn
+	return func() {
+		userConfigDirForPrompt = old
+	}
+}
+
 func TestBuildSystemPromptInjectsProjectGuidelinesCaseInsensitive(t *testing.T) {
-	// Git tracks AGENTS.MD (uppercase MD) on a case-sensitive filesystem; the
+	// Git tracks AGENTS.md on a case-sensitive filesystem; the
 	// loader must still resolve it when the cwd lookup uses lowercase.
 	cwd := t.TempDir()
 	if err := os.WriteFile(filepath.Join(cwd, "AGENTS.MD"), []byte("Always run `make lint`."), 0o644); err != nil {
@@ -256,6 +358,30 @@ func TestBuildSystemPromptProjectGuidelinesTruncatesAtTotalCap(t *testing.T) {
 	}
 }
 
+func TestTruncateGuidelineContentStaysWithinLimit(t *testing.T) {
+	limit := 64
+	content := strings.Repeat("x", limit+1)
+	got := truncateGuidelineContent(content, limit)
+	if len(got) > limit {
+		t.Fatalf("truncated result is %d bytes, want at most %d", len(got), limit)
+	}
+	if !strings.Contains(got, "truncated") {
+		t.Fatalf("expected truncation marker, got %q", got)
+	}
+}
+
+func TestTruncateGuidelineContentStaysWithinLimitForSmallBudgets(t *testing.T) {
+	// Limits at or below the truncation marker's own length can't fit the
+	// marker at all; the helper must still never exceed the requested limit.
+	content := strings.Repeat("x", 100)
+	for _, limit := range []int{0, 1, len(truncationMarker) - 1, len(truncationMarker), len(truncationMarker) + 1} {
+		got := truncateGuidelineContent(content, limit)
+		if len(got) > limit {
+			t.Fatalf("truncateGuidelineContent(_, %d) = %q (%d bytes), want at most %d bytes", limit, got, len(got), limit)
+		}
+	}
+}
+
 func TestProjectGuidelineDirsOrdersRootToLeaf(t *testing.T) {
 	root := filepath.Join("r")
 	leaf := filepath.Join(root, "a", "b")

internal/agent/types.go

@@ -174,6 +174,16 @@ type SpecialistInfo struct {
 	WhenToUse string
 }
 
+// SkillInfo is a one-line summary of a reusable, on-demand skill (its name and
+// frontmatter description) surfaced to the system prompt so the model can invoke
+// the right skill with the skill tool on the first try instead of guessing a name
+// and reading the failure. Like SpecialistInfo it is plain data, so the agent
+// package needs no dependency on internal/skills.
+type SkillInfo struct {
+	Name        string
+	Description string
+}
+
 type Options struct {
 	MaxTurns int
 	// DeferThreshold activates deferred MCP-tool loading: when the number of
@@ -189,6 +199,12 @@ type Options struct {
 	// populated only where the Task tool is actually registered, so an empty slice
 	// (the default) reproduces the previous prompt byte-for-byte.
 	Specialists []SpecialistInfo
+	// Skills lists the reusable skills installed for this run (the default skills
+	// dir merged with any plugin skill roots). When non-empty the system prompt
+	// gains an <available_skills> block naming them so the model loads the right one
+	// via the skill tool on the first try. Empty (the default) reproduces the
+	// previous prompt byte-for-byte.
+	Skills []SkillInfo
 	// Specialist/sub-agent metadata is carried through exec now and consumed by
 	// the specialist runtime in later slices.
 	SessionID        string
@@ -269,6 +285,14 @@ type Options struct {
 	// ceiling and the autonomy gate. nil disables it entirely (the loop is
 	// byte-identical to before). One instance per run — it holds attempt state.
 	SelfCorrect *SelfCorrector
+	// FileDiagnostics, when set, checks files changed by mutating tools for
+	// error-severity language diagnostics IN THE BACKGROUND and appends any
+	// errors as a nudge before the model's next request — the model still sees
+	// an error it introduced at its next decision point, but no tool call ever
+	// blocks on the language server (the old inline path stalled every edit on
+	// a ≥300ms debounce, 10s cap). Build one with NewFileDiagnostics. nil
+	// disables post-edit diagnostics.
+	FileDiagnostics func(ctx context.Context, absPath string) string
 
 	// RequireCompletionSignal gates run completion for HEADLESS exec. Without it,
 	// any assistant turn that produces text but no tool call is accepted as the

internal/agenteval/benchmark_test.go

@@ -455,11 +455,15 @@ func TestHarnessTimeoutCancelsBlockedAgent(t *testing.T) {
 
 	// The timeout is longer than materialization of the small fixture even on
 	// slower Windows CI runners, so the agent is reached before it blocks until
-	// the deadline fires.
+	// the deadline fires. 10s wasn't generous enough in practice — it failed
+	// on an otherwise-unrelated PR merge (materialization alone consumed the
+	// full budget on a contended Windows runner, so the agent was never
+	// reached); widened for headroom, since the assertions below still bound
+	// worst-case test time via that same deadline, not this constant.
 	report := harness.Run(context.Background(), suitePath, suite, BenchmarkInput{
 		TaskID:   "document-stream-json-verify-events",
 		WorkRoot: t.TempDir(),
-		Timeout:  10 * time.Second,
+		Timeout:  30 * time.Second,
 	})
 
 	if !agentReached {

internal/cli/acp.go

@@ -43,8 +43,11 @@ func runACP(args []string, stdout io.Writer, stderr io.Writer, deps appDeps) int
 	conn := acp.NewConn(deps.stdin, stdout)
 	acp.NewAgent(conn, acp.Deps{
 		ResolveConfig: deps.resolveConfig,
-		NewProvider:   deps.newProvider,
-		RunAgent:      agent.Run,
+		// deps.newProvider is wrapped in fillAppDeps to apply the stored API key,
+		// so ACP is authenticated for apiKeyStored profiles like every other
+		// surface — no ACP-specific credential handling needed.
+		NewProvider: deps.newProvider,
+		RunAgent:    agent.Run,
 		// Build the SCOPED registry + sandbox engine per workspace, exactly like the
 		// exec surface, so ACP shell/file tools are confined — never run unconfined.
 		BuildWorkspace: func(workspaceRoot string, resolved config.ResolvedConfig) (*tools.Registry, *sandbox.Engine, error) {

internal/cli/app.go

@@ -12,6 +12,7 @@ import (
 	"os/exec"
 	"path/filepath"
 	"strings"
+	"sync"
 	"time"
 
 	"github.com/Gitlawb/zero/internal/agent"
@@ -19,9 +20,11 @@ import (
 	"github.com/Gitlawb/zero/internal/hooks"
 	"github.com/Gitlawb/zero/internal/localcontrol"
 	"github.com/Gitlawb/zero/internal/mcp"
+	"github.com/Gitlawb/zero/internal/modelregistry"
 	"github.com/Gitlawb/zero/internal/observability"
 	"github.com/Gitlawb/zero/internal/plugins"
 	"github.com/Gitlawb/zero/internal/providerhealth"
+	"github.com/Gitlawb/zero/internal/providermodeldiscovery"
 	"github.com/Gitlawb/zero/internal/provideronboarding"
 	"github.com/Gitlawb/zero/internal/providers"
 	"github.com/Gitlawb/zero/internal/redaction"
@@ -43,37 +46,46 @@ import (
 var version = "dev"
 
 type appDeps struct {
-	getwd                 func() (string, error)
-	stdin                 io.Reader
-	userConfigPath        func() (string, error)
-	resolveConfig         func(workspaceRoot string, overrides config.Overrides) (config.ResolvedConfig, error)
-	resolveMCPConfig      func(workspaceRoot string) (config.MCPConfig, error)
-	newProvider           func(config.ProviderProfile) (zeroruntime.Provider, error)
-	probeProviderHealth   func(context.Context, providerhealth.Options) providerhealth.Result
-	detectLocalRuntimes   func(context.Context, provideronboarding.LocalDetectOptions) []provideronboarding.DetectedLocalRuntime
-	newSessionStore       func() *sessions.Store
-	loadPlugins           func(plugins.LoadOptions) (plugins.LoadResult, error)
-	loadHooks             func(hooks.LoadOptions) (hooks.LoadResult, error)
-	skillsDir             func() string
-	pluginsDir            func() string
-	toolsDir              func() string
-	newMCPStore           func() (*mcp.PermissionStore, error)
-	newMCPTokenStore      func() (*mcp.TokenStore, error)
-	newSandboxStore       func() (*sandbox.GrantStore, error)
-	selectSandboxBackend  func(sandbox.BackendOptions) sandbox.Backend
-	runSandboxSetupHelper func(path string, args []string, stdout io.Writer, stderr io.Writer) error
-	registerMCPTools      func(context.Context, *tools.Registry, config.MCPConfig, mcp.RegisterOptions) (mcpToolRuntime, error)
-	prepareWorktree       func(context.Context, worktrees.Options) (worktrees.Result, error)
-	detectVerifyPlan      func(string) (verify.Plan, error)
-	runVerify             func(context.Context, verify.Plan, verify.RunOptions) verify.Report
-	runSelfVerify         func(context.Context, verify.Plan, selfverify.Options) selfverify.Report
-	runAgentEval          func(context.Context, agentEvalOptions) (agentEvalReport, error)
-	inspectChanges        func(context.Context, zerogit.InspectOptions) (zerogit.ChangeSummary, error)
-	commitChanges         func(context.Context, zerogit.CommitOptions) (zerogit.CommitResult, error)
-	runTUI                func(context.Context, tui.Options) int
-	runEditor             func(string) error
-	checkUpdate           func(context.Context, update.Options) (update.Result, error)
-	now                   func() time.Time
+	getwd            func() (string, error)
+	stdin            io.Reader
+	userConfigPath   func() (string, error)
+	resolveConfig    func(workspaceRoot string, overrides config.Overrides) (config.ResolvedConfig, error)
+	resolveMCPConfig func(workspaceRoot string) (config.MCPConfig, error)
+	newProvider      func(config.ProviderProfile) (zeroruntime.Provider, error)
+	// exportActiveProvider pins spawned children to the run's provider (production:
+	// config.SetActiveProviderEnv, set in defaultAppDeps — deliberately NOT filled
+	// by fillAppDeps, so tests never mutate the process environment unless they
+	// inject it). nil ⇒ no export.
+	exportActiveProvider   func(providerName string)
+	probeProviderHealth    func(context.Context, providerhealth.Options) providerhealth.Result
+	discoverProviderModels func(context.Context, config.ProviderProfile) ([]providermodeldiscovery.Model, error)
+	detectLocalRuntimes    func(context.Context, provideronboarding.LocalDetectOptions) []provideronboarding.DetectedLocalRuntime
+	newSessionStore        func() *sessions.Store
+	loadPlugins            func(plugins.LoadOptions) (plugins.LoadResult, error)
+	loadHooks              func(hooks.LoadOptions) (hooks.LoadResult, error)
+	skillsDir              func() string
+	pluginsDir             func() string
+	toolsDir               func() string
+	newMCPStore            func() (*mcp.PermissionStore, error)
+	newMCPTokenStore       func() (*mcp.TokenStore, error)
+	newSandboxStore        func() (*sandbox.GrantStore, error)
+	selectSandboxBackend   func(sandbox.BackendOptions) sandbox.Backend
+	runSandboxSetupHelper  func(path string, args []string, stdout io.Writer, stderr io.Writer) error
+	registerMCPTools       func(context.Context, *tools.Registry, config.MCPConfig, mcp.RegisterOptions) (mcpToolRuntime, error)
+	prepareWorktree        func(context.Context, worktrees.Options) (worktrees.Result, error)
+	detectVerifyPlan       func(string) (verify.Plan, error)
+	runVerify              func(context.Context, verify.Plan, verify.RunOptions) verify.Report
+	runSelfVerify          func(context.Context, verify.Plan, selfverify.Options) selfverify.Report
+	runAgentEval           func(context.Context, agentEvalOptions) (agentEvalReport, error)
+	inspectChanges         func(context.Context, zerogit.InspectOptions) (zerogit.ChangeSummary, error)
+	commitChanges          func(context.Context, zerogit.CommitOptions) (zerogit.CommitResult, error)
+	pushChanges            func(context.Context, zerogit.PushOptions) (zerogit.PushResult, error)
+	createPR               func(context.Context, zerogit.PROptions) (zerogit.PRResult, error)
+	runTUI                 func(context.Context, tui.Options) int
+	runEditor              func(string) error
+	checkUpdate            func(context.Context, update.Options) (update.Result, error)
+	applyUpdate            func(context.Context, update.Options) (update.ApplyResult, error)
+	now                    func() time.Time
 }
 
 type mcpToolRuntime interface {
@@ -99,9 +111,10 @@ func Run(args []string, stdout io.Writer, stderr io.Writer) int {
 
 func defaultAppDeps() appDeps {
 	return appDeps{
-		getwd:          os.Getwd,
-		stdin:          os.Stdin,
-		userConfigPath: config.DefaultUserConfigPath,
+		getwd:                os.Getwd,
+		stdin:                os.Stdin,
+		userConfigPath:       config.DefaultUserConfigPath,
+		exportActiveProvider: config.SetActiveProviderEnv,
 		resolveConfig: func(workspaceRoot string, overrides config.Overrides) (config.ResolvedConfig, error) {
 			options, err := config.DefaultResolveOptions(workspaceRoot)
 			if err != nil {
@@ -118,13 +131,19 @@ func defaultAppDeps() appDeps {
 			return config.ResolveMCP(options)
 		},
 		newProvider: func(profile config.ProviderProfile) (zeroruntime.Provider, error) {
+			// Resolve the OAuth login ONCE: the bearer resolver and the login key it
+			// bound must describe the same login (the key is passed on to the Codex
+			// account-header resolver so it never re-selects independently).
+			resolver, loginKey := oauthLoginForProfile(profile)
 			return providers.New(profile, providers.Options{
 				UserAgent:     userAgent(),
-				OAuthResolver: oauthResolverForProfile(profile),
+				OAuthResolver: resolver,
+				OAuthLoginKey: loginKey,
 			})
 		},
-		probeProviderHealth: providerhealth.Probe,
-		detectLocalRuntimes: provideronboarding.DetectLocalRuntimes,
+		probeProviderHealth:    providerhealth.Probe,
+		discoverProviderModels: defaultDiscoverProviderModels,
+		detectLocalRuntimes:    provideronboarding.DetectLocalRuntimes,
 		newSessionStore: func() *sessions.Store {
 			return sessions.NewStore(sessions.StoreOptions{})
 		},
@@ -163,9 +182,12 @@ func defaultAppDeps() appDeps {
 		runAgentEval:     defaultRunAgentEval,
 		inspectChanges:   zerogit.Inspect,
 		commitChanges:    zerogit.Commit,
+		pushChanges:      zerogit.Push,
+		createPR:         zerogit.CreatePR,
 		runTUI:           tui.Run,
 		runEditor:        openEditor,
 		checkUpdate:      update.Check,
+		applyUpdate:      update.Apply,
 		now:              time.Now,
 	}
 }
@@ -213,13 +235,19 @@ func runWithDeps(args []string, stdout io.Writer, stderr io.Writer, deps appDeps
 	defer observability.Recover(observability.DefaultCrashDir(), "cli", stderr, &exitCode)
 	deps = fillAppDeps(deps)
 
+	// CLI runs opt into the models.dev overlay (cached live context limits and
+	// pricing on top of the curated catalog). Explicitly enabled here — and only
+	// here — so library consumers and hermetic tests are never perturbed by a
+	// cache file on the machine. The refresh itself is fired in exec/TUI startup.
+	modelregistry.EnableModelsDevOverlay()
+
 	addDirs, args, err := splitLeadingAddDirFlags(args)
 	if err != nil {
 		return writeAppError(stderr, err.Error(), 1)
 	}
-	// --theme {auto|dark|light} selects the TUI palette non-interactively (populates
-	// tui.Options.Theme, which resolveThemeMode prefers over ZERO_THEME). Re-split
-	// --add-dir afterward so it may appear on either side of --theme.
+	// --theme <name> selects the TUI palette non-interactively (auto or any registered
+	// theme; populates tui.Options.Theme, which resolveThemeMode prefers over
+	// ZERO_THEME). Re-split --add-dir afterward so it may appear on either side of --theme.
 	theme, args, err := splitLeadingThemeFlag(args)
 	if err != nil {
 		return writeAppError(stderr, err.Error(), 1)
@@ -369,6 +397,8 @@ func runWithDeps(args []string, stdout io.Writer, stderr io.Writer, deps appDeps
 		return runSandbox(args[1:], stdout, stderr, deps)
 	case "update":
 		return runUpdate(args[1:], stdout, stderr, deps)
+	case "upgrade":
+		return runUpgrade(args[1:], stdout, stderr, deps)
 	case "worktrees", "worktree":
 		return runWorktrees(args[1:], stdout, stderr, deps)
 	case "verify":
@@ -391,6 +421,14 @@ func runWithDeps(args []string, stdout io.Writer, stderr io.Writer, deps appDeps
 		if _, err := fmt.Fprintf(stderr, "unknown command %q\n", args[0]); err != nil {
 			return 1
 		}
+		// First-run users reach for `zero login`/`zero logout` (reported in the
+		// wild); point them at the real command instead of a bare usage pointer.
+		switch strings.ToLower(args[0]) {
+		case "login", "logout":
+			if _, err := fmt.Fprintf(stderr, "did you mean %q?\n", "zero auth "+strings.ToLower(args[0])); err != nil {
+				return 1
+			}
+		}
 		if _, err := fmt.Fprintln(stderr, "Run zero --help for usage."); err != nil {
 			return 1
 		}
@@ -421,6 +459,9 @@ func fillAppDeps(deps appDeps) appDeps {
 	if deps.probeProviderHealth == nil {
 		deps.probeProviderHealth = defaults.probeProviderHealth
 	}
+	if deps.discoverProviderModels == nil {
+		deps.discoverProviderModels = defaults.discoverProviderModels
+	}
 	if deps.detectLocalRuntimes == nil {
 		deps.detectLocalRuntimes = defaults.detectLocalRuntimes
 	}
@@ -481,6 +522,12 @@ func fillAppDeps(deps appDeps) appDeps {
 	if deps.commitChanges == nil {
 		deps.commitChanges = defaults.commitChanges
 	}
+	if deps.pushChanges == nil {
+		deps.pushChanges = defaults.pushChanges
+	}
+	if deps.createPR == nil {
+		deps.createPR = defaults.createPR
+	}
 	if deps.runTUI == nil {
 		deps.runTUI = defaults.runTUI
 	}
@@ -490,9 +537,23 @@ func fillAppDeps(deps appDeps) appDeps {
 	if deps.checkUpdate == nil {
 		deps.checkUpdate = defaults.checkUpdate
 	}
+	if deps.applyUpdate == nil {
+		deps.applyUpdate = defaults.applyUpdate
+	}
 	if deps.now == nil {
 		deps.now = defaults.now
 	}
+	// Wrap newProvider ONCE, after all defaults are filled, so every surface that
+	// builds the runtime provider gets the credential-store key applied. The
+	// resolver is pure and leaves apiKeyStored profiles keyless; an unwrapped
+	// build sends unauthenticated requests. config.ApplyStoredAPIKey is a no-op
+	// when the key is already set or the profile isn't stored-key, so this is
+	// safe and idempotent for every profile kind and every injected newProvider.
+	baseNewProvider := deps.newProvider
+	userConfigPath := deps.userConfigPath
+	deps.newProvider = func(profile config.ProviderProfile) (zeroruntime.Provider, error) {
+		return baseNewProvider(applyStoredProviderKeyAt(profile, userConfigPath))
+	}
 	return deps
 }
 
@@ -501,6 +562,11 @@ func runInteractiveTUI(stderr io.Writer, deps appDeps, permissionMode agent.Perm
 }
 
 func runInteractiveTUIWithSetup(stderr io.Writer, deps appDeps, permissionMode agent.PermissionMode, addDirs []string, theme string, forceSetup bool) int {
+	// Refresh the models.dev pricing/limits cache in the background when stale;
+	// the overlay is read at registry construction from the cache file, so this
+	// benefits the next run and never blocks or fails this one.
+	go func() { _ = modelregistry.RefreshModelsDevCache(context.Background()) }()
+
 	workspaceRoot, err := deps.getwd()
 	if err != nil {
 		return writeAppError(stderr, "failed to resolve workspace: "+err.Error(), 1)
@@ -508,7 +574,32 @@ func runInteractiveTUIWithSetup(stderr io.Writer, deps appDeps, permissionMode a
 
 	resolved, err := deps.resolveConfig(workspaceRoot, config.Overrides{})
 	if err != nil {
-		return writeAppError(stderr, err.Error(), 1)
+		// A resolve failure the setup wizard can FIX is not fatal for the
+		// interactive TUI: drop into the wizard with an empty config so the user
+		// can onboard/repair, instead of exiting with an error they can only fix
+		// by hand-editing config.json. That covers a missing/unresolvable active
+		// provider and an active provider without a model (custom endpoints have
+		// no catalog default) — previously the second shape bricked bare `zero`
+		// and `zero setup`, the exact commands that could have fixed it. Any
+		// other error (malformed JSON, directory conflict, etc.) is still fatal,
+		// and headless commands (zero config / zero exec) still fail with the
+		// actionable message.
+		if !errors.Is(err, config.ErrNoActiveProvider) && !errors.Is(err, config.ErrProviderRequiresModel) {
+			return writeAppError(stderr, err.Error(), 1)
+		}
+		// ErrNoActiveProvider can mean "nothing configured yet" (needs onboarding)
+		// or "providers ARE configured, just none marked active" (e.g. config.json's
+		// activeProvider is blank/stale). In the second case resolved.Providers still
+		// carries the already-normalized list — prefer falling back to one of those
+		// over wiping everything and forcing the user to re-enter credentials they
+		// already saved.
+		if usable, ok := firstUsableProvider(resolved.Providers); errors.Is(err, config.ErrNoActiveProvider) && ok {
+			resolved.Provider = usable
+			resolved.ActiveProvider = usable.Name
+		} else {
+			resolved = config.ResolvedConfig{}
+			forceSetup = true
+		}
 	}
 	userConfigPath, err := deps.userConfigPath()
 	if err != nil {
@@ -641,6 +732,7 @@ func runInteractiveTUIWithSetup(stderr io.Writer, deps appDeps, permissionMode a
 	return deps.runTUI(context.Background(), tui.Options{
 		Cwd:                  workspaceRoot,
 		Theme:                theme,
+		SavedTheme:           resolved.Preferences.Theme,
 		UserConfigPath:       userConfigPath,
 		DoctorUserConfigPath: doctorUserConfigPath,
 		ProjectConfigPath:    projectConfigPath,
@@ -689,9 +781,19 @@ func runInteractiveTUIWithSetup(stderr io.Writer, deps appDeps, permissionMode a
 			Hooks:          newHookDispatcherWithExtra(workspaceRoot, pluginActivation.hooks),
 			DeferThreshold: resolved.Tools.DeferThreshold,
 			Specialists:    specialistRuntime.specialists,
+			Skills:         pluginActivation.skillInfos(deps.skillsDir()),
 		},
+		// LoadSkills backs /skills and direct /<skill-name> invocation in the TUI.
+		// It resolves against the same merged set (default dir + plugin skill
+		// roots) as the skill tool and the system-prompt list, re-read per use so
+		// newly installed skills work without a restart.
+		LoadSkills: cachedSkillsLoader(func() []skills.Skill {
+			merged, _ := plugins.MergedSkillsLoaded(deps.skillsDir(), pluginActivation.skillRoots)
+			return merged
+		}),
 		PermissionMode: permissionMode,
 		Notify:         resolved.Notify,
+		KeyBindings:    resolved.KeyBindings,
 		Setup: tui.SetupOptions{
 			Visible:    setupVisible,
 			Required:   needsSetup,
@@ -719,20 +821,50 @@ func tuiSandboxSetupCommand(backend sandbox.Backend, deps appDeps) func(context.
 	}
 }
 
+// buildProvider constructs the run's provider at STARTUP — it is called only from
+// the two launch paths (interactive TUI and headless exec), never from mid-run
+// rebuilds (escalation, wizard, ACP go through deps.newProvider directly).
 func buildProvider(resolved config.ResolvedConfig, deps appDeps) (zeroruntime.Provider, error) {
 	if !config.HasProviderProfile(resolved.Provider) {
 		return nil, nil
 	}
-	// Load the API key from the encrypted credential store when it isn't already
-	// resolved from inline config or an env var. This keeps the resolver pure (no
-	// I/O) and is a no-op for env/inline-key providers and when nothing is stored.
-	profile := resolved.Provider
-	if path, perr := deps.userConfigPath(); perr == nil {
+	// deps.newProvider is wrapped in fillAppDeps to apply the stored key, so this
+	// (and every other newProvider caller) needs no per-site key handling.
+	provider, err := deps.newProvider(resolved.Provider)
+	if err != nil {
+		return nil, err
+	}
+	// Pin spawned children (sub-agents / swarm members inherit the environment) to
+	// THIS run's provider from launch, not only after an in-session switch. Without
+	// the launch-time export a child re-resolves config.json at spawn time, so a
+	// provider switch persisted by ANOTHER zero process mid-session would silently
+	// move new children onto a different provider (and different credentials) than
+	// the parent is running. Runtime switches (/model, /provider, wizard,
+	// onboarding) re-export on commit, keeping the pin current. Injected via deps
+	// (nil in tests) because it mutates the process environment.
+	if deps.exportActiveProvider != nil {
+		deps.exportActiveProvider(resolved.Provider.Name)
+	}
+	return provider, nil
+}
+
+// applyStoredProviderKeyAt loads the API key from the encrypted credential store
+// when the resolver left it empty (the resolver is pure — no I/O — so an
+// apiKeyStored profile reaches this layer keyless). No-op for inline/env-key
+// providers and when nothing is stored. This is applied once, centrally, by the
+// deps.newProvider wrapper in fillAppDeps so EVERY surface that builds a runtime
+// provider (headless exec, the ACP builder, exec's mid-run escalation switcher,
+// and any future caller) is covered without a per-site invariant to forget.
+func applyStoredProviderKeyAt(profile config.ProviderProfile, userConfigPath func() (string, error)) config.ProviderProfile {
+	if userConfigPath == nil {
+		return profile
+	}
+	if path, perr := userConfigPath(); perr == nil {
 		if store, err := config.ProviderKeyStoreAt(filepath.Dir(path)); err == nil {
 			profile = config.ApplyStoredAPIKey(profile, store)
 		}
 	}
-	return deps.newProvider(profile)
+	return profile
 }
 
 func newCoreRegistry(workspaceRoot string) *tools.Registry {
@@ -1033,9 +1165,10 @@ func splitLeadingAddDirFlags(args []string) ([]string, []string, error) {
 	return addDirs, args, nil
 }
 
-// splitLeadingThemeFlag strips a leading --theme {auto|dark|light} (space or =form)
-// from the root argument list and validates it. The last occurrence wins. A value
-// outside the allowed set is a loud error rather than a silent fallback.
+// splitLeadingThemeFlag strips a leading --theme <auto|theme-name> (space or =form)
+// from the root argument list and validates it against the registered themes. The
+// last occurrence wins. A value outside the allowed set is a loud error rather than
+// a silent fallback.
 func splitLeadingThemeFlag(args []string) (string, []string, error) {
 	theme := ""
 	for len(args) > 0 {
@@ -1043,7 +1176,7 @@ func splitLeadingThemeFlag(args []string) (string, []string, error) {
 		switch {
 		case args[0] == "--theme":
 			if len(args) < 2 {
-				return "", nil, errors.New("--theme requires a value (auto, dark, or light)")
+				return "", nil, errors.New("--theme requires a value (auto or a theme name; try --theme auto)")
 			}
 			value = strings.TrimSpace(args[1])
 			args = args[2:]
@@ -1053,12 +1186,10 @@ func splitLeadingThemeFlag(args []string) (string, []string, error) {
 		default:
 			return theme, args, nil
 		}
-		switch strings.ToLower(value) {
-		case "auto", "dark", "light":
-			theme = strings.ToLower(value)
-		default:
-			return "", nil, fmt.Errorf("--theme must be auto, dark, or light (got %q)", value)
+		if !tui.ValidThemeArg(value) {
+			return "", nil, fmt.Errorf("--theme must be auto or a registered theme name (got %q)", value)
 		}
+		theme = strings.ToLower(value)
 	}
 	return theme, args, nil
 }
@@ -1078,7 +1209,8 @@ func profileHasCredential(profile config.ProviderProfile) bool {
 }
 
 // baseURLIsLoopback reports whether a provider base_url points at a loopback host
-// (a keyless local provider like Ollama/LM Studio), which needs no API key.
+// or a private-network host (192.168.x.y / 10.x.y.z / 172.16.x.y) — a local
+// provider like Ollama/LM Studio that needs no API key.
 func baseURLIsLoopback(baseURL string) bool {
 	u := strings.TrimSpace(baseURL)
 	if u == "" {
@@ -1093,7 +1225,7 @@ func baseURLIsLoopback(baseURL string) bool {
 		return true
 	}
 	if addr, err := netip.ParseAddr(host); err == nil {
-		return addr.IsLoopback()
+		return addr.IsLoopback() || addr.IsPrivate()
 	}
 	return false
 }
@@ -1154,3 +1286,26 @@ Flags:
 `)
 	return err
 }
+
+// cachedSkillsLoader memoizes a skills loader for a short interval. The TUI
+// calls the loader from autocomplete on every "/x" keystroke; without a cache
+// each keystroke would re-read every SKILL.md body. Two seconds is fresh enough
+// that a newly installed skill still shows up "immediately" while typing.
+func cachedSkillsLoader(load func() []skills.Skill) func() []skills.Skill {
+	var (
+		mu     sync.Mutex
+		at     time.Time
+		cached []skills.Skill
+	)
+	const ttl = 2 * time.Second
+	return func() []skills.Skill {
+		mu.Lock()
+		defer mu.Unlock()
+		if cached != nil && time.Since(at) < ttl {
+			return cached
+		}
+		cached = load()
+		at = time.Now()
+		return cached
+	}
+}

internal/cli/app_test.go

@@ -5,6 +5,7 @@ import (
 	"context"
 	"encoding/json"
 	"errors"
+	"fmt"
 	"io"
 	"os"
 	"path/filepath"
@@ -157,6 +158,141 @@ func TestRunNoArgsLaunchesSetupTUIWithNilProviderWhenNoProviderConfigured(t *tes
 	assertAgentOptions(t, launchedOptions, 12, agent.PermissionModeAsk)
 }
 
+func TestRunNoArgsEntersSetupWhenResolveReportsNoActiveProvider(t *testing.T) {
+	var stdout bytes.Buffer
+	var stderr bytes.Buffer
+	cwd := t.TempDir()
+	setCLIUserConfigRoot(t)
+	userConfigPath := filepath.Join(t.TempDir(), "zero", "config.json")
+	var launchedOptions tui.Options
+	launched := false
+
+	exitCode := runWithDeps([]string{}, &stdout, &stderr, appDeps{
+		getwd: func() (string, error) {
+			return cwd, nil
+		},
+		resolveConfig: func(workspaceRoot string, overrides config.Overrides) (config.ResolvedConfig, error) {
+			return config.ResolvedConfig{}, fmt.Errorf("%w: active provider %q not found", config.ErrNoActiveProvider, "ghost")
+		},
+		newProvider: func(config.ProviderProfile) (zeroruntime.Provider, error) {
+			t.Fatal("newProvider should not be called without a resolved provider")
+			return nil, nil
+		},
+		userConfigPath: func() (string, error) {
+			return userConfigPath, nil
+		},
+		registerMCPTools: func(context.Context, *tools.Registry, config.MCPConfig, mcp.RegisterOptions) (mcpToolRuntime, error) {
+			return noopMCPRuntime{}, nil
+		},
+		runTUI: func(ctx context.Context, options tui.Options) int {
+			launched = true
+			launchedOptions = options
+			return 0
+		},
+	})
+
+	if exitCode != 0 {
+		t.Fatalf("exit code = %d, want 0 (setup TUI launched), stderr=%q", exitCode, stderr.String())
+	}
+	if !launched {
+		t.Fatal("TUI was not launched; expected fallback into setup instead of a fatal error")
+	}
+	if !launchedOptions.Setup.Visible || !launchedOptions.Setup.Required {
+		t.Fatalf("Setup = %#v, want visible required setup", launchedOptions.Setup)
+	}
+	if launchedOptions.Provider != nil {
+		t.Fatalf("Provider = %#v, want nil for no-provider fallback", launchedOptions.Provider)
+	}
+}
+
+func TestRunNoArgsFallsBackToUsableProviderWhenNoneMarkedActive(t *testing.T) {
+	var stdout bytes.Buffer
+	var stderr bytes.Buffer
+	cwd := t.TempDir()
+	setCLIUserConfigRoot(t)
+	userConfigPath := filepath.Join(t.TempDir(), "zero", "config.json")
+	var launchedOptions tui.Options
+	launched := false
+	var providerProfile config.ProviderProfile
+	fake := &cliFakeProvider{}
+
+	usable := config.ProviderProfile{
+		Name:         "work",
+		ProviderKind: config.ProviderKindOpenAI,
+		BaseURL:      config.OpenAIBaseURL,
+		APIKey:       "sk-test",
+		Model:        "gpt-test",
+	}
+
+	exitCode := runWithDeps([]string{}, &stdout, &stderr, appDeps{
+		getwd: func() (string, error) {
+			return cwd, nil
+		},
+		resolveConfig: func(workspaceRoot string, overrides config.Overrides) (config.ResolvedConfig, error) {
+			// config.json has providers configured but none marked active (e.g. a
+			// blank/stale activeProvider field) — Resolve returns the successfully
+			// normalized providers list alongside ErrNoActiveProvider.
+			return config.ResolvedConfig{Providers: []config.ProviderProfile{usable}},
+				fmt.Errorf("%w: active provider %q not found", config.ErrNoActiveProvider, "")
+		},
+		newProvider: func(profile config.ProviderProfile) (zeroruntime.Provider, error) {
+			providerProfile = profile
+			return fake, nil
+		},
+		userConfigPath: func() (string, error) {
+			return userConfigPath, nil
+		},
+		registerMCPTools: func(context.Context, *tools.Registry, config.MCPConfig, mcp.RegisterOptions) (mcpToolRuntime, error) {
+			return noopMCPRuntime{}, nil
+		},
+		runTUI: func(ctx context.Context, options tui.Options) int {
+			launched = true
+			launchedOptions = options
+			return 0
+		},
+	})
+
+	if exitCode != 0 {
+		t.Fatalf("exit code = %d, want 0, stderr=%q", exitCode, stderr.String())
+	}
+	if !launched {
+		t.Fatal("TUI was not launched")
+	}
+	if launchedOptions.Setup.Visible || launchedOptions.Setup.Required {
+		t.Fatalf("Setup = %#v, want no setup wizard — a usable saved provider exists", launchedOptions.Setup)
+	}
+	if providerProfile.Name != "work" {
+		t.Fatalf("provider used = %q, want fallback to the usable saved provider %q", providerProfile.Name, "work")
+	}
+}
+
+func TestRunNoArgsFailsWhenResolveErrorIsNotProviderRelated(t *testing.T) {
+	var stdout bytes.Buffer
+	var stderr bytes.Buffer
+	cwd := t.TempDir()
+	setCLIUserConfigRoot(t)
+
+	exitCode := runWithDeps([]string{}, &stdout, &stderr, appDeps{
+		getwd: func() (string, error) {
+			return cwd, nil
+		},
+		resolveConfig: func(workspaceRoot string, overrides config.Overrides) (config.ResolvedConfig, error) {
+			return config.ResolvedConfig{}, fmt.Errorf("invalid config JSON")
+		},
+		runTUI: func(ctx context.Context, options tui.Options) int {
+			t.Fatal("TUI must not launch on a non-provider resolve error")
+			return 0
+		},
+	})
+
+	if exitCode == 0 {
+		t.Fatal("exit code = 0, want non-zero for fatal config error")
+	}
+	if !strings.Contains(stderr.String(), "invalid config JSON") {
+		t.Fatalf("stderr = %q, want the underlying config error", stderr.String())
+	}
+}
+
 func TestRunNoArgsLaunchesTUIWithMCPState(t *testing.T) {
 	var stdout bytes.Buffer
 	var stderr bytes.Buffer
@@ -1192,6 +1328,153 @@ func TestRunUpdateReportsUpToDate(t *testing.T) {
 	}
 }
 
+func TestRunUpdateApplyTextAndJSON(t *testing.T) {
+	applyResult := update.ApplyResult{
+		Result: update.Result{
+			CurrentVersion:  "dev",
+			LatestVersion:   "0.2.0",
+			UpdateAvailable: true,
+		},
+		Applied:       true,
+		InstallMethod: update.InstallMethodStandalone,
+		BinaryPath:    "/usr/local/bin/zero",
+		Message:       "updated to 0.2.0",
+	}
+	var gotOptions update.Options
+	deps := appDeps{
+		applyUpdate: func(_ context.Context, options update.Options) (update.ApplyResult, error) {
+			gotOptions = options
+			return applyResult, nil
+		},
+		checkUpdate: func(context.Context, update.Options) (update.Result, error) {
+			t.Fatal("checkUpdate should not run for --apply")
+			return update.Result{}, nil
+		},
+	}
+
+	var stdout bytes.Buffer
+	var stderr bytes.Buffer
+	exitCode := runWithDeps([]string{"update", "--apply"}, &stdout, &stderr, deps)
+	if exitCode != exitSuccess {
+		t.Fatalf("expected exit code %d, got %d: %s", exitSuccess, exitCode, stderr.String())
+	}
+	if !strings.Contains(stdout.String(), "updated to 0.2.0") {
+		t.Fatalf("unexpected apply text: %q", stdout.String())
+	}
+	if gotOptions.CurrentVersion != "dev" {
+		t.Fatalf("unexpected options passed to applyUpdate: %#v", gotOptions)
+	}
+
+	stdout.Reset()
+	stderr.Reset()
+	exitCode = runWithDeps([]string{"update", "--apply", "--json"}, &stdout, &stderr, deps)
+	if exitCode != exitSuccess {
+		t.Fatalf("expected exit code %d, got %d: %s", exitSuccess, exitCode, stderr.String())
+	}
+	var payload struct {
+		Applied       bool   `json:"applied"`
+		InstallMethod string `json:"installMethod"`
+		BinaryPath    string `json:"binaryPath"`
+	}
+	if err := json.Unmarshal(stdout.Bytes(), &payload); err != nil {
+		t.Fatalf("apply JSON did not decode: %v\n%s", err, stdout.String())
+	}
+	if !payload.Applied || payload.InstallMethod != string(update.InstallMethodStandalone) || payload.BinaryPath != applyResult.BinaryPath {
+		t.Fatalf("unexpected apply JSON: %#v", payload)
+	}
+}
+
+func TestRunUpgradeDefaultsToApply(t *testing.T) {
+	var applyCalled bool
+	deps := appDeps{
+		applyUpdate: func(context.Context, update.Options) (update.ApplyResult, error) {
+			applyCalled = true
+			return update.ApplyResult{Message: "already up to date"}, nil
+		},
+		checkUpdate: func(context.Context, update.Options) (update.Result, error) {
+			t.Fatal("checkUpdate should not run for `zero upgrade`")
+			return update.Result{}, nil
+		},
+	}
+
+	var stdout bytes.Buffer
+	var stderr bytes.Buffer
+	exitCode := runWithDeps([]string{"upgrade"}, &stdout, &stderr, deps)
+	if exitCode != exitSuccess {
+		t.Fatalf("expected exit code %d, got %d: %s", exitSuccess, exitCode, stderr.String())
+	}
+	if !applyCalled {
+		t.Fatal("expected `zero upgrade` to call applyUpdate")
+	}
+}
+
+func TestRunUpdateRejectsCheckAndApplyTogether(t *testing.T) {
+	var stdout bytes.Buffer
+	var stderr bytes.Buffer
+
+	exitCode := runWithDeps([]string{"update", "--check", "--apply"}, &stdout, &stderr, appDeps{
+		checkUpdate: func(context.Context, update.Options) (update.Result, error) {
+			t.Fatal("checkUpdate should not run")
+			return update.Result{}, nil
+		},
+		applyUpdate: func(context.Context, update.Options) (update.ApplyResult, error) {
+			t.Fatal("applyUpdate should not run")
+			return update.ApplyResult{}, nil
+		},
+	})
+
+	if exitCode != exitUsage {
+		t.Fatalf("expected usage exit code, got %d", exitCode)
+	}
+	if got := stderr.String(); !strings.Contains(got, "only one of --check or --apply") {
+		t.Fatalf("expected --check/--apply usage error, got %q", got)
+	}
+}
+
+func TestRunUpdateRejectsTargetWithApply(t *testing.T) {
+	for _, args := range [][]string{
+		{"update", "--apply", "--target", "linux-arm64"},
+		{"upgrade", "--target", "linux-arm64"},
+	} {
+		t.Run(strings.Join(args, " "), func(t *testing.T) {
+			var stdout bytes.Buffer
+			var stderr bytes.Buffer
+
+			exitCode := runWithDeps(args, &stdout, &stderr, appDeps{
+				applyUpdate: func(context.Context, update.Options) (update.ApplyResult, error) {
+					t.Fatal("applyUpdate should not run when --target is combined with --apply")
+					return update.ApplyResult{}, nil
+				},
+			})
+
+			if exitCode != exitUsage {
+				t.Fatalf("expected usage exit code, got %d", exitCode)
+			}
+			if got := stderr.String(); !strings.Contains(got, "--target cannot be combined with --apply") {
+				t.Fatalf("expected --target/--apply usage error, got %q", got)
+			}
+		})
+	}
+}
+
+func TestRunUpdateReportsApplyError(t *testing.T) {
+	var stdout bytes.Buffer
+	var stderr bytes.Buffer
+
+	exitCode := runWithDeps([]string{"update", "--apply"}, &stdout, &stderr, appDeps{
+		applyUpdate: func(context.Context, update.Options) (update.ApplyResult, error) {
+			return update.ApplyResult{}, errors.New("download failed")
+		},
+	})
+
+	if exitCode == exitSuccess {
+		t.Fatalf("expected non-success exit code, got %d", exitCode)
+	}
+	if got := stderr.String(); !strings.Contains(got, "download failed") {
+		t.Fatalf("expected apply error, got %q", got)
+	}
+}
+
 func assertHelpOutput(t *testing.T, args []string) {
 	t.Helper()
 
@@ -1397,7 +1680,79 @@ func TestRunThemeFlagRejectsBadValue(t *testing.T) {
 	if launched {
 		t.Fatal("the TUI must not launch on a bad --theme value")
 	}
-	if !strings.Contains(stderr.String(), "--theme must be auto, dark, or light") {
+	if !strings.Contains(stderr.String(), "--theme must be auto or a registered theme name") {
 		t.Fatalf("expected a clear --theme error, got %q", stderr.String())
 	}
 }
+
+// A custom endpoint saved without a model previously bricked bare `zero` and
+// `zero setup` — the exact commands that could have fixed it (the resolve
+// error escaped before the wizard could open). The interactive TUI now treats
+// the requires-model failure as "needs onboarding", same as a missing active
+// provider. The error comes from a REAL Resolve over a real config file, so
+// this exercises the production sentinel wrapping, not a hand-built error.
+func TestRunNoArgsEntersSetupWhenActiveProviderMissesModel(t *testing.T) {
+	var stdout bytes.Buffer
+	var stderr bytes.Buffer
+	cwd := t.TempDir()
+	setCLIUserConfigRoot(t)
+	brokenConfig := filepath.Join(t.TempDir(), "config.json")
+	if err := os.WriteFile(brokenConfig, []byte(`{
+		"activeProvider": "gw",
+		"providers": [{"name": "gw", "provider_kind": "openai-compatible", "baseURL": "https://gw.example/v1", "apiKey": "sk-x"}]
+	}`), 0o600); err != nil {
+		t.Fatal(err)
+	}
+	launched := false
+	var launchedOptions tui.Options
+
+	exitCode := runWithDeps([]string{}, &stdout, &stderr, appDeps{
+		getwd: func() (string, error) { return cwd, nil },
+		resolveConfig: func(string, config.Overrides) (config.ResolvedConfig, error) {
+			return config.Resolve(config.ResolveOptions{UserConfigPath: brokenConfig, Env: map[string]string{}})
+		},
+		newProvider: func(config.ProviderProfile) (zeroruntime.Provider, error) {
+			t.Fatal("newProvider must not be called without a resolved provider")
+			return nil, nil
+		},
+		userConfigPath: func() (string, error) { return filepath.Join(t.TempDir(), "zero", "config.json"), nil },
+		registerMCPTools: func(context.Context, *tools.Registry, config.MCPConfig, mcp.RegisterOptions) (mcpToolRuntime, error) {
+			return noopMCPRuntime{}, nil
+		},
+		runTUI: func(_ context.Context, options tui.Options) int {
+			launched = true
+			launchedOptions = options
+			return 0
+		},
+	})
+
+	if exitCode != 0 {
+		t.Fatalf("exit code = %d, want 0 (setup TUI launched), stderr=%q", exitCode, stderr.String())
+	}
+	if !launched {
+		t.Fatal("TUI was not launched; a fixable requires-model config must fall into setup, not exit fatally")
+	}
+	if !launchedOptions.Setup.Visible || !launchedOptions.Setup.Required {
+		t.Fatalf("Setup = %#v, want visible required setup", launchedOptions.Setup)
+	}
+}
+
+// `zero login`/`zero logout` don't exist (it's `zero auth login`); first-run
+// users try them (reported in the wild), so the unknown-command error points at
+// the real command.
+func TestRunUnknownLoginSuggestsAuthLogin(t *testing.T) {
+	var stdout, stderr bytes.Buffer
+	exitCode := runWithDeps([]string{"login"}, &stdout, &stderr, appDeps{})
+	if exitCode != 2 {
+		t.Fatalf("exit = %d, want 2 (unknown command)", exitCode)
+	}
+	if !strings.Contains(stderr.String(), `"zero auth login"`) {
+		t.Fatalf("stderr = %q, want a zero auth login suggestion", stderr.String())
+	}
+	// An unrelated unknown command gets no misleading suggestion.
+	stderr.Reset()
+	runWithDeps([]string{"frobnicate"}, &stdout, &stderr, appDeps{})
+	if strings.Contains(stderr.String(), "did you mean") {
+		t.Fatalf("stderr = %q, unrelated commands must not get the auth hint", stderr.String())
+	}
+}

internal/cli/auth.go

@@ -283,6 +283,9 @@ func newAuthManager(deps appDeps, out io.Writer) (*oauth.Manager, error) {
 		// carries no token material. A real browser launch is intentionally not
 		// performed (the sandbox/headless contexts make printing the safer default).
 		OpenBrowser: func(string) error { return nil },
+		// `zero auth login <preset>` (e.g. xai) should resolve the baked-in preset
+		// without the operator exporting ZERO_OAUTH_ALLOW_PRESETS first.
+		AllowPresets: true,
 	})
 }
 
@@ -298,12 +301,27 @@ func runAuthLogin(args []string, stdout io.Writer, stderr io.Writer, deps appDep
 	if len(parsed.positional) != 1 {
 		return writeExecUsageError(stderr, "usage: zero auth login <provider> [--device] [--scope <scope>]")
 	}
+	provider := parsed.positional[0]
+	// ChatGPT (Codex) requires a fixed redirect_uri (http://localhost:1455/
+	// auth/callback) and mandatory authorize params (id_token_add_organizations,
+	// codex_cli_simplified_flow, originator) that the generic loopback flow
+	// cannot supply. Route it to the dedicated ChatGPT login so
+	// `zero auth login chatgpt` behaves identically to `zero auth chatgpt`.
+	if strings.EqualFold(provider, "chatgpt") {
+		if parsed.device {
+			return writeExecUsageError(stderr, "ChatGPT login does not support --device (it is loopback-only)")
+		}
+		if len(parsed.scopes) > 0 {
+			return writeExecUsageError(stderr, "ChatGPT login does not support --scope (the required scopes are fixed by the Codex client registration)")
+		}
+		return runAuthChatGPT(nil, stdout, stderr, deps)
+	}
 	manager, err := newAuthManager(deps, stdout)
 	if err != nil {
 		return writeAppError(stderr, redaction.ErrorMessage(err, redaction.Options{}), exitCrash)
 	}
 	status, err := manager.Login(context.Background(), oauth.LoginOptions{
-		Provider:    parsed.positional[0],
+		Provider:    provider,
 		Device:      parsed.device,
 		ExtraScopes: parsed.scopes,
 	})
@@ -486,8 +504,9 @@ Commands:
 A provider is any OAuth 2.0 / OIDC server. "openrouter" ('zero auth openrouter')
 works out of the box. "xai" ('zero auth login xai') uses a built-in preset that is
 off by default — enable it with ZERO_OAUTH_ALLOW_PRESETS=1, or set the
-ZERO_OAUTH_XAI_* vars yourself. Any preset field is overridable via the env vars
-below. For a custom provider named <name>, set:
+ZERO_OAUTH_XAI_* vars yourself. "chatgpt" ('zero auth login chatgpt' or
+'zero auth chatgpt') uses a fixed-port loopback flow against the Codex backend.
+Any preset field is overridable via the env vars below. For a custom provider named <name>, set:
   ZERO_OAUTH_<NAME>_CLIENT_ID       (required)
   ZERO_OAUTH_<NAME>_CLIENT_SECRET   (optional)
   ZERO_OAUTH_<NAME>_AUTHORIZE_URL   ZERO_OAUTH_<NAME>_TOKEN_URL

internal/cli/auth_propagation_test.go

@@ -0,0 +1,180 @@
+package cli
+
+import (
+	"bytes"
+	"errors"
+	"os"
+	"path/filepath"
+	"testing"
+
+	"github.com/Gitlawb/zero/internal/config"
+	"github.com/Gitlawb/zero/internal/zeroruntime"
+)
+
+// seedStoredProviderKey writes key into a credential store rooted beside the
+// returned config path, and returns deps fields wired at that path — the same
+// layout applyStoredProviderKey resolves through in production.
+func seedStoredProviderKey(t *testing.T, providerName, key string) (userConfigPath string) {
+	t.Helper()
+	t.Setenv("ZERO_CRED_STORAGE", "encrypted-file") // never touch the real OS keychain in tests
+	dir := t.TempDir()
+	store, err := config.ProviderKeyStoreAt(dir)
+	if err != nil {
+		t.Fatalf("ProviderKeyStoreAt: %v", err)
+	}
+	if err := store.Set(providerName, key); err != nil {
+		t.Fatalf("seed stored key: %v", err)
+	}
+	return filepath.Join(dir, "config.json")
+}
+
+// applyStoredProviderKeyAt must fill APIKey from the credential store for an
+// apiKeyStored profile — the single central helper the newProvider wrapper uses.
+func TestApplyStoredProviderKeyAtFillsFromCredstore(t *testing.T) {
+	configPath := seedStoredProviderKey(t, "echo", "sk-stored-test")
+	ucp := func() (string, error) { return configPath, nil }
+
+	stored := applyStoredProviderKeyAt(config.ProviderProfile{Name: "echo", APIKeyStored: true}, ucp)
+	if stored.APIKey != "sk-stored-test" {
+		t.Fatalf("APIKey = %q, want the stored key", stored.APIKey)
+	}
+
+	// A profile that did NOT opt into stored-key auth must be left unchanged —
+	// a stale store entry must not silently reactivate credentials.
+	plain := applyStoredProviderKeyAt(config.ProviderProfile{Name: "echo"}, ucp)
+	if plain.APIKey != "" {
+		t.Fatalf("non-stored profile APIKey = %q, want empty", plain.APIKey)
+	}
+}
+
+// fillAppDeps wraps newProvider so EVERY surface that builds a runtime provider
+// (buildProvider, the ACP builder — which is just deps.newProvider — and exec's
+// escalation switcher) gets the stored key applied. Passing the raw resolved
+// profile previously sent unauthenticated requests for apiKeyStored profiles,
+// the default onboarding outcome. Testing the wrap covers all those surfaces.
+func TestFillAppDepsWrapsNewProviderWithStoredKey(t *testing.T) {
+	configPath := seedStoredProviderKey(t, "echo", "sk-stored-wrap")
+
+	var captured config.ProviderProfile
+	deps := fillAppDeps(appDeps{
+		userConfigPath: func() (string, error) { return configPath, nil },
+		newProvider: func(profile config.ProviderProfile) (zeroruntime.Provider, error) {
+			captured = profile
+			return nil, nil // only the captured profile matters here
+		},
+	})
+
+	if _, err := deps.newProvider(config.ProviderProfile{Name: "echo", APIKeyStored: true}); err != nil {
+		t.Fatalf("wrapped newProvider: %v", err)
+	}
+	if captured.APIKey != "sk-stored-wrap" {
+		t.Fatalf("wrapped newProvider built with APIKey = %q, want the stored key (unauthenticated regression)", captured.APIKey)
+	}
+}
+
+// buildProvider (the TUI/exec STARTUP construction site) must export
+// ZERO_PROVIDER so children spawned at any point in the run are pinned to the
+// parent's provider from launch — not only after an in-session switch. Without
+// this, a provider switch persisted by another zero process mid-session moves
+// new children onto a different provider (and credentials) than the parent.
+func TestBuildProviderExportsActiveProviderEnv(t *testing.T) {
+	t.Setenv(config.ActiveProviderEnv, "stale-from-elsewhere")
+	// The exporter is injected (defaultAppDeps wires config.SetActiveProviderEnv;
+	// fillAppDeps deliberately leaves it nil so ordinary tests never mutate the
+	// process env). Inject the real one here to assert the actual env effect.
+	deps := appDeps{
+		exportActiveProvider: config.SetActiveProviderEnv,
+		newProvider: func(config.ProviderProfile) (zeroruntime.Provider, error) {
+			return nil, nil
+		},
+	}
+
+	resolved := config.ResolvedConfig{Provider: config.ProviderProfile{Name: "pinned", ProviderKind: config.ProviderKindAnthropic, Model: "m"}}
+	if _, err := buildProvider(resolved, deps); err != nil {
+		t.Fatalf("buildProvider: %v", err)
+	}
+	if got := os.Getenv(config.ActiveProviderEnv); got != "pinned" {
+		t.Fatalf("%s = %q after startup build, want %q (children would spawn on the stale provider)", config.ActiveProviderEnv, got, "pinned")
+	}
+
+	// A FAILED build must not move the pin: the process never committed to the
+	// new provider, so children must keep resolving whatever was in effect.
+	t.Setenv(config.ActiveProviderEnv, "still-current")
+	failing := appDeps{
+		exportActiveProvider: config.SetActiveProviderEnv,
+		newProvider: func(config.ProviderProfile) (zeroruntime.Provider, error) {
+			return nil, errors.New("boom")
+		},
+	}
+	if _, err := buildProvider(resolved, failing); err == nil {
+		t.Fatal("expected build error")
+	}
+	if got := os.Getenv(config.ActiveProviderEnv); got != "still-current" {
+		t.Fatalf("%s = %q after failed build, want it untouched", config.ActiveProviderEnv, got)
+	}
+
+	// No provider profile at all → no provider, no pin change.
+	if _, err := buildProvider(config.ResolvedConfig{}, deps); err != nil {
+		t.Fatalf("buildProvider(empty): %v", err)
+	}
+	if got := os.Getenv(config.ActiveProviderEnv); got != "still-current" {
+		t.Fatalf("%s = %q after profile-less build, want it untouched", config.ActiveProviderEnv, got)
+	}
+}
+
+// A mid-run --allow-escalation model switch previously rebuilt the provider
+// from the PURE resolved profile (the stored key lives only in buildProvider's
+// local copy), so the escalated provider was keyless and the next completion
+// call 401'd mid-run. Every profile handed to newProvider on this path must
+// carry the stored key.
+func TestRunExecEscalationSwitchKeepsStoredKey(t *testing.T) {
+	dataHome := t.TempDir()
+	t.Setenv("XDG_DATA_HOME", dataHome)
+	cwd := t.TempDir()
+
+	if _, ok := mustUpgradeTarget(t, "claude-haiku-4.5"); !ok {
+		t.Skip("registry has no upgrade target for claude-haiku-4.5")
+	}
+
+	configPath := seedStoredProviderKey(t, "echo", "sk-stored-escalation")
+
+	var builtProfiles []config.ProviderProfile
+	var stdout, stderr bytes.Buffer
+	exitCode := runWithDeps([]string{
+		"exec",
+		"--allow-escalation",
+		"--model", "claude-haiku-4.5",
+		"--init-session-id", "escalation_stored_key_run",
+		"escalate please",
+	}, &stdout, &stderr, appDeps{
+		getwd:          func() (string, error) { return cwd, nil },
+		userConfigPath: func() (string, error) { return configPath, nil },
+		resolveConfig: func(_ string, overrides config.Overrides) (config.ResolvedConfig, error) {
+			model := "claude-haiku-4.5"
+			if overrides.Provider.Model != "" {
+				model = overrides.Provider.Model
+			}
+			cfg := execResolvedConfig()
+			cfg.Provider.ProviderKind = config.ProviderKindAnthropic
+			cfg.Provider.Model = model
+			cfg.Provider.APIKeyStored = true // key lives ONLY in the credstore
+			cfg.MaxTurns = 3
+			return cfg, nil
+		},
+		newProvider: func(profile config.ProviderProfile) (zeroruntime.Provider, error) {
+			builtProfiles = append(builtProfiles, profile)
+			return &usageEmittingEscalatingProvider{escalate: len(builtProfiles) == 1}, nil
+		},
+	})
+	if exitCode != exitSuccess {
+		t.Fatalf("exit = %d, want success; stderr=%s", exitCode, stderr.String())
+	}
+	if len(builtProfiles) < 2 {
+		t.Fatalf("want at least 2 provider builds (initial + escalated), got %d", len(builtProfiles))
+	}
+	for i, profile := range builtProfiles {
+		if profile.APIKey != "sk-stored-escalation" {
+			t.Fatalf("provider build %d APIKey = %q, want the stored key (escalated build went out keyless)", i, profile.APIKey)
+		}
+	}
+}

internal/cli/auth_test.go

@@ -156,3 +156,44 @@ func TestRunAuthHelp(t *testing.T) {
 		}
 	}
 }
+
+// TestRunAuthLoginChatGPTRoutesToDedicatedFlow verifies `zero auth login
+// chatgpt` reaches the dedicated ChatGPT login (fixed-port loopback + mandatory
+// authorize params), not the generic manager path. The generic login accepts
+// --device, so a ChatGPT-specific rejection proves the routing took effect.
+// See issue #430: the generic path built a random-port 127.0.0.1 redirect_uri
+// without the required extra params, so OpenAI's authorize endpoint rejected it.
+func TestRunAuthLoginChatGPTRoutesToDedicatedFlow(t *testing.T) {
+	withAuthStore(t)
+	var stdout, stderr bytes.Buffer
+	if code := runWithDeps([]string{"auth", "login", "chatgpt", "--device"}, &stdout, &stderr, appDeps{}); code == exitSuccess {
+		t.Fatal("auth login chatgpt --device should be rejected (ChatGPT is loopback-only)")
+	}
+	if !strings.Contains(stderr.String(), "ChatGPT login does not support --device") {
+		t.Fatalf("stderr = %q, want the ChatGPT-specific --device rejection", stderr.String())
+	}
+	// Case-insensitive provider name should still route.
+	stdout.Reset()
+	stderr.Reset()
+	if code := runWithDeps([]string{"auth", "login", "ChatGPT", "--device"}, &stdout, &stderr, appDeps{}); code == exitSuccess {
+		t.Fatal("auth login ChatGPT --device should be rejected")
+	}
+	if !strings.Contains(stderr.String(), "ChatGPT login does not support --device") {
+		t.Fatalf("stderr = %q, want the ChatGPT-specific rejection (case-insensitive)", stderr.String())
+	}
+}
+
+// TestRunAuthLoginChatGPTRejectsScope mirrors the --device rejection: --scope
+// must not be silently dropped on the ChatGPT path. The Codex client
+// registration pins a fixed scope set (incl. api.connectors.*), so custom
+// scopes are rejected up front rather than plumbed through.
+func TestRunAuthLoginChatGPTRejectsScope(t *testing.T) {
+	withAuthStore(t)
+	var stdout, stderr bytes.Buffer
+	if code := runWithDeps([]string{"auth", "login", "chatgpt", "--scope", "custom-scope"}, &stdout, &stderr, appDeps{}); code == exitSuccess {
+		t.Fatal("auth login chatgpt --scope should be rejected")
+	}
+	if !strings.Contains(stderr.String(), "ChatGPT login does not support --scope") {
+		t.Fatalf("stderr = %q, want the ChatGPT-specific --scope rejection", stderr.String())
+	}
+}

internal/cli/command_center.go

@@ -81,6 +81,9 @@ func runProviders(args []string, stdout io.Writer, stderr io.Writer, deps appDep
 	if command == "detect" {
 		return runProvidersDetect(args, stdout, stderr, deps)
 	}
+	if command == "models" {
+		return runProvidersModels(args, stdout, stderr, deps)
+	}
 	if command != "list" && command != "current" && command != "catalog" {
 		return writeExecUsageError(stderr, fmt.Sprintf("unknown providers command %q", command))
 	}
@@ -465,9 +468,11 @@ func writeProvidersHelp(w io.Writer) error {
   zero providers use <name> [flags]
   zero providers setup <catalog-id> [flags]
   zero providers detect [flags]
+  zero providers models [name] [flags]
 
 Inspects resolved provider profiles and provider catalog descriptors without printing secrets.
 Detect probes for running local runtimes (Ollama, LM Studio) and prints adopt commands plus per-provider next steps.
+Models probes a provider's live model-listing endpoint (e.g. an OpenAI-compatible /v1/models) and lists the models it serves — including custom OpenAI-/Anthropic-compatible endpoints — so a self-hosted provider needs no per-model config.
 
 Flags:
       --json                    Print JSON summary

internal/cli/exec.go

@@ -12,6 +12,7 @@ import (
 
 	"github.com/Gitlawb/zero/internal/agent"
 	"github.com/Gitlawb/zero/internal/config"
+	"github.com/Gitlawb/zero/internal/errhint"
 	"github.com/Gitlawb/zero/internal/imageinput"
 	"github.com/Gitlawb/zero/internal/lsp"
 	"github.com/Gitlawb/zero/internal/modelregistry"
@@ -135,6 +136,11 @@ func runExec(args []string, stdout io.Writer, stderr io.Writer, deps appDeps) in
 		return exitSuccess
 	}
 
+	// Refresh the models.dev pricing/limits cache in the background when stale;
+	// the overlay is read at registry construction from the cache file, so this
+	// benefits the next run and never blocks or fails this one.
+	go func() { _ = modelregistry.RefreshModelsDevCache(context.Background()) }()
+
 	// A mode seeds model/effort/max-turns/tool filters as a preset. Expand it up
 	// front — before tool-filter validation and the --list-tools branch — so a
 	// mode-injected tool filter is validated and reflected in --list-tools, and a
@@ -365,6 +371,9 @@ func runExec(args []string, stdout io.Writer, stderr io.Writer, deps appDeps) in
 	var modelSwitcher func(context.Context, string) (agent.Provider, error)
 	if options.allowEscalation {
 		modelSwitcher = func(_ context.Context, modelID string) (agent.Provider, error) {
+			// deps.newProvider is wrapped (fillAppDeps) to apply the stored key, so
+			// the escalated provider is authenticated even though resolved.Provider
+			// is the pure profile — no per-site key handling here.
 			switchedProfile := resolved.Provider
 			switchedProfile.Model = modelID
 			switchedProvider, err := deps.newProvider(switchedProfile)
@@ -496,15 +505,17 @@ func runExec(args []string, stdout io.Writer, stderr io.Writer, deps appDeps) in
 	// by default the corrector is nil, leaving the agent loop byte-identical. When
 	// on we verify with both the workspace test plan and LSP diagnostics over the
 	// changed files; the autonomy gate inside the corrector still decides whether
-	// failures auto-fix or just report. lspShutdown tears down any language-server
-	// sessions the LSP half spawned (no-op when self-correct is off).
-	selfCorrector, lspShutdown := newExecSelfCorrector(options.selfCorrect, workspaceRoot, options.autonomy)
+	// failures auto-fix or just report. fileDiagnostics (always on, lazy) gives
+	// edit_file/write_file inline error diagnostics for the file they just wrote.
+	// lspShutdown tears down any language-server sessions either half spawned.
+	selfCorrector, fileDiagnostics, lspShutdown := newExecSelfCorrector(options.selfCorrect, workspaceRoot, options.autonomy)
 	defer lspShutdown()
 	result, err := agent.Run(runCtx, agentPrompt, provider, agent.Options{
 		MaxTurns:         resolved.MaxTurns,
 		ContextWindow:    resolveAgentContextWindow(runCtx, modelRegistry, resolved.Provider),
 		DeferThreshold:   effectiveDeferThreshold,
 		Specialists:      specialistRuntime.specialistInfos(),
+		Skills:           pluginActivation.skillInfos(deps.skillsDir()),
 		SessionID:        preparedSession.Session.SessionID,
 		CallingSessionID: options.callingSessionID,
 		CallingToolUseID: options.callingToolUseID,
@@ -521,6 +532,7 @@ func runExec(args []string, stdout io.Writer, stderr io.Writer, deps appDeps) in
 		PermissionMode:   permissionMode,
 		Autonomy:         options.autonomy,
 		SelfCorrect:      selfCorrector,
+		FileDiagnostics:  fileDiagnostics,
 		// Headless exec: don't accept a no-tool-call turn as "done" while work
 		// clearly remains (pending plan items / a mid-step continuation cue) —
 		// nudge to continue, and finalize as INCOMPLETE rather than false success
@@ -532,6 +544,7 @@ func runExec(args []string, stdout io.Writer, stderr io.Writer, deps appDeps) in
 		EnabledTools:            options.enabledTools,
 		DisabledTools:           options.disabledTools,
 		OnText:                  writer.text,
+		OnReasoning:             writer.reasoning,
 		OnToolCall: func(call agent.ToolCall) {
 			writer.toolCall(call, registry)
 			sessionRecorder.append(sessions.EventToolCall, map[string]any{
@@ -696,23 +709,29 @@ func runExec(args []string, stdout io.Writer, stderr io.Writer, deps appDeps) in
 // unless a changed file's language actually has one installed on PATH. The
 // returned cleanup shuts that manager down (terminating any spawned server
 // sessions); it is a no-op when self-correct is off.
-func newExecSelfCorrector(enabled bool, workspaceRoot string, autonomy string) (*agent.SelfCorrector, func()) {
-	if !enabled {
-		return nil, func() {}
-	}
+func newExecSelfCorrector(enabled bool, workspaceRoot string, autonomy string) (*agent.SelfCorrector, func(context.Context, string) string, func()) {
+	// The manager is created regardless of --self-correct: it also backs the
+	// always-on background post-edit diagnostics (agent.NewFileDiagnostics,
+	// collected off the tool-call path by the loop), and it stays lazy — no
+	// language server is spawned unless an edited file's language actually has
+	// one installed on PATH.
 	manager := lsp.NewManager(workspaceRoot)
+	cleanup := func() {
+		ctx, cancel := context.WithTimeout(context.Background(), 5*time.Second)
+		defer cancel()
+		_ = manager.Shutdown(ctx)
+	}
+	fileDiagnostics := agent.NewFileDiagnostics(manager, workspaceRoot)
+	if !enabled {
+		return nil, fileDiagnostics, cleanup
+	}
 	corrector := agent.NewSelfCorrector(workspaceRoot, agent.NewLSPDiagnosticsChecker(manager), agent.NewProjectVerifier(workspaceRoot), agent.SelfCorrectConfig{
 		Enabled:      true,
 		IncludeTests: true,
 		IncludeLSP:   true,
 		Autonomy:     autonomy,
 	})
-	cleanup := func() {
-		ctx, cancel := context.WithTimeout(context.Background(), 5*time.Second)
-		defer cancel()
-		_ = manager.Shutdown(ctx)
-	}
-	return corrector, cleanup
+	return corrector, fileDiagnostics, cleanup
 }
 
 func deferredEligibleCount(registry *tools.Registry, permissionMode agent.PermissionMode, enabledTools []string, disabledTools []string) int {
@@ -945,6 +964,14 @@ func writeExecProviderError(stdout io.Writer, stderr io.Writer, format execOutpu
 	if _, err := fmt.Fprintf(stderr, "[zero] %s\n", message); err != nil {
 		return exitCrash
 	}
+	// Append a one-line next step for recognized provider failures (auth /
+	// rate-limit / connectivity / …). The classifier gates on a provider-origin
+	// marker, so non-provider codes (sandbox_error, mcp_error) never draw a hint.
+	if hint := errhint.CLIHint(errors.New(message)); hint != "" {
+		if _, err := fmt.Fprintf(stderr, "[zero] %s\n", hint); err != nil {
+			return exitCrash
+		}
+	}
 	return exitProvider
 }
 

internal/cli/exec_protocol_test.go

@@ -529,6 +529,45 @@ func TestRunExecStreamJSONRunStartUsesResolvedAPIModel(t *testing.T) {
 	}
 }
 
+func TestRunExecStreamJSONEmitsReasoningEvents(t *testing.T) {
+	t.Setenv("XDG_DATA_HOME", t.TempDir())
+	cwd := t.TempDir()
+	var stdout bytes.Buffer
+	var stderr bytes.Buffer
+
+	exitCode := runWithDeps([]string{"exec", "--output-format", "stream-json", "think then answer"}, &stdout, &stderr, appDeps{
+		getwd: func() (string, error) {
+			return cwd, nil
+		},
+		resolveConfig: func(_ string, _ config.Overrides) (config.ResolvedConfig, error) {
+			return execResolvedConfig(), nil
+		},
+		newProvider: func(config.ProviderProfile) (zeroruntime.Provider, error) {
+			return reasoningExecProvider{}, nil
+		},
+	})
+
+	if exitCode != exitSuccess {
+		t.Fatalf("expected exit code %d, got %d: %s", exitSuccess, exitCode, stderr.String())
+	}
+	events := decodeJSONLines(t, stdout.String())
+	reasoningIdx, reasoning := findJSONEventIndex(t, events, "reasoning")
+	if reasoning["delta"] != "Thinking. " {
+		t.Fatalf("unexpected reasoning event: %#v", reasoning)
+	}
+	textIdx, text := findJSONEventIndex(t, events, "text")
+	if text["delta"] != "done" {
+		t.Fatalf("unexpected text event: %#v", text)
+	}
+	if reasoningIdx >= textIdx {
+		t.Fatalf("expected reasoning event before text event, got indices %d and %d", reasoningIdx, textIdx)
+	}
+	final := findJSONEvent(t, events, "final")
+	if final["text"] != "done" {
+		t.Fatalf("reasoning must not be folded into final answer: %#v", final)
+	}
+}
+
 func TestExecEventWriterTruncatesStreamJSONToolResults(t *testing.T) {
 	var stdout bytes.Buffer
 	writer := execEventWriter{
@@ -751,6 +790,17 @@ type toolCallingExecProvider struct {
 	answer     string
 }
 
+type reasoningExecProvider struct{}
+
+func (reasoningExecProvider) StreamCompletion(context.Context, zeroruntime.CompletionRequest) (<-chan zeroruntime.StreamEvent, error) {
+	ch := make(chan zeroruntime.StreamEvent, 3)
+	ch <- zeroruntime.StreamEvent{Type: zeroruntime.StreamEventReasoning, Content: "Thinking. "}
+	ch <- zeroruntime.StreamEvent{Type: zeroruntime.StreamEventText, Content: "done"}
+	ch <- zeroruntime.StreamEvent{Type: zeroruntime.StreamEventDone}
+	close(ch)
+	return ch, nil
+}
+
 func (provider toolCallingExecProvider) StreamCompletion(ctx context.Context, request zeroruntime.CompletionRequest) (<-chan zeroruntime.StreamEvent, error) {
 	for _, message := range request.Messages {
 		if message.Role == zeroruntime.MessageRoleTool {
@@ -777,13 +827,19 @@ func (provider toolCallingExecProvider) StreamCompletion(ctx context.Context, re
 
 func findJSONEvent(t *testing.T, events []map[string]any, eventType string) map[string]any {
 	t.Helper()
-	for _, event := range events {
+	_, event := findJSONEventIndex(t, events, eventType)
+	return event
+}
+
+func findJSONEventIndex(t *testing.T, events []map[string]any, eventType string) (int, map[string]any) {
+	t.Helper()
+	for idx, event := range events {
 		if event["type"] == eventType {
-			return event
+			return idx, event
 		}
 	}
 	t.Fatalf("event %q not found in %#v", eventType, events)
-	return nil
+	return -1, nil
 }
 
 func findSessionEvent(t *testing.T, events []sessions.Event, eventType sessions.EventType) sessions.Event {

internal/cli/exec_provider_hint_test.go

@@ -0,0 +1,50 @@
+package cli
+
+import (
+	"bytes"
+	"strings"
+	"testing"
+)
+
+// A recognized provider error in text mode prints the raw message plus a one-line
+// actionable hint; a non-provider error (or JSON mode) prints no hint.
+func TestWriteExecProviderErrorAppendsHint(t *testing.T) {
+	t.Run("provider auth error gets a hint", func(t *testing.T) {
+		var stdout, stderr bytes.Buffer
+		code := writeExecProviderError(&stdout, &stderr, execOutputText, "provider_error",
+			"auth error: your API key is missing or invalid")
+		if code != exitProvider {
+			t.Fatalf("exit code = %d, want exitProvider", code)
+		}
+		out := stderr.String()
+		if !strings.Contains(out, "auth error:") {
+			t.Fatalf("expected raw message, got %q", out)
+		}
+		if !strings.Contains(out, "zero auth") {
+			t.Fatalf("expected an actionable hint referencing `zero auth`, got %q", out)
+		}
+	})
+
+	t.Run("non-provider error gets no hint", func(t *testing.T) {
+		var stdout, stderr bytes.Buffer
+		writeExecProviderError(&stdout, &stderr, execOutputText, "sandbox_error",
+			"sandbox setup failed: permission denied")
+		out := stderr.String()
+		// Exactly one "[zero]" line — no spurious hint attached to a local error.
+		if n := strings.Count(out, "[zero]"); n != 1 {
+			t.Fatalf("expected exactly one [zero] line for a non-provider error, got %d:\n%s", n, out)
+		}
+	})
+
+	t.Run("json mode never appends a hint line", func(t *testing.T) {
+		var stdout, stderr bytes.Buffer
+		writeExecProviderError(&stdout, &stderr, execOutputJSON, "provider_error",
+			"auth error: bad key")
+		if stderr.Len() != 0 {
+			t.Fatalf("json mode must not write to stderr, got %q", stderr.String())
+		}
+		if strings.Contains(stdout.String(), "zero auth") {
+			t.Fatalf("json mode must not inject a hint into the structured payload, got %q", stdout.String())
+		}
+	})
+}

internal/cli/exec_writer.go

@@ -80,6 +80,16 @@ func (writer *execEventWriter) text(delta string) {
 	writer.writeStdout(delta)
 }
 
+func (writer *execEventWriter) reasoning(delta string) {
+	if writer.format == execOutputJSON {
+		writer.writeJSON(map[string]any{"type": "reasoning", "delta": delta})
+		return
+	}
+	if writer.format == execOutputStreamJSON {
+		writer.writeStreamJSON(streamjson.Event{Type: streamjson.EventReasoning, RunID: writer.runID, Delta: delta})
+	}
+}
+
 func (writer *execEventWriter) toolCall(call agent.ToolCall, registry *tools.Registry) {
 	if writer.format == execOutputJSON {
 		writer.writeJSON(map[string]any{

internal/cli/oauth_credential_test.go

@@ -24,6 +24,7 @@ func TestProviderHasOAuthLogin(t *testing.T) {
 
 func TestSetupRequiredRecognizesOAuthLogin(t *testing.T) {
 	path := filepath.Join(t.TempDir(), "tok.json")
+	t.Setenv("ZERO_OAUTH_STORAGE", "file") // an inherited "keyring" would ignore the temp path and hit the OS keychain
 	t.Setenv("ZERO_OAUTH_TOKENS_PATH", path)
 	store, err := oauth.NewStore(oauth.StoreOptions{FilePath: path})
 	if err != nil {
@@ -45,3 +46,88 @@ func TestSetupRequiredRecognizesOAuthLogin(t *testing.T) {
 		t.Fatal("a keyless provider with no OAuth login must require onboarding")
 	}
 }
+
+// A profile renamed away from its catalog ID must still get a runtime OAuth
+// resolver when the login is stored under the catalog name — and a CASE-VARIANT
+// profile name must not drop the catalog-ID candidate (the store is
+// case-sensitive; a case-insensitive dedupe would swallow it).
+func TestOAuthResolverForProfileFallsBackToCatalogID(t *testing.T) {
+	path := filepath.Join(t.TempDir(), "tok.json")
+	t.Setenv("ZERO_OAUTH_STORAGE", "file") // an inherited "keyring" would ignore the temp path and hit the OS keychain
+	t.Setenv("ZERO_OAUTH_TOKENS_PATH", path)
+	store, err := oauth.NewStore(oauth.StoreOptions{FilePath: path})
+	if err != nil {
+		t.Fatalf("NewStore: %v", err)
+	}
+	if err := store.Save(oauth.ProviderKey("chatgpt"), oauth.Token{AccessToken: "tok", ExpiresAt: time.Now().Add(time.Hour)}); err != nil {
+		t.Fatalf("seed token: %v", err)
+	}
+
+	// A renamed profile resolves the catalog-ID login, and the returned key MUST be
+	// that login's key — the same key the Codex account-header resolver reads from,
+	// so bearer and account can never come from different logins.
+	if resolver, key := oauthLoginForProfile(config.ProviderProfile{Name: "codex", CatalogID: "chatgpt"}); resolver == nil {
+		t.Fatal("renamed profile with a catalog-ID login must get a resolver (unauthenticated-children regression)")
+	} else if key != oauth.ProviderKey("chatgpt") {
+		t.Fatalf("bound key = %q, want the chatgpt login key (bearer/account must share it)", key)
+	}
+	// Name differs from catalog ID only in CASE: the exact-case store key is
+	// "provider:chatgpt", so "chatgpt" must survive as a distinct candidate.
+	if resolver, key := oauthLoginForProfile(config.ProviderProfile{Name: "ChatGPT", CatalogID: "chatgpt"}); resolver == nil {
+		t.Fatal("case-variant profile name must still resolve the catalog-ID login (case-insensitive dedupe regression)")
+	} else if key != oauth.ProviderKey("chatgpt") {
+		t.Fatalf("case-variant bound key = %q, want the chatgpt login key", key)
+	}
+	if resolver, _ := oauthLoginForProfile(config.ProviderProfile{Name: "chatgpt", CatalogID: "chatgpt"}); resolver == nil {
+		t.Fatal("catalog-named profile must get a resolver")
+	}
+	if resolver, key := oauthLoginForProfile(config.ProviderProfile{Name: "openai", CatalogID: "openai"}); resolver != nil || key != "" {
+		t.Fatalf("a profile with no stored login must get no resolver and an empty key, got key=%q", key)
+	}
+}
+
+// The catalog-ID fallback must NOT reach a sibling profile's OAuth login when
+// this profile has its own key — withBearer would erase the key and silently
+// bill the model call under the sibling's account (a real cross-profile
+// override the fallback introduced). A profile WITH its own resolved key gets no
+// resolver even when a catalog-shared login exists.
+func TestOAuthResolverForProfileDoesNotOverrideExplicitKey(t *testing.T) {
+	path := filepath.Join(t.TempDir(), "tok.json")
+	t.Setenv("ZERO_OAUTH_STORAGE", "file") // an inherited "keyring" would ignore the temp path and hit the OS keychain
+	t.Setenv("ZERO_OAUTH_TOKENS_PATH", path)
+	store, err := oauth.NewStore(oauth.StoreOptions{FilePath: path})
+	if err != nil {
+		t.Fatalf("NewStore: %v", err)
+	}
+	// A personal login stored under the catalog name.
+	if err := store.Save(oauth.ProviderKey("anthropic"), oauth.Token{AccessToken: "personal-tok", ExpiresAt: time.Now().Add(time.Hour)}); err != nil {
+		t.Fatalf("seed token: %v", err)
+	}
+
+	// Work profile: same catalog, its OWN resolved API key, deliberately NOT the
+	// personal OAuth account. It must keep using its key (no resolver attached).
+	work := config.ProviderProfile{Name: "anthropic-work", CatalogID: "anthropic", APIKey: "sk-work-key"}
+	if resolver, key := oauthLoginForProfile(work); resolver != nil || key != "" {
+		t.Fatalf("a profile with its own API key must not get a catalog-shared OAuth resolver (cross-profile override), got key=%q", key)
+	}
+	// A raw auth-header credential must be protected the same way.
+	header := config.ProviderProfile{Name: "anthropic-hdr", CatalogID: "anthropic", AuthHeaderValue: "Bearer byo"}
+	if resolver, _ := oauthLoginForProfile(header); resolver != nil {
+		t.Fatal("a profile with its own auth header must not get a catalog-shared OAuth resolver")
+	}
+	// A same-NAME login must not override an explicit key either (the name
+	// candidate is gated on the profile being keyless, not just the catalog one).
+	if err := store.Save(oauth.ProviderKey("anthropic-work"), oauth.Token{AccessToken: "name-tok", ExpiresAt: time.Now().Add(time.Hour)}); err != nil {
+		t.Fatalf("seed name token: %v", err)
+	}
+	if resolver, _ := oauthLoginForProfile(work); resolver != nil {
+		t.Fatal("a profile with its own API key must not get an OAuth resolver even from a same-name login")
+	}
+	// Sanity: a keyless sibling on the same catalog DOES resolve the shared login.
+	keyless := config.ProviderProfile{Name: "anthropic-oauth", CatalogID: "anthropic"}
+	if resolver, key := oauthLoginForProfile(keyless); resolver == nil {
+		t.Fatal("a keyless profile should still resolve the catalog-shared login")
+	} else if key != oauth.ProviderKey("anthropic") {
+		t.Fatalf("keyless sibling bound key = %q, want the catalog login key", key)
+	}
+}

internal/cli/oauth_provider.go

@@ -4,7 +4,6 @@ import (
 	"context"
 	"errors"
 	"net/http"
-	"strings"
 	"time"
 
 	"github.com/Gitlawb/zero/internal/config"
@@ -12,42 +11,48 @@ import (
 	"github.com/Gitlawb/zero/internal/providers/providerio"
 )
 
-// providerOAuthName resolves the OAuth login name for a provider profile. It is
-// the name a user logs in under (`zero auth login <name>`) whose token then
-// authenticates this provider's model calls. For now it is the profile name;
-// the provider presets map well-known kinds (openai/anthropic) to canonical
-// names on top of this.
-func providerOAuthName(profile config.ProviderProfile) string {
-	return strings.TrimSpace(profile.Name)
-}
-
-// oauthResolverForProfile returns a TokenResolver that authenticates a provider's
-// model calls with the user's OAuth login, or nil when no login exists for this
-// provider. Returning nil for the no-login case keeps API-key users free of any
-// per-request store lookups: the resolver is only attached once a login is
-// present at construction time.
-func oauthResolverForProfile(profile config.ProviderProfile) providerio.TokenResolver {
-	name := providerOAuthName(profile)
-	if name == "" {
-		return nil
+// oauthLoginForProfile resolves the user's OAuth login for a provider ONCE and
+// returns both a TokenResolver that authenticates model calls with it and the
+// credential-store key it bound to. It returns (nil, "") when no login exists —
+// keeping API-key users free of any per-request store lookups, since the resolver
+// is only attached when a login is present at construction time.
+//
+// The returned key is the single source of truth for "which login is this
+// provider using": callers pass it to providers.Options.OAuthLoginKey so the
+// Codex chatgpt-account-id header reads its account from the exact login that
+// issued the bearer token, instead of doing a second, independent lookup that
+// could select a different login (a backend-rejected mismatch).
+//
+// Candidate login names (profile name, then a catalog-ID fallback, both gated on
+// the profile having no own configured credential) come from the shared
+// ProviderProfile.OAuthLoginCandidates so the runtime resolver, the Codex account
+// resolver, and the onboarding presence check never diverge.
+func oauthLoginForProfile(profile config.ProviderProfile) (providerio.TokenResolver, string) {
+	candidates := profile.OAuthLoginCandidates()
+	if len(candidates) == 0 {
+		return nil, ""
 	}
 	store, err := oauth.NewStore(oauth.StoreOptions{})
 	if err != nil {
-		return nil
+		return nil, ""
 	}
-	key := oauth.ProviderKey(name)
-	if _, ok, loadErr := store.Load(key); loadErr != nil || !ok {
-		// No login (or an unreadable/invalid key) → API-key auth, no resolver.
-		return nil
+	_, key, ok := oauth.FirstStored(store, candidates)
+	if !ok {
+		// No login under any candidate (or unreadable/invalid keys) → API-key
+		// auth, no resolver.
+		return nil, ""
 	}
 	manager, err := oauth.NewManager(oauth.ManagerOptions{
 		Store:      store,
 		HTTPClient: &http.Client{Timeout: 30 * time.Second},
+		// Refreshing a token the user logged into (possibly a preset provider like
+		// xAI) re-resolves that provider's OAuth config, which needs the preset.
+		AllowPresets: true,
 	})
 	if err != nil {
-		return nil
+		return nil, ""
 	}
-	return func(ctx context.Context, forceRefresh bool) (string, string, bool, error) {
+	resolver := func(ctx context.Context, forceRefresh bool) (string, string, bool, error) {
 		var token string
 		var rerr error
 		if forceRefresh {
@@ -64,4 +69,5 @@ func oauthResolverForProfile(profile config.ProviderProfile) providerio.TokenRes
 		}
 		return "Authorization", "Bearer " + token, true, nil
 	}
+	return resolver, key
 }

internal/cli/plugin_activate.go

@@ -3,7 +3,9 @@ package cli
 import (
 	"fmt"
 	"io"
+	"strings"
 
+	"github.com/Gitlawb/zero/internal/agent"
 	"github.com/Gitlawb/zero/internal/hooks"
 	"github.com/Gitlawb/zero/internal/plugins"
 	"github.com/Gitlawb/zero/internal/tools"
@@ -57,6 +59,27 @@ func activatePlugins(workspaceRoot string, registry *tools.Registry, deps appDep
 	return pluginActivation{hooks: result.Hooks, skillRoots: result.SkillRoots}
 }
 
+// skillInfos resolves the reusable skills the model can load via the skill tool —
+// the default skills dir merged with any plugin skill roots, the same set the
+// plugin-aware skill tool resolves against — as plain data for the agent's system
+// prompt. It returns nil when no skills are installed, so a skill-less run leaves
+// the prompt byte-identical.
+func (a pluginActivation) skillInfos(defaultDir string) []agent.SkillInfo {
+	merged, _ := plugins.MergedSkills(defaultDir, a.skillRoots)
+	if len(merged) == 0 {
+		return nil
+	}
+	infos := make([]agent.SkillInfo, 0, len(merged))
+	for _, skill := range merged {
+		name := strings.TrimSpace(skill.Name)
+		if name == "" {
+			continue
+		}
+		infos = append(infos, agent.SkillInfo{Name: name, Description: strings.TrimSpace(skill.Description)})
+	}
+	return infos
+}
+
 // formatLoadDiagnostic renders a plugin load diagnostic into a single warning
 // line, prefixing the diagnostic kind and appending the most specific locator
 // available (manifest path, then plugin path, then plugin ID) so an operator can

internal/cli/provider_catalog_onboarding_test.go

@@ -18,6 +18,7 @@ func TestProviderCatalogRuntimeProvidersShowOnboardingSetupHints(t *testing.T) {
 	}{
 		{id: "openai", name: "OpenAI", transport: "openai", defaultModel: "gpt-4.1"},
 		{id: "groq", name: "Groq", transport: "openai-compatible", defaultModel: "llama-3.3-70b-versatile"},
+		{id: "longcat", name: "LongCat", transport: "openai-compatible", defaultModel: "LongCat-2.0"},
 		{id: "ollama-cloud", name: "Ollama Cloud", transport: "openai-compatible", defaultModel: "qwen3-coder:480b"},
 		{id: "ollama", name: "Ollama Local", transport: "openai-compatible", defaultModel: "llama3.1"},
 	}
@@ -46,6 +47,21 @@ func TestProviderCatalogRuntimeProvidersShowOnboardingSetupHints(t *testing.T) {
 	}
 }
 
+func TestProviderCatalogLongCatRequiresAPIKey(t *testing.T) {
+	output := runProviderCatalogOnboarding(t)
+	block := providerCatalogOnboardingBlock(t, output, "longcat")
+
+	for _, want := range []string{
+		"requiresAuth=true",
+		"authEnvVars=LONGCAT_API_KEY",
+		"setup: zero providers setup longcat --set-active",
+	} {
+		if !strings.Contains(block, want) {
+			t.Fatalf("expected LongCat catalog block to contain %q, got:\n%s", want, block)
+		}
+	}
+}
+
 func TestProviderCatalogOllamaCloudRequiresAPIKey(t *testing.T) {
 	output := runProviderCatalogOnboarding(t)
 	block := providerCatalogOnboardingBlock(t, output, "ollama-cloud")

internal/cli/provider_models.go

@@ -0,0 +1,148 @@
+package cli
+
+import (
+	"context"
+	"fmt"
+	"io"
+	"os"
+	"strings"
+
+	"github.com/Gitlawb/zero/internal/config"
+	"github.com/Gitlawb/zero/internal/providermodeldiscovery"
+)
+
+type providerModelsOptions struct {
+	name string
+	json bool
+}
+
+// runProvidersModels lists the models a saved provider actually serves by probing
+// its live model-discovery endpoint (e.g. an OpenAI-compatible `/v1/models`). It
+// works for custom OpenAI-/Anthropic-compatible providers too: discovery runs off
+// the profile's base URL + credentials, so a self-hosted endpoint serving a dozen
+// models no longer needs a config object per model — configure the provider once,
+// then run any listed model with `zero exec --model <id>` (Zero passes unknown
+// model ids through to the provider).
+func runProvidersModels(args []string, stdout io.Writer, stderr io.Writer, deps appDeps) int {
+	options, help, err := parseProviderModelsArgs(args)
+	if err != nil {
+		return writeExecUsageError(stderr, err.Error())
+	}
+	if help {
+		if err := writeProvidersHelp(stdout); err != nil {
+			return exitCrash
+		}
+		return exitSuccess
+	}
+
+	resolved, exitCode := resolveCommandCenterConfig(stderr, deps)
+	if exitCode != exitSuccess {
+		return exitCode
+	}
+	profile, err := selectProviderForCheck(resolved, options.name)
+	if err != nil {
+		return writeExecUsageError(stderr, err.Error())
+	}
+
+	ctx, stop := signalContext()
+	defer stop()
+	models, err := deps.discoverProviderModels(ctx, discoveryCredentialProfile(profile))
+	if err != nil {
+		return writeAppError(stderr, err.Error(), exitProvider)
+	}
+
+	if options.json {
+		items := make([]map[string]any, 0, len(models))
+		for _, model := range models {
+			entry := map[string]any{"id": model.ID}
+			if description := strings.TrimSpace(model.Description); description != "" {
+				entry["description"] = description
+			}
+			items = append(items, entry)
+		}
+		payload := map[string]any{
+			"provider": profile.Name,
+			"count":    len(items),
+			"models":   items,
+		}
+		if err := writePrettyJSON(stdout, payload); err != nil {
+			return exitCrash
+		}
+		return exitSuccess
+	}
+
+	name := strings.TrimSpace(profile.Name)
+	if name == "" {
+		name = "provider"
+	}
+	if _, err := fmt.Fprintf(stdout, "Provider models (%s)\n", name); err != nil {
+		return exitCrash
+	}
+	for _, model := range models {
+		line := strings.TrimSpace(model.ID)
+		if description := strings.TrimSpace(model.Description); description != "" {
+			line += " — " + description
+		}
+		if _, err := fmt.Fprintln(stdout, line); err != nil {
+			return exitCrash
+		}
+	}
+	suffix := "s"
+	if len(models) == 1 {
+		suffix = ""
+	}
+	if _, err := fmt.Fprintf(stdout, "%d model%s discovered\n", len(models), suffix); err != nil {
+		return exitCrash
+	}
+	if len(models) > 0 {
+		if _, err := fmt.Fprintf(stdout, "next: zero exec %q --model %s\n", "hello", setupCommandArg(models[0].ID)); err != nil {
+			return exitCrash
+		}
+	}
+	return exitSuccess
+}
+
+// discoveryCredentialProfile resolves the profile's API key the same way the
+// runtime does — inline, then the stored credential, then the configured env var —
+// so a `providers models` probe authenticates exactly like a real request. Mirrors
+// discoveredModelContextWindow's credential resolution.
+func discoveryCredentialProfile(profile config.ProviderProfile) config.ProviderProfile {
+	authed := profile
+	if strings.TrimSpace(authed.APIKey) == "" {
+		if store, err := config.ProviderKeyStore(); err == nil {
+			authed = config.ApplyStoredAPIKey(authed, store)
+		}
+	}
+	if strings.TrimSpace(authed.APIKey) == "" && strings.TrimSpace(authed.APIKeyEnv) != "" {
+		authed.APIKey = strings.TrimSpace(os.Getenv(authed.APIKeyEnv))
+	}
+	return authed
+}
+
+// defaultDiscoverProviderModels is the production discovery hook: a live probe of
+// the provider's model-listing endpoint with no curated-catalog merge or
+// coding-model filtering, so a custom provider's full model list is returned.
+func defaultDiscoverProviderModels(ctx context.Context, profile config.ProviderProfile) ([]providermodeldiscovery.Model, error) {
+	return providermodeldiscovery.Discover(ctx, profile, providermodeldiscovery.Options{})
+}
+
+func parseProviderModelsArgs(args []string) (providerModelsOptions, bool, error) {
+	options := providerModelsOptions{}
+	for index := 0; index < len(args); index++ {
+		arg := args[index]
+		switch {
+		case arg == "-h" || arg == "--help" || arg == "help":
+			return options, true, nil
+		case arg == "--json":
+			options.json = true
+		case strings.HasPrefix(arg, "-"):
+			return options, false, execUsageError{fmt.Sprintf("unknown flag %q", arg)}
+		default:
+			if options.name != "" {
+				return options, false, execUsageError{fmt.Sprintf("unexpected argument %q", arg)}
+			}
+			options.name = arg
+		}
+	}
+	return options, false, nil
+}

internal/cli/provider_models_test.go

@@ -0,0 +1,151 @@
+package cli
+
+import (
+	"bytes"
+	"context"
+	"encoding/json"
+	"errors"
+	"strings"
+	"testing"
+
+	"github.com/Gitlawb/zero/internal/config"
+	"github.com/Gitlawb/zero/internal/providermodeldiscovery"
+)
+
+func TestRunProvidersModelsListsDiscoveredModels(t *testing.T) {
+	var stdout bytes.Buffer
+	var stderr bytes.Buffer
+	deps := commandCenterDeps(t)
+	var probed config.ProviderProfile
+	deps.discoverProviderModels = func(_ context.Context, profile config.ProviderProfile) ([]providermodeldiscovery.Model, error) {
+		probed = profile
+		return []providermodeldiscovery.Model{
+			{ID: "team/model-a", Description: "first"},
+			{ID: "team/model-b"},
+		}, nil
+	}
+
+	exitCode := runWithDeps([]string{"providers", "models"}, &stdout, &stderr, deps)
+
+	if exitCode != exitSuccess {
+		t.Fatalf("exit = %d, want %d: %s", exitCode, exitSuccess, stderr.String())
+	}
+	// No name given → the active provider is probed, with its resolved credential.
+	if probed.Name != "work" {
+		t.Fatalf("probed provider = %q, want active provider work", probed.Name)
+	}
+	out := stdout.String()
+	for _, want := range []string{"Provider models (work)", "team/model-a — first", "team/model-b", "2 models discovered", "--model team/model-a"} {
+		if !strings.Contains(out, want) {
+			t.Fatalf("output missing %q:\n%s", want, out)
+		}
+	}
+}
+
+func TestRunProvidersModelsJSON(t *testing.T) {
+	var stdout bytes.Buffer
+	var stderr bytes.Buffer
+	deps := commandCenterDeps(t)
+	deps.discoverProviderModels = func(_ context.Context, _ config.ProviderProfile) ([]providermodeldiscovery.Model, error) {
+		return []providermodeldiscovery.Model{{ID: "m1", Description: "d1"}, {ID: "m2"}}, nil
+	}
+
+	exitCode := runWithDeps([]string{"providers", "models", "--json"}, &stdout, &stderr, deps)
+
+	if exitCode != exitSuccess {
+		t.Fatalf("exit = %d, want %d: %s", exitCode, exitSuccess, stderr.String())
+	}
+	var payload struct {
+		Provider string `json:"provider"`
+		Count    int    `json:"count"`
+		Models   []struct {
+			ID          string `json:"id"`
+			Description string `json:"description"`
+		} `json:"models"`
+	}
+	if err := json.Unmarshal(stdout.Bytes(), &payload); err != nil {
+		t.Fatalf("decode json: %v\n%s", err, stdout.String())
+	}
+	if payload.Provider != "work" || payload.Count != 2 || len(payload.Models) != 2 {
+		t.Fatalf("payload = %#v", payload)
+	}
+	if payload.Models[0].ID != "m1" || payload.Models[0].Description != "d1" {
+		t.Fatalf("model[0] = %#v", payload.Models[0])
+	}
+	// A model with no description omits the field rather than emitting an empty string.
+	if payload.Models[1].ID != "m2" || payload.Models[1].Description != "" {
+		t.Fatalf("model[1] = %#v", payload.Models[1])
+	}
+}
+
+func TestRunProvidersModelsSelectsNamedProvider(t *testing.T) {
+	var stdout bytes.Buffer
+	var stderr bytes.Buffer
+	deps := commandCenterDeps(t)
+	custom := config.ProviderProfile{
+		Name:         "self-hosted",
+		ProviderKind: config.ProviderKindOpenAICompatible,
+		BaseURL:      "https://models.internal/v1",
+		APIKey:       "sk-internal",
+		Model:        "house-model",
+	}
+	deps.resolveConfig = func(string, config.Overrides) (config.ResolvedConfig, error) {
+		work := config.ProviderProfile{Name: "work", ProviderKind: config.ProviderKindOpenAI, BaseURL: config.OpenAIBaseURL, APIKey: "sk", Model: "gpt-4.1"}
+		return config.ResolvedConfig{ActiveProvider: "work", Provider: work, Providers: []config.ProviderProfile{work, custom}}, nil
+	}
+	var probed config.ProviderProfile
+	deps.discoverProviderModels = func(_ context.Context, profile config.ProviderProfile) ([]providermodeldiscovery.Model, error) {
+		probed = profile
+		return []providermodeldiscovery.Model{{ID: "house-model"}}, nil
+	}
+
+	exitCode := runWithDeps([]string{"providers", "models", "self-hosted"}, &stdout, &stderr, deps)
+
+	if exitCode != exitSuccess {
+		t.Fatalf("exit = %d, want %d: %s", exitCode, exitSuccess, stderr.String())
+	}
+	if probed.Name != "self-hosted" || probed.BaseURL != "https://models.internal/v1" {
+		t.Fatalf("probed = %#v, want the named custom provider", probed)
+	}
+}
+
+func TestRunProvidersModelsUnknownProviderFails(t *testing.T) {
+	var stdout bytes.Buffer
+	var stderr bytes.Buffer
+	deps := commandCenterDeps(t)
+	called := false
+	deps.discoverProviderModels = func(_ context.Context, _ config.ProviderProfile) ([]providermodeldiscovery.Model, error) {
+		called = true
+		return nil, nil
+	}
+
+	exitCode := runWithDeps([]string{"providers", "models", "ghost"}, &stdout, &stderr, deps)
+
+	if exitCode == exitSuccess {
+		t.Fatalf("exit = %d, want non-zero for unknown provider", exitCode)
+	}
+	if called {
+		t.Fatal("discovery must not run for an unknown provider")
+	}
+	if !strings.Contains(stderr.String(), `provider "ghost" not found`) {
+		t.Fatalf("stderr = %q, want provider-not-found", stderr.String())
+	}
+}
+
+func TestRunProvidersModelsSurfacesDiscoveryError(t *testing.T) {
+	var stdout bytes.Buffer
+	var stderr bytes.Buffer
+	deps := commandCenterDeps(t)
+	deps.discoverProviderModels = func(_ context.Context, _ config.ProviderProfile) ([]providermodeldiscovery.Model, error) {
+		return nil, errors.New("models endpoint returned 404")
+	}
+
+	exitCode := runWithDeps([]string{"providers", "models"}, &stdout, &stderr, deps)
+
+	if exitCode != exitProvider {
+		t.Fatalf("exit = %d, want exitProvider %d", exitCode, exitProvider)
+	}
+	if !strings.Contains(stderr.String(), "models endpoint returned 404") {
+		t.Fatalf("stderr = %q, want the discovery error", stderr.String())
+	}
+}

internal/cli/provider_setup.go

@@ -488,7 +488,7 @@ func validateProviderRuntimeReady(profile config.ProviderProfile) error {
 }
 
 func providerProfileHasCredential(profile config.ProviderProfile) bool {
-	return strings.TrimSpace(profile.APIKey) != "" || profile.APIKeyStored || strings.TrimSpace(profile.AuthHeaderValue) != ""
+	return profile.HasConfiguredCredential()
 }
 
 func providerKindForDescriptor(descriptor providercatalog.Descriptor) config.ProviderKind {

internal/cli/repoinfo.go

@@ -88,6 +88,15 @@ func formatRepoInfo(info repoinfo.Info) string {
 	fmt.Fprintf(&b, "Files:        %d\n", info.FileCount)
 	fmt.Fprintf(&b, "Directories:  %d (max depth %d)\n", info.DirectoryCount, info.MaxDepth)
 	fmt.Fprintf(&b, "Est. LOC:     %d\n", info.LOCEstimate)
+	if info.CommitCount != nil {
+		fmt.Fprintf(&b, "Commits:      %d\n", *info.CommitCount)
+	}
+	if info.BranchCount != nil {
+		fmt.Fprintf(&b, "Branches:     %d\n", *info.BranchCount)
+	}
+	if info.TagCount != nil {
+		fmt.Fprintf(&b, "Tags:         %d\n", *info.TagCount)
+	}
 	if info.PrimaryLanguage != "" {
 		fmt.Fprintf(&b, "Primary lang: %s\n", info.PrimaryLanguage)
 	}

internal/cli/repoinfo_test.go

@@ -33,7 +33,7 @@ func TestFormatRepoInfoText(t *testing.T) {
 	contrib := 3
 	info := repoInfoTestData(age, contrib)
 	out := formatRepoInfo(info)
-	for _, want := range []string{"Files", "TypeScript", "main", "GitHub Actions", "42", "3"} {
+	for _, want := range []string{"Files", "TypeScript", "main", "GitHub Actions", "42", "3", "Commits:      100", "Branches:     5", "Tags:         10"} {
 		if !strings.Contains(out, want) {
 			t.Fatalf("text output missing %q:\n%s", want, out)
 		}
@@ -113,12 +113,16 @@ func initTempGitRepo(t *testing.T) string {
 }
 
 func repoInfoTestData(age, contrib int) repoinfo.Info {
+	commits := 100
+	branches := 5
+	tags := 10
 	return repoinfo.Info{
 		FileCount: 10, DirectoryCount: 3, MaxDepth: 2, LOCEstimate: 500,
 		Languages:       []repoinfo.LangStat{{Name: "TypeScript", LOCEstimate: 400, FileCount: 4}},
 		PrimaryLanguage: "TypeScript", LanguageCount: 1,
 		WorkspaceType: "none", CICD: []string{"GitHub Actions"},
 		HasGit: true, Branch: "main", AgeDays: &age, Contributors90d: &contrib,
+		CommitCount: &commits, BranchCount: &branches, TagCount: &tags,
 	}
 }
 

internal/cli/setup.go

@@ -3,6 +3,7 @@ package cli
 import (
 	"fmt"
 	"io"
+	"net"
 	"net/url"
 	"strconv"
 	"strings"
@@ -314,10 +315,14 @@ func setupRequired(resolved config.ResolvedConfig) bool {
 }
 
 // providerHasOAuthLogin reports whether a stored OAuth login exists for the
-// provider, keyed by its profile name or catalog id.
+// provider. It uses the same ProviderProfile.OAuthLoginCandidates the runtime
+// resolver does, so the "authenticated in the UI" gate and the "authenticated at
+// runtime" resolver can never diverge. This is only consulted for a profile with
+// no usable API-key credential (setupRequired returns early otherwise), where
+// the candidate set is permissive (profile name + catalog ID).
 func providerHasOAuthLogin(profile config.ProviderProfile, oauthLogins map[string]bool) bool {
-	for _, name := range []string{strings.TrimSpace(profile.Name), strings.TrimSpace(profile.CatalogID)} {
-		if name != "" && oauthLogins[name] {
+	for _, name := range profile.OAuthLoginCandidates() {
+		if oauthLogins[name] {
 			return true
 		}
 	}
@@ -449,6 +454,11 @@ func providerProfileIsLocal(profile config.ProviderProfile) bool {
 	case "localhost", "127.0.0.1", "::1":
 		return true
 	}
+	// Also treat explicit private-network IPs (192.168.x.y, 10.x.y.z,
+	// 172.16-31.x.y) as local — the user is pointing at their own LAN box.
+	if ip := net.ParseIP(parsed.Hostname()); ip != nil && ip.IsPrivate() {
+		return true
+	}
 	return false
 }
 

internal/cli/specialist_test.go

@@ -8,6 +8,8 @@ import (
 	"runtime"
 	"strings"
 	"testing"
+
+	"github.com/Gitlawb/zero/internal/config"
 )
 
 func TestRunSpecialistListShowAndPath(t *testing.T) {
@@ -326,7 +328,7 @@ func setSpecialistConfigRoot(t *testing.T) string {
 	default:
 		t.Setenv("XDG_CONFIG_HOME", root)
 	}
-	configRoot, err := os.UserConfigDir()
+	configRoot, err := config.UserConfigDir()
 	if err != nil {
 		t.Fatalf("UserConfigDir() error = %v", err)
 	}

internal/cli/update.go

@@ -12,6 +12,7 @@ import (
 
 type updateOptions struct {
 	check      bool
+	apply      bool
 	json       bool
 	repository string
 	endpoint   string
@@ -20,6 +21,18 @@ type updateOptions struct {
 }
 
 func runUpdate(args []string, stdout io.Writer, stderr io.Writer, deps appDeps) int {
+	return runUpdateCommand(args, stdout, stderr, deps, false)
+}
+
+func runUpgrade(args []string, stdout io.Writer, stderr io.Writer, deps appDeps) int {
+	return runUpdateCommand(args, stdout, stderr, deps, true)
+}
+
+// runUpdateCommand backs both `zero update` and `zero upgrade`. defaultApply
+// makes `--apply` the implicit behavior for `zero upgrade` when neither
+// `--check` nor `--apply` is passed explicitly; `zero update` keeps requiring
+// one of the two so existing scripts around `zero update --check` don't change.
+func runUpdateCommand(args []string, stdout io.Writer, stderr io.Writer, deps appDeps, defaultApply bool) int {
 	options, help, err := parseUpdateArgs(args)
 	if err != nil {
 		return writeUsageError(stderr, err.Error())
@@ -30,10 +43,19 @@ func runUpdate(args []string, stdout io.Writer, stderr io.Writer, deps appDeps)
 		}
 		return exitSuccess
 	}
-	if !options.check {
-		return writeUsageError(stderr, "Only `zero update --check` is available right now.")
+	if options.check && options.apply {
+		return writeUsageError(stderr, "Pass only one of --check or --apply.")
 	}
-	checkOptions := update.Options{
+	if !options.check && !options.apply {
+		if !defaultApply {
+			return writeUsageError(stderr, "Pass --check to check for updates or --apply to install one (or use `zero upgrade`).")
+		}
+		options.apply = true
+	}
+	if options.apply && options.target != "" {
+		return writeUsageError(stderr, "--target cannot be combined with --apply; it only verifies release assets for --check.")
+	}
+	updateOptions := update.Options{
 		CurrentVersion: version,
 		Repository:     options.repository,
 		Endpoint:       options.endpoint,
@@ -44,10 +66,26 @@ func runUpdate(args []string, stdout io.Writer, stderr io.Writer, deps appDeps)
 		if err != nil {
 			return writeUsageError(stderr, err.Error())
 		}
-		checkOptions.GOOS = target.GOOS
-		checkOptions.GOARCH = target.GOARCH
+		updateOptions.GOOS = target.GOOS
+		updateOptions.GOARCH = target.GOARCH
 	}
-	result, err := deps.checkUpdate(context.Background(), checkOptions)
+	if options.apply {
+		result, err := deps.applyUpdate(context.Background(), updateOptions)
+		if err != nil {
+			return writeAppError(stderr, "Could not install update: "+err.Error(), exitCrash)
+		}
+		if options.json {
+			if err := writePrettyJSON(stdout, result); err != nil {
+				return exitCrash
+			}
+			return exitSuccess
+		}
+		if _, err := fmt.Fprintln(stdout, update.FormatApply(result)); err != nil {
+			return exitCrash
+		}
+		return exitSuccess
+	}
+	result, err := deps.checkUpdate(context.Background(), updateOptions)
 	if err != nil {
 		return writeAppError(stderr, "Could not check for updates: "+err.Error(), exitCrash)
 	}
@@ -72,6 +110,8 @@ func parseUpdateArgs(args []string) (updateOptions, bool, error) {
 			return options, true, nil
 		case arg == "--check":
 			options.check = true
+		case arg == "--apply":
+			options.apply = true
 		case arg == "--json":
 			options.json = true
 		case arg == "--repo":
@@ -155,14 +195,17 @@ func parseUpdateTarget(value string) (string, error) {
 func writeUpdateHelp(w io.Writer) error {
 	_, err := fmt.Fprint(w, `Usage:
   zero update --check [flags]
+  zero update --apply [flags]
+  zero upgrade [flags]
 
 Flags:
       --check                 Check the latest GitHub release without installing
-      --json                  Print the update check result as JSON
+      --apply                 Download, verify, and install the latest release
+      --json                  Print the update result as JSON
       --repo <owner/repo>     Repository to check when no endpoint is provided
       --endpoint <url|repo>   Release API URL or owner/repo slug to check
       --timeout <duration>    Release check timeout (default 5s)
-      --target <platform>     Release target to verify (for example windows-x64)
+      --target <platform>     Release target to verify with --check (for example windows-x64); not valid with --apply
   -h, --help                  Show this help
 `)
 	return err

internal/cli/workflow_test.go

@@ -662,3 +662,144 @@ func TestRunExecRejectsWorktreeDirWithoutWorktree(t *testing.T) {
 		t.Fatalf("expected empty stdout, got %q", stdout.String())
 	}
 }
+
+func TestParseChangesArgsAuto(t *testing.T) {
+	// 1. Correct parsing of auto
+	for _, args := range [][]string{
+		{"--auto"},
+		{"-a"},
+	} {
+		options, help, err := parseChangesArgs(args, "commit")
+		if err != nil {
+			t.Fatalf("parseChangesArgs(%v) error: %v", args, err)
+		}
+		if help {
+			t.Fatalf("parseChangesArgs(%v) returned help", args)
+		}
+		if !options.auto {
+			t.Fatalf("auto = false, want true (args %v)", args)
+		}
+	}
+
+	// 2. Reject --auto on inspect
+	_, _, err := parseChangesArgs([]string{"--auto"}, "inspect")
+	if err == nil || !strings.Contains(err.Error(), "--auto") {
+		t.Fatalf("expected --auto rejection on inspect, got %v", err)
+	}
+
+	// 3. Reject both --message and --auto on commit
+	_, _, err = parseChangesArgs([]string{"--message", "foo", "--auto"}, "commit")
+	if err == nil || !strings.Contains(err.Error(), "cannot specify both --message and --auto") {
+		t.Fatalf("expected --message and --auto conflict error, got %v", err)
+	}
+
+	// 4. Reject both empty message and --auto on commit
+	_, _, err = parseChangesArgs([]string{"--message=", "--auto"}, "commit")
+	if err == nil || !strings.Contains(err.Error(), "cannot specify both --message and --auto") {
+		t.Fatalf("expected empty --message and --auto conflict error, got %v", err)
+	}
+}
+
+type mockCommitMsgProvider struct {
+	response string
+	req      zeroruntime.CompletionRequest
+}
+
+func (p *mockCommitMsgProvider) StreamCompletion(ctx context.Context, request zeroruntime.CompletionRequest) (<-chan zeroruntime.StreamEvent, error) {
+	p.req = request
+	events := make(chan zeroruntime.StreamEvent, 2)
+	events <- zeroruntime.StreamEvent{Type: zeroruntime.StreamEventText, Content: p.response}
+	events <- zeroruntime.StreamEvent{Type: zeroruntime.StreamEventDone}
+	close(events)
+	return events, nil
+}
+
+func TestRunChangesCommitAuto(t *testing.T) {
+	cwd := t.TempDir()
+	summary := zerogit.ChangeSummary{
+		Root:   cwd,
+		Branch: "main",
+		Files:  []zerogit.FileChange{{Path: "README.md", Status: "modified"}},
+		Diff:   "some diff content with ghp_SECRETKEYHERE",
+	}
+
+	mockProv := &mockCommitMsgProvider{
+		response: "```\nfeat: auto commit message\n```",
+	}
+
+	var stdout bytes.Buffer
+	var stderr bytes.Buffer
+	commitCalled := false
+
+	exitCode := runWithDeps([]string{"changes", "commit", "--auto"}, &stdout, &stderr, appDeps{
+		getwd: func() (string, error) { return cwd, nil },
+		inspectChanges: func(ctx context.Context, options zerogit.InspectOptions) (zerogit.ChangeSummary, error) {
+			return summary, nil
+		},
+		resolveConfig: func(workspaceRoot string, overrides config.Overrides) (config.ResolvedConfig, error) {
+			cfg := execResolvedConfig()
+			cfg.Provider.Name = "openai"
+			cfg.Provider.Model = "gpt-4o"
+			return cfg, nil
+		},
+		newProvider: func(profile config.ProviderProfile) (zeroruntime.Provider, error) {
+			return mockProv, nil
+		},
+		commitChanges: func(ctx context.Context, options zerogit.CommitOptions) (zerogit.CommitResult, error) {
+			commitCalled = true
+			if options.Message != "feat: auto commit message" {
+				t.Fatalf("expected commit message 'feat: auto commit message', got %q", options.Message)
+			}
+			return zerogit.CommitResult{
+				Root:       cwd,
+				Message:    options.Message,
+				Committed:  true,
+				CommitHash: "abc1234",
+				Before:     summary,
+			}, nil
+		},
+	})
+
+	if exitCode != exitSuccess {
+		t.Fatalf("expected exit code %d, got %d: %s", exitSuccess, exitCode, stderr.String())
+	}
+	if !commitCalled {
+		t.Fatal("expected commitChanges to be called")
+	}
+	// Verify that secret in the diff was redacted
+	promptContent := mockProv.req.Messages[0].Content
+	if strings.Contains(promptContent, "ghp_SECRETKEYHERE") {
+		t.Fatal("expected secret in diff to be redacted, but it was found in the prompt")
+	}
+	if !strings.Contains(promptContent, "[REDACTED]") && !strings.Contains(promptContent, "REDACTED") {
+		t.Fatalf("expected redacted diff content in prompt, got: %q", promptContent)
+	}
+	if !strings.Contains(stdout.String(), "Generating commit message using LLM...") {
+		t.Fatalf("expected status message in stdout, got %q", stdout.String())
+	}
+
+	t.Run("EmptyLLMResponse", func(t *testing.T) {
+		mockProvEmpty := &mockCommitMsgProvider{
+			response: "   \n\n   ",
+		}
+		var stdout, stderr bytes.Buffer
+		code := runWithDeps([]string{"changes", "commit", "--auto"}, &stdout, &stderr, appDeps{
+			getwd: func() (string, error) { return cwd, nil },
+			inspectChanges: func(ctx context.Context, options zerogit.InspectOptions) (zerogit.ChangeSummary, error) {
+				return summary, nil
+			},
+			resolveConfig: func(workspaceRoot string, overrides config.Overrides) (config.ResolvedConfig, error) {
+				return execResolvedConfig(), nil
+			},
+			newProvider: func(profile config.ProviderProfile) (zeroruntime.Provider, error) {
+				return mockProvEmpty, nil
+			},
+		})
+		if code == exitSuccess {
+			t.Fatal("expected command to fail when LLM returns empty response, got success")
+		}
+		if !strings.Contains(stderr.String(), "empty commit message") {
+			t.Fatalf("expected empty commit message error in stderr, got %q", stderr.String())
+		}
+	})
+}

internal/cli/workflows.go

@@ -6,13 +6,16 @@ import (
 	"io"
 	"strconv"
 	"strings"
+	"time"
 
+	"github.com/Gitlawb/zero/internal/config"
 	"github.com/Gitlawb/zero/internal/redaction"
 	"github.com/Gitlawb/zero/internal/selfverify"
 	"github.com/Gitlawb/zero/internal/testrunner"
 	"github.com/Gitlawb/zero/internal/verify"
 	"github.com/Gitlawb/zero/internal/worktrees"
 	"github.com/Gitlawb/zero/internal/zerogit"
+	"github.com/Gitlawb/zero/internal/zeroruntime"
 )
 
 type worktreeCommandOptions struct {
@@ -35,8 +38,17 @@ type changesCommandOptions struct {
 	cwd          string
 	baseRef      string
 	message      string
+	hasMessage   bool
 	dryRun       bool
 	maxDiffBytes int
+	remote       string
+	force        bool
+	title        string
+	body         string
+	fill         bool
+	draft        bool
+	yes          bool
+	auto         bool
 }
 
 func runWorktrees(args []string, stdout io.Writer, stderr io.Writer, deps appDeps) int {
@@ -198,9 +210,52 @@ func runChanges(args []string, stdout io.Writer, stderr io.Writer, deps appDeps)
 		}
 		return exitSuccess
 	case "commit":
+		var message string
+		if options.auto {
+			summary, err := deps.inspectChanges(context.Background(), zerogit.InspectOptions{
+				Cwd:          workspaceRoot,
+				MaxDiffBytes: options.maxDiffBytes,
+			})
+			if err != nil {
+				return writeExecUsageError(stderr, fmt.Sprintf("failed to inspect changes: %v", err))
+			}
+			if summary.Clean {
+				return writeExecUsageError(stderr, "no changes to commit")
+			}
+
+			resolved, err := deps.resolveConfig(workspaceRoot, config.Overrides{})
+			if err != nil {
+				return writeExecUsageError(stderr, fmt.Sprintf("failed to resolve config: %v", err))
+			}
+			if !config.HasProviderProfile(resolved.Provider) {
+				return writeExecUsageError(stderr, "no provider configured for auto-commit message")
+			}
+			provider, err := deps.newProvider(resolved.Provider)
+			if err != nil {
+				return writeExecUsageError(stderr, fmt.Sprintf("failed to create provider: %v", err))
+			}
+
+			if !options.json {
+				if _, err := fmt.Fprintln(stdout, "Generating commit message using LLM..."); err != nil {
+					return exitCrash
+				}
+			}
+
+			safeSummary := redactChangeSummary(summary)
+			genCtx, cancel := context.WithTimeout(context.Background(), 60*time.Second)
+			defer cancel()
+			msg, err := generateAutoCommitMessage(genCtx, provider, resolved.Provider.Model, safeSummary)
+			if err != nil {
+				return writeExecUsageError(stderr, fmt.Sprintf("failed to generate commit message: %v", err))
+			}
+			message = msg
+		} else {
+			message = options.message
+		}
+
 		result, err := deps.commitChanges(context.Background(), zerogit.CommitOptions{
 			Cwd:          workspaceRoot,
-			Message:      options.message,
+			Message:      message,
 			DryRun:       options.dryRun,
 			MaxDiffBytes: options.maxDiffBytes,
 		})
@@ -218,6 +273,10 @@ func runChanges(args []string, stdout io.Writer, stderr io.Writer, deps appDeps)
 			return exitCrash
 		}
 		return exitSuccess
+	case "push":
+		return runChangesPush(args, stdout, stderr, deps)
+	case "pr", "pull-request":
+		return runChangesPR(args, stdout, stderr, deps)
 	default:
 		return writeExecUsageError(stderr, fmt.Sprintf("unknown changes command %q", command))
 	}
@@ -395,9 +454,11 @@ func parseChangesArgs(args []string, command string) (changesCommandOptions, boo
 				return options, false, err
 			}
 			options.message = value
+			options.hasMessage = true
 			index = next
 		case strings.HasPrefix(arg, "--message="):
 			options.message = strings.TrimSpace(strings.TrimPrefix(arg, "--message="))
+			options.hasMessage = true
 		case arg == "--diff-bytes":
 			value, next, err := nextFlagValue(args, index, arg)
 			if err != nil {
@@ -415,18 +476,73 @@ func parseChangesArgs(args []string, command string) (changesCommandOptions, boo
 				return options, false, err
 			}
 			options.maxDiffBytes = maxDiffBytes
+		case arg == "--remote":
+			value, next, err := nextFlagValue(args, index, arg)
+			if err != nil {
+				return options, false, err
+			}
+			options.remote = strings.TrimSpace(value)
+			index = next
+		case strings.HasPrefix(arg, "--remote="):
+			options.remote = strings.TrimSpace(strings.TrimPrefix(arg, "--remote="))
+		case arg == "--force":
+			options.force = true
+		case arg == "--title":
+			value, next, err := nextFlagValue(args, index, arg)
+			if err != nil {
+				return options, false, err
+			}
+			options.title = value
+			index = next
+		case strings.HasPrefix(arg, "--title="):
+			options.title = strings.TrimPrefix(arg, "--title=")
+		case arg == "--body":
+			value, next, err := nextFlagValue(args, index, arg)
+			if err != nil {
+				return options, false, err
+			}
+			options.body = value
+			index = next
+		case strings.HasPrefix(arg, "--body="):
+			options.body = strings.TrimPrefix(arg, "--body=")
+		case arg == "--fill":
+			options.fill = true
+		case arg == "--draft":
+			options.draft = true
+		case arg == "--yes":
+			options.yes = true
+		case arg == "-a" || arg == "--auto":
+			options.auto = true
 		case strings.HasPrefix(arg, "-"):
 			return options, false, execUsageError{fmt.Sprintf("unknown changes flag %q", arg)}
 		default:
 			return options, false, execUsageError{fmt.Sprintf("unexpected changes argument %q", arg)}
 		}
 	}
-	if command != "commit" && (options.message != "" || options.dryRun) {
-		return options, false, execUsageError{"--message and --dry-run are only valid with `zero changes commit`"}
+	if command != "commit" && options.message != "" {
+		return options, false, execUsageError{"--message is only valid with `zero changes commit`"}
 	}
-	if command == "commit" && options.baseRef != "" {
+	if command != "commit" && (options.hasMessage || options.dryRun || options.auto) {
+		return options, false, execUsageError{"--message, --dry-run, and --auto are only valid with `zero changes commit`"}
+	}
+	if command == "commit" && options.hasMessage && options.auto {
+		return options, false, execUsageError{"cannot specify both --message and --auto"}
+	}
+	if command != "commit" && command != "push" && options.dryRun {
+		return options, false, execUsageError{"--dry-run is only valid with commit or push"}
+	}
+	if command != "inspect" && options.baseRef != "" {
 		return options, false, execUsageError{"--base is only valid with `zero changes inspect`"}
 	}
+	if command != "push" && command != "pr" && (options.remote != "" || options.force) {
+		return options, false, execUsageError{"--remote and --force are only valid with push or pr"}
+	}
+	if command != "pr" && (options.title != "" || options.body != "" || options.fill || options.draft) {
+		return options, false, execUsageError{"--title, --body, --fill, and --draft are only valid with `zero changes pr`"}
+	}
+	if command != "push" && command != "pr" && options.yes {
+		return options, false, execUsageError{"--yes is only valid with push or pr"}
+	}
 	return options, false, nil
 }
 
@@ -704,17 +820,187 @@ func writeChangesHelp(w io.Writer) error {
 	_, err := fmt.Fprint(w, `Usage:
   zero changes inspect [flags]
   zero changes commit [flags]
+  zero changes push [flags]
+  zero changes pr [flags]
 
-Inspects local git changes and optionally creates a commit.
+Inspects, commits, pushes, and creates pull requests for local git changes.
 
 Flags:
   -C, --cwd <path>        Workspace directory
       --base <ref>        Diff against <ref>...HEAD instead of the working tree
       --diff-bytes <n>    Maximum diff bytes to include
   -m, --message <text>    Commit message for `+"`zero changes commit`"+`
+      --dry-run           Preview commit metadata / push without mutating git state
+      --remote <name>     Remote to push to (defaults to upstream tracked branch or origin)
+      --force             Use force-with-lease when pushing
+      --title <text>      PR title
+      --body <text>       PR body
+      --fill              Automatically populate PR title and body from commits
+      --draft             Create PR as a draft
+      --yes               Confirm pushing to a default/protected branch
+  -a, --auto              Auto-generate commit message using LLM (use --dry-run to preview)
       --dry-run           Preview commit metadata without mutating git state
       --json              Print JSON output
   -h, --help              Show this help
 `)
 	return err
 }
+
+func runChangesPush(args []string, stdout io.Writer, stderr io.Writer, deps appDeps) int {
+	options, help, err := parseChangesArgs(args, "push")
+	if err != nil {
+		return writeExecUsageError(stderr, err.Error())
+	}
+	if help {
+		if err := writeChangesHelp(stdout); err != nil {
+			return exitCrash
+		}
+		return exitSuccess
+	}
+	workspaceRoot, err := resolveWorkspaceRoot(options.cwd, deps)
+	if err != nil {
+		return writeExecUsageError(stderr, err.Error())
+	}
+
+	result, err := deps.pushChanges(context.Background(), zerogit.PushOptions{
+		Cwd:                    workspaceRoot,
+		Remote:                 options.remote,
+		Force:                  options.force,
+		DryRun:                 options.dryRun,
+		AllowPushDefaultBranch: options.yes,
+	})
+	if err != nil {
+		return writeExecUsageError(stderr, err.Error())
+	}
+
+	if options.json {
+		if err := writePrettyJSON(stdout, result); err != nil {
+			return exitCrash
+		}
+		return exitSuccess
+	}
+
+	dryRunStr := ""
+	if options.dryRun {
+		dryRunStr = " (dry run)"
+	}
+	if _, err := fmt.Fprintf(stdout, "Pushed branch %s to remote %s%s\n", result.Branch, result.Remote, dryRunStr); err != nil {
+		return exitCrash
+	}
+	if result.Output != "" {
+		if _, err := fmt.Fprintln(stdout, result.Output); err != nil {
+			return exitCrash
+		}
+	}
+	return exitSuccess
+}
+
+func runChangesPR(args []string, stdout io.Writer, stderr io.Writer, deps appDeps) int {
+	options, help, err := parseChangesArgs(args, "pr")
+	if err != nil {
+		return writeExecUsageError(stderr, err.Error())
+	}
+	if help {
+		if err := writeChangesHelp(stdout); err != nil {
+			return exitCrash
+		}
+		return exitSuccess
+	}
+	if !options.fill && options.title == "" {
+		return writeExecUsageError(stderr, "must provide either --fill or --title to run non-interactively")
+	}
+	workspaceRoot, err := resolveWorkspaceRoot(options.cwd, deps)
+	if err != nil {
+		return writeExecUsageError(stderr, err.Error())
+	}
+
+	if !options.json {
+		if _, err := fmt.Fprintln(stdout, "Pushing current branch to set upstream..."); err != nil {
+			return exitCrash
+		}
+	}
+	pushResult, err := deps.pushChanges(context.Background(), zerogit.PushOptions{
+		Cwd:                    workspaceRoot,
+		Remote:                 options.remote,
+		Force:                  options.force,
+		AllowPushDefaultBranch: options.yes,
+	})
+	if err != nil {
+		return writeExecUsageError(stderr, fmt.Sprintf("auto-push failed: %v", err))
+	}
+	if !options.json {
+		if _, err := fmt.Fprintf(stdout, "Pushed branch %s to remote %s\n", pushResult.Branch, pushResult.Remote); err != nil {
+			return exitCrash
+		}
+	}
+
+	prResult, err := deps.createPR(context.Background(), zerogit.PROptions{
+		Cwd:   workspaceRoot,
+		Fill:  options.fill,
+		Draft: options.draft,
+		Title: options.title,
+		Body:  options.body,
+	})
+	if err != nil {
+		return writeExecUsageError(stderr, err.Error())
+	}
+
+	if options.json {
+		type prJSONResult struct {
+			Branch string `json:"branch"`
+			Remote string `json:"remote"`
+			Output string `json:"output"`
+		}
+		res := prJSONResult{
+			Branch: pushResult.Branch,
+			Remote: pushResult.Remote,
+			Output: strings.TrimSpace(prResult.Output),
+		}
+		if err := writePrettyJSON(stdout, res); err != nil {
+			return exitCrash
+		}
+		return exitSuccess
+	}
+
+	if _, err := fmt.Fprintln(stdout, strings.TrimSpace(prResult.Output)); err != nil {
+		return exitCrash
+	}
+	return exitSuccess
+}
+
+func generateAutoCommitMessage(ctx context.Context, provider zeroruntime.Provider, model string, summary zerogit.ChangeSummary) (string, error) {
+	var promptBuilder strings.Builder
+	promptBuilder.WriteString("Analyze the following git diff and generate a concise, conventional commit message.\n")
+	promptBuilder.WriteString("The commit message subject line must be 72 characters or fewer, starting with a conventional commit type (e.g., feat, fix, docs, style, refactor, test, chore) followed by a colon and space, and a lowercase description.\n")
+	promptBuilder.WriteString("You may optionally include a blank line and a bulleted list of changes for the body if there are multiple files or complex changes.\n")
+	promptBuilder.WriteString("Output ONLY the raw commit message text. Do not wrap the message in markdown code block fence, do not include quotes or any introduction/explanation.\n\n")
+	promptBuilder.WriteString("Git Diff:\n")
+	promptBuilder.WriteString(summary.Diff)
+
+	request := zeroruntime.CompletionRequest{
+		Messages: []zeroruntime.Message{
+			{Role: zeroruntime.MessageRoleUser, Content: promptBuilder.String()},
+		},
+	}
+	stream, err := provider.StreamCompletion(ctx, request)
+	if err != nil {
+		return "", err
+	}
+	collected := zeroruntime.CollectStream(ctx, stream)
+	if collected.Error != "" {
+		return "", fmt.Errorf("%s", collected.Error)
+	}
+
+	msg := strings.TrimSpace(collected.Text)
+	if strings.HasPrefix(msg, "```") {
+		if idx := strings.Index(msg, "\n"); idx != -1 {
+			msg = msg[idx+1:]
+		}
+	}
+	msg = strings.TrimSuffix(msg, "```")
+	msg = strings.TrimSpace(msg)
+	if msg == "" {
+		return "", fmt.Errorf("provider returned empty commit message")
+	}
+	return msg, nil
+}

internal/config/active_provider_env_test.go

@@ -0,0 +1,49 @@
+package config
+
+import (
+	"os"
+	"testing"
+)
+
+// SetActiveProviderEnv must export the provider name so a child process (whose
+// resolution reads the inherited environment via the nil-map -> os.Getenv path in
+// applyEnv) resolves the SAME active provider. A blank name CLEARS it.
+func TestSetActiveProviderEnvRoundTrips(t *testing.T) {
+	t.Setenv(ActiveProviderEnv, "") // register cleanup; restores the original after
+
+	if err := os.Unsetenv(ActiveProviderEnv); err != nil {
+		t.Fatalf("unset: %v", err)
+	}
+	SetActiveProviderEnv("") // blank with nothing set stays unset
+	if got := os.Getenv(ActiveProviderEnv); got != "" {
+		t.Fatalf("empty name set the env to %q, want unset", got)
+	}
+
+	// Regression: switching from a named provider back to an unnamed/default profile
+	// must CLEAR a previously exported value, not leave it stale for children.
+	if err := os.Setenv(ActiveProviderEnv, "stale-provider"); err != nil {
+		t.Fatalf("seed stale: %v", err)
+	}
+	SetActiveProviderEnv("") // blank must unset the stale value
+	if got := os.Getenv(ActiveProviderEnv); got != "" {
+		t.Fatalf("blank name left stale env %q, want cleared", got)
+	}
+	cfgStale := &FileConfig{}
+	applyEnv(cfgStale, nil)
+	if cfgStale.ActiveProvider != "" {
+		t.Fatalf("applyEnv still resolved stale provider %q after clear", cfgStale.ActiveProvider)
+	}
+
+	SetActiveProviderEnv("ollama-cloud")
+	if got := os.Getenv(ActiveProviderEnv); got != "ollama-cloud" {
+		t.Fatalf("env = %q, want ollama-cloud", got)
+	}
+
+	// applyEnv with a nil map is the exact path a child uses: it falls back to
+	// os.Getenv, so the inherited value selects the active provider.
+	cfg := &FileConfig{}
+	applyEnv(cfg, nil)
+	if cfg.ActiveProvider != "ollama-cloud" {
+		t.Fatalf("applyEnv ActiveProvider = %q, want ollama-cloud (inherited env not read)", cfg.ActiveProvider)
+	}
+}

internal/config/credentials.go

@@ -155,6 +155,63 @@ func MigratePlaintextProviderKeys(path string, store APIKeySetter) (int, error)
 	return migrated, nil
 }
 
+// HasConfiguredCredential reports whether the profile is set up to authenticate
+// with its own API key (inline, stored, or a raw auth header). It is the single
+// definition of "this profile is key-authed", shared by OAuthLoginCandidates and
+// the cli/tui setup surfaces so they never disagree about whether a profile is
+// keyed. APIKeyEnv is deliberately NOT included: an env var may be unset at
+// runtime while the profile actually relies on an OAuth login, so an
+// env-configured-but-empty profile must still be allowed to resolve a token.
+func (profile ProviderProfile) HasConfiguredCredential() bool {
+	return strings.TrimSpace(profile.APIKey) != "" ||
+		profile.APIKeyStored ||
+		strings.TrimSpace(profile.AuthHeaderValue) != ""
+}
+
+// OAuthLoginCandidates returns the login names to try, in order, when resolving
+// this profile's OAuth token — used by the runtime bearer resolver, the Codex
+// account-header resolver, and the onboarding presence check so all three agree.
+//
+// It returns NO candidates for a profile that carries its own configured
+// credential: attaching an OAuth resolver to such a profile makes withBearer
+// erase the key the instant a token resolves, silently routing/billing the call
+// through the OAuth account instead of the intended key (the cross-profile
+// override the review caught — and, via the profile name, a same-name login
+// overriding an explicit key). A stored-key profile that reaches this keyless
+// (e.g. a transiently unreadable keyring) is likewise treated as key-auth, so it
+// fails with a clear missing-credential error rather than silently borrowing an
+// unrelated OAuth login. APIKeyEnv is intentionally not part of that gate (an
+// env var may be unset while the profile relies on OAuth).
+//
+// For a keyless profile the profile name comes first (a login under your own
+// profile name is unambiguously yours), then the catalog ID as a FALLBACK (so a
+// profile renamed away from its catalog ID — e.g. {name:"codex",
+// catalogID:"chatgpt"} — still finds a `zero auth login chatgpt` token).
+// Candidates are trimmed, blank-skipped, and de-duplicated CASE-SENSITIVELY — the
+// OAuth token store is a case-sensitive map, so "ChatGPT" and "chatgpt" are
+// distinct keys and both must survive.
+func (profile ProviderProfile) OAuthLoginCandidates() []string {
+	if profile.HasConfiguredCredential() {
+		return nil
+	}
+	var names []string
+	add := func(s string) {
+		s = strings.TrimSpace(s)
+		if s == "" {
+			return
+		}
+		for _, existing := range names {
+			if existing == s { // case-sensitive: the store lookup is case-sensitive
+				return
+			}
+		}
+		names = append(names, s)
+	}
+	add(profile.Name)
+	add(profile.CatalogID)
+	return names
+}
+
 // ApplyStoredAPIKey fills profile.APIKey from the credential store when it is not
 // already resolved — an inline config key or a resolved APIKeyEnv always wins, so
 // this only supplies a key for providers whose secret lives in the store. A nil

internal/config/credentials_test.go

@@ -172,3 +172,69 @@ func TestProviderProfileAPIKeyStoredRoundTrips(t *testing.T) {
 		t.Fatalf("no inline key expected, got %q", p.APIKey)
 	}
 }
+
+// OAuthLoginCandidates always offers the profile name, adds the catalog ID as a
+// fallback ONLY when the profile has no effective own credential, and dedupes
+// case-sensitively (the OAuth store is a case-sensitive map).
+func TestOAuthLoginCandidates(t *testing.T) {
+	cases := []struct {
+		name    string
+		profile ProviderProfile
+		want    []string
+	}{
+		{
+			name:    "renamed keyless profile falls back to catalog id",
+			profile: ProviderProfile{Name: "codex", CatalogID: "chatgpt"},
+			want:    []string{"codex", "chatgpt"},
+		},
+		{
+			name:    "case-variant name keeps distinct catalog-id candidate",
+			profile: ProviderProfile{Name: "ChatGPT", CatalogID: "chatgpt"},
+			want:    []string{"ChatGPT", "chatgpt"},
+		},
+		{
+			name:    "exact-duplicate name and catalog id collapse",
+			profile: ProviderProfile{Name: "chatgpt", CatalogID: "chatgpt"},
+			want:    []string{"chatgpt"},
+		},
+		{
+			// A configured key must block ALL candidates (name included): a login
+			// under the profile's own name would otherwise erase the key too.
+			name:    "own inline key blocks every candidate",
+			profile: ProviderProfile{Name: "anthropic-work", CatalogID: "anthropic", APIKey: "sk-work"},
+			want:    nil,
+		},
+		{
+			name:    "own auth header blocks every candidate",
+			profile: ProviderProfile{Name: "acme", CatalogID: "anthropic", AuthHeaderValue: "Bearer x"},
+			want:    nil,
+		},
+		{
+			// APIKeyStored means key-auth even if the key isn't loaded here (e.g. a
+			// transiently unreadable keyring): don't silently borrow an OAuth login.
+			name:    "stored-key profile blocks every candidate",
+			profile: ProviderProfile{Name: "openai-stored", CatalogID: "openai", APIKeyStored: true},
+			want:    nil,
+		},
+		{
+			// APIKeyEnv is NOT a configured credential for gating: the env var may be
+			// unset while the profile relies on an OAuth login, so candidates stand.
+			name:    "env-only profile still yields candidates",
+			profile: ProviderProfile{Name: "xai", CatalogID: "xai", APIKeyEnv: "XAI_API_KEY"},
+			want:    []string{"xai"},
+		},
+		{
+			name:    "empty profile yields no candidates",
+			profile: ProviderProfile{},
+			want:    nil,
+		},
+	}
+	for _, c := range cases {
+		t.Run(c.name, func(t *testing.T) {
+			got := c.profile.OAuthLoginCandidates()
+			if strings.Join(got, ",") != strings.Join(c.want, ",") {
+				t.Fatalf("OAuthLoginCandidates() = %#v, want %#v", got, c.want)
+			}
+		})
+	}
+}

internal/config/max_turns_env_test.go

@@ -0,0 +1,56 @@
+package config
+
+import (
+	"os"
+	"testing"
+)
+
+// SetMaxTurnsEnv must export the per-run turn budget so a child process (whose
+// resolution reads the inherited environment via the nil-map -> os.Getenv path in
+// applyEnv) runs with the SAME budget the user set via /turns. n <= 0 is a no-op,
+// and a non-numeric/zero env value must not clobber the configured budget.
+func TestSetMaxTurnsEnvRoundTrips(t *testing.T) {
+	t.Setenv(MaxTurnsEnv, "") // register cleanup; restores the original after
+
+	if err := os.Unsetenv(MaxTurnsEnv); err != nil {
+		t.Fatalf("unset: %v", err)
+	}
+	SetMaxTurnsEnv(0) // non-positive must not set anything
+	if got := os.Getenv(MaxTurnsEnv); got != "" {
+		t.Fatalf("n<=0 set the env to %q, want unset", got)
+	}
+
+	SetMaxTurnsEnv(150)
+	if got := os.Getenv(MaxTurnsEnv); got != "150" {
+		t.Fatalf("env = %q, want 150", got)
+	}
+
+	// applyEnv with a nil map is the exact path a child uses: it falls back to
+	// os.Getenv, so the inherited value sets the budget.
+	cfg := &FileConfig{MaxTurns: defaultMaxTurns}
+	applyEnv(cfg, nil)
+	if cfg.MaxTurns != 150 {
+		t.Fatalf("applyEnv MaxTurns = %d, want 150 (inherited env not read)", cfg.MaxTurns)
+	}
+
+	// A garbage env value must leave the configured budget untouched.
+	if err := os.Setenv(MaxTurnsEnv, "not-a-number"); err != nil {
+		t.Fatalf("set: %v", err)
+	}
+	cfg2 := &FileConfig{MaxTurns: defaultMaxTurns}
+	applyEnv(cfg2, nil)
+	if cfg2.MaxTurns != defaultMaxTurns {
+		t.Fatalf("garbage env clobbered MaxTurns to %d, want default %d", cfg2.MaxTurns, defaultMaxTurns)
+	}
+
+	// An over-ceiling env value (e.g. a raw shell export that bypasses the /turns
+	// clamp) must be clamped to MaxTurnsCeiling at the read site.
+	if err := os.Setenv(MaxTurnsEnv, "999999"); err != nil {
+		t.Fatalf("set: %v", err)
+	}
+	cfg3 := &FileConfig{MaxTurns: defaultMaxTurns}
+	applyEnv(cfg3, nil)
+	if cfg3.MaxTurns != MaxTurnsCeiling {
+		t.Fatalf("applyEnv MaxTurns = %d, want clamped to ceiling %d", cfg3.MaxTurns, MaxTurnsCeiling)
+	}
+}

internal/config/paths.go

@@ -4,6 +4,7 @@ import (
 	"fmt"
 	"os"
 	"path/filepath"
+	"runtime"
 	"strings"
 )
 
@@ -33,13 +34,32 @@ func DefaultResolveOptions(workspaceRoot string) (ResolveOptions, error) {
 }
 
 func DefaultUserConfigPath() (string, error) {
-	userConfigDir, err := os.UserConfigDir()
+	userConfigDir, err := UserConfigDir()
 	if err != nil {
 		return "", fmt.Errorf("resolve user config directory: %w", err)
 	}
 	return filepath.Join(userConfigDir, "zero", "config.json"), nil
 }
 
+// UserConfigDir returns the base directory Zero stores per-user config under.
+// It mirrors os.UserConfigDir everywhere except macOS: there Go defaults to
+// ~/Library/Application Support, but Zero deliberately uses ~/.config (XDG-style,
+// matching Linux and Claude Code) so a single config path works cross-platform.
+// $XDG_CONFIG_HOME still wins on macOS when set.
+func UserConfigDir() (string, error) {
+	if runtime.GOOS == "darwin" {
+		if xdg := strings.TrimSpace(os.Getenv("XDG_CONFIG_HOME")); xdg != "" {
+			return xdg, nil
+		}
+		home, err := os.UserHomeDir()
+		if err != nil {
+			return "", fmt.Errorf("resolve home directory: %w", err)
+		}
+		return filepath.Join(home, ".config"), nil
+	}
+	return os.UserConfigDir()
+}
+
 func existingConfigFile(path string) (string, error) {
 	info, err := os.Stat(path)
 	if err != nil {

internal/config/paths_test.go

@@ -105,7 +105,7 @@ func setUserConfigRoot(t *testing.T) string {
 		t.Setenv("XDG_CONFIG_HOME", root)
 	}
 
-	configRoot, err := os.UserConfigDir()
+	configRoot, err := UserConfigDir()
 	if err != nil {
 		t.Fatalf("UserConfigDir() error = %v", err)
 	}
@@ -130,3 +130,21 @@ func mkdirAll(t *testing.T, path string) {
 		t.Fatalf("create directory %s: %v", path, err)
 	}
 }
+
+func TestDefaultUserConfigPathUsesXDGConfigOnMacOS(t *testing.T) {
+	if runtime.GOOS != "darwin" {
+		t.Skip("macOS-specific config path behavior")
+	}
+	home := t.TempDir()
+	t.Setenv("HOME", home)
+	t.Setenv("XDG_CONFIG_HOME", "")
+
+	path, err := DefaultUserConfigPath()
+	if err != nil {
+		t.Fatalf("DefaultUserConfigPath() error = %v", err)
+	}
+	want := filepath.Join(home, ".config", "zero", "config.json")
+	if path != want {
+		t.Fatalf("DefaultUserConfigPath() = %q, want %q", path, want)
+	}
+}

internal/config/resolver.go

@@ -2,9 +2,11 @@ package config
 
 import (
 	"encoding/json"
+	"errors"
 	"fmt"
 	"net/url"
 	"os"
+	"strconv"
 	"strings"
 
 	"github.com/Gitlawb/zero/internal/modelregistry"
@@ -13,7 +15,40 @@ import (
 	"github.com/Gitlawb/zero/internal/sandbox"
 )
 
-const defaultMaxTurns = 30
+// ErrNoActiveProvider marks a resolve failure caused solely by a missing or
+// unresolvable active provider. The interactive TUI treats this as "needs
+// onboarding" (drop into the setup wizard) rather than a fatal config error.
+var ErrNoActiveProvider = errors.New("no active provider configured")
+
+// ErrProviderRequiresModel marks a resolve failure caused solely by the active
+// provider missing a model with no catalog default to fall back on (custom
+// openai-/anthropic-compatible endpoints — Zero cannot guess a gateway's model).
+// Like ErrNoActiveProvider, the interactive TUI treats it as "needs onboarding"
+// and drops into the setup wizard so the user can fix it; headless commands
+// (zero config, zero exec) still fail with the actionable message.
+var ErrProviderRequiresModel = errors.New("provider requires model")
+
+// setupFixableError tags an error with a sentinel for errors.Is WITHOUT
+// changing its message: the sentinel's own text must not prefix what the user
+// sees (the wrapped message is already complete and actionable).
+type setupFixableError struct {
+	err      error
+	sentinel error
+}
+
+func (e *setupFixableError) Error() string { return e.err.Error() }
+
+func (e *setupFixableError) Unwrap() []error { return []error{e.err, e.sentinel} }
+
+// defaultMaxTurns is the per-run tool-turn budget when none is configured. 30 was
+// too low for real multi-step agentic work (agents ran out mid-task before reaching
+// later steps); 50 matches the old "deep" preset. Raise per-session with /turns.
+const defaultMaxTurns = 50
+
+// MaxTurnsCeiling caps the per-run tool-turn budget so a stray env value or typo
+// can't set an absurd ceiling. Shared between applyEnv (read site) and the /turns
+// command (write site) so the bound holds even if the env is set by a raw shell.
+const MaxTurnsCeiling = 500
 
 // defaultDeferThreshold is the number of deferred-eligible (MCP) tools at which
 // Zero collapses their full JSON schemas into compact `tool_search` reminder
@@ -95,7 +130,11 @@ func Resolve(options ResolveOptions) (ResolvedConfig, error) {
 
 	providers, active, err := normalizeProviders(cfg.Providers, cfg.ActiveProvider, options.Env)
 	if err != nil {
-		return ResolvedConfig{}, err
+		// On ErrNoActiveProvider, providers may still hold the successfully
+		// normalized (but active-less) profile list — keep it so a caller can fall
+		// back to an already-configured usable provider instead of treating this
+		// like a config with nothing set up at all.
+		return ResolvedConfig{Providers: providers}, err
 	}
 
 	return ResolvedConfig{
@@ -109,6 +148,7 @@ func Resolve(options ResolveOptions) (ResolvedConfig, error) {
 		Tools:          cfg.Tools,
 		Swarm:          cfg.Swarm,
 		Preferences:    cfg.Preferences,
+		KeyBindings:    cfg.KeyBindings,
 		LocalControl:   cfg.LocalControl,
 	}, nil
 }
@@ -186,7 +226,11 @@ func mergeConfig(dst *FileConfig, src FileConfig) {
 	if src.Preferences.Recaps != nil {
 		dst.Preferences.Recaps = src.Preferences.Recaps
 	}
+	if strings.TrimSpace(src.Preferences.Theme) != "" {
+		dst.Preferences.Theme = strings.TrimSpace(src.Preferences.Theme)
+	}
 	mergeLocalControlConfig(&dst.LocalControl, src.LocalControl)
+	mergeKeyBindings(&dst.KeyBindings, src.KeyBindings)
 }
 
 func mergeProjectConfig(dst *FileConfig, src FileConfig) error {
@@ -208,8 +252,17 @@ func mergeProjectConfig(dst *FileConfig, src FileConfig) error {
 	// config: a cloned repo's .zero/config.json must not be able to grant
 	// itself write access outside the workspace. Global config and CLI flags
 	// are the only config sources for write roots.
+	//
+	// Sandbox.Network from project config may only TIGHTEN (→ "deny"), never
+	// WEAKEN (→ "allow"). A malicious repo must not be able to open network
+	// access for shell commands. Matches the AdditionalWriteRoots posture.
 	if network := strings.TrimSpace(src.Sandbox.Network); network != "" {
-		dst.Sandbox.Network = network
+		// Normalize: anything that isn't literally "allow" counts as deny.
+		if sandbox.NormalizeNetworkMode(sandbox.NetworkMode(network)) != sandbox.NetworkAllow {
+			dst.Sandbox.Network = network
+		}
+		// Silently ignore "allow" from project config — not an error, just
+		// a privilege the project scope does not have.
 	}
 	if src.Sandbox.BlockUnixSockets {
 		dst.Sandbox.BlockUnixSockets = true
@@ -230,6 +283,7 @@ func mergeProjectConfig(dst *FileConfig, src FileConfig) error {
 	if src.Swarm.MaxTeamSize != 0 {
 		dst.Swarm.MaxTeamSize = src.Swarm.MaxTeamSize
 	}
+	mergeKeyBindings(&dst.KeyBindings, src.KeyBindings)
 	// Local control is intentionally user-config/override only. A cloned project
 	// must not be able to make browser, desktop, or terminal automation tools
 	// appear in the model's tool surface.
@@ -358,6 +412,18 @@ func effectiveProviderKind(profile ProviderProfile) ProviderKind {
 	return ""
 }
 
+// catalogDefaultModel returns the default model of a catalog provider ("" when
+// the id is unknown), used to fill a missing profile.Model for the official-API
+// kinds so a hand-written profile without a model resolves instead of bricking
+// every command that resolves config up front (zero config, bare zero setup).
+func catalogDefaultModel(catalogID string) string {
+	descriptor, ok := providercatalog.Get(catalogID)
+	if !ok {
+		return ""
+	}
+	return strings.TrimSpace(descriptor.DefaultModel)
+}
+
 func catalogDescriptorForProfile(profile ProviderProfile) (providercatalog.Descriptor, bool) {
 	catalogID := providercatalog.NormalizeID(profile.CatalogID)
 	if catalogID == "" {
@@ -431,11 +497,53 @@ func mergeProfile(base ProviderProfile, next ProviderProfile) ProviderProfile {
 	return base
 }
 
+// ActiveProviderEnv selects the active provider profile by name (read in applyEnv).
+const ActiveProviderEnv = "ZERO_PROVIDER"
+
+// SetActiveProviderEnv exports the active provider name to the process environment
+// so a spawned child process (which inherits the environment) resolves the SAME
+// provider profile — and therefore the same credentials (env key / stored key /
+// OAuth) — as its parent. Without this a sub-agent re-resolves config.json's
+// default provider and can land on one whose credentials don't match the parent's
+// live selection, failing auth the instant it spawns. A blank name CLEARS the
+// variable: switching back to an unnamed/default profile must not keep exporting a
+// stale provider to children.
+func SetActiveProviderEnv(name string) {
+	name = strings.TrimSpace(name)
+	if name == "" {
+		_ = os.Unsetenv(ActiveProviderEnv)
+		return
+	}
+	_ = os.Setenv(ActiveProviderEnv, name)
+}
+
+// MaxTurnsEnv overrides the per-run tool-turn budget by name (read in applyEnv).
+const MaxTurnsEnv = "ZERO_MAX_TURNS"
+
+// SetMaxTurnsEnv exports the per-run tool-turn budget to the process environment so
+// a spawned child (sub-agent / swarm member, which inherits the environment) runs
+// with the SAME budget the user set via /turns. Without it a child re-resolves
+// config.json's default and a large delegated task can exhaust its turns mid-run
+// (exit 4 / max-turns). No-op for n <= 0.
+func SetMaxTurnsEnv(n int) {
+	if n > 0 {
+		_ = os.Setenv(MaxTurnsEnv, strconv.Itoa(n))
+	}
+}
+
 func applyEnv(cfg *FileConfig, env map[string]string) {
-	activeProvider := strings.TrimSpace(envValue(env, "ZERO_PROVIDER"))
+	activeProvider := strings.TrimSpace(envValue(env, ActiveProviderEnv))
 	if activeProvider != "" {
 		cfg.ActiveProvider = activeProvider
 	}
+	if maxTurns := strings.TrimSpace(envValue(env, MaxTurnsEnv)); maxTurns != "" {
+		if n, err := strconv.Atoi(maxTurns); err == nil && n > 0 {
+			if n > MaxTurnsCeiling {
+				n = MaxTurnsCeiling
+			}
+			cfg.MaxTurns = n
+		}
+	}
 
 	applyProviderEnv(cfg, ProviderKindOpenAI, envProfile{
 		Name:    string(ProviderKindOpenAI),
@@ -482,9 +590,17 @@ func applyProviderEnv(cfg *FileConfig, providerKind ProviderKind, env envProfile
 	if profile.Name == "" {
 		profile.Name = string(providerKind)
 	}
+	// A non-official baseURL from the environment signals a proxy/gateway, so promote
+	// the first-party kind to its -compatible transport (which accepts a custom URL).
+	// Env-only by design: a custom baseURL in a project config.json is left as-is and
+	// rejected by Resolve, so an untrusted project can't redirect a first-party key to
+	// another host.
 	if providerKind == ProviderKindOpenAI && baseURL != "" && !isOfficialOpenAIBaseURL(baseURL) {
 		profile.ProviderKind = ProviderKindOpenAICompatible
 	}
+	if providerKind == ProviderKindAnthropic && baseURL != "" && !isOfficialAnthropicBaseURL(baseURL) {
+		profile.ProviderKind = ProviderKindAnthropicCompat
+	}
 	// When the env supplies only credentials (no baseURL) for a provider name that
 	// already exists, don't force the standard transport kind — a same-named
 	// openai-compatible/anthropic-compatible proxy or gateway must keep its kind.
@@ -564,6 +680,7 @@ func applyOverrides(cfg *FileConfig, overrides Overrides) {
 		cfg.Tools.deferThresholdSet = true
 	}
 	mergeLocalControlConfig(&cfg.LocalControl, overrides.LocalControl)
+	mergeKeyBindings(&cfg.KeyBindings, overrides.KeyBindings)
 	for _, provider := range overrides.Providers {
 		mergeProvider(cfg, provider)
 	}
@@ -599,6 +716,24 @@ func mergeLocalControlDriverConfig(dst *LocalControlDriverConfig, src LocalContr
 	}
 }
 
+func mergeKeyBindings(dst *KeyBindingsConfig, src KeyBindingsConfig) {
+	if src.ToggleDetailed != "" {
+		dst.ToggleDetailed = src.ToggleDetailed
+	}
+	if src.ToggleMouse != "" {
+		dst.ToggleMouse = src.ToggleMouse
+	}
+	if src.CycleReasoning != "" {
+		dst.CycleReasoning = src.CycleReasoning
+	}
+	if src.TogglePlan != "" {
+		dst.TogglePlan = src.TogglePlan
+	}
+	if src.ToggleSidebar != "" {
+		dst.ToggleSidebar = src.ToggleSidebar
+	}
+}
+
 func mergeMCPConfig(dst *MCPConfig, src MCPConfig) {
 	if len(src.Servers) == 0 {
 		return
@@ -677,11 +812,16 @@ func unionStrings(base []string, extra []string) []string {
 }
 
 type normalizeOptions struct {
-	defaultOpenAIModel bool
+	// defaultModels fills a missing profile.Model for the official-API provider
+	// kinds (openai from the model registry; anthropic/google from their catalog
+	// descriptors). Off for provider-command configs (command.go), which must
+	// surface exactly what the external command returned rather than inventing
+	// a model behind its back.
+	defaultModels bool
 }
 
 func normalizeProviders(providers []ProviderProfile, activeName string, envMaps ...map[string]string) ([]ProviderProfile, ProviderProfile, error) {
-	return normalizeProvidersWithOptions(providers, activeName, normalizeOptions{defaultOpenAIModel: true}, envMaps...)
+	return normalizeProvidersWithOptions(providers, activeName, normalizeOptions{defaultModels: true}, envMaps...)
 }
 
 func normalizeProvidersWithoutModelDefaults(providers []ProviderProfile, activeName string, envMaps ...map[string]string) ([]ProviderProfile, ProviderProfile, error) {
@@ -696,7 +836,7 @@ func normalizeProvidersWithOptions(providers []ProviderProfile, activeName strin
 	}
 	if len(providers) == 0 {
 		if activeName != "" {
-			return nil, ProviderProfile{}, fmt.Errorf("active provider %q not found", activeName)
+			return nil, ProviderProfile{}, fmt.Errorf("%w: active provider %q not found", ErrNoActiveProvider, activeName)
 		}
 		return []ProviderProfile{}, ProviderProfile{}, nil
 	}
@@ -728,10 +868,17 @@ func normalizeProvidersWithOptions(providers []ProviderProfile, activeName strin
 	}
 
 	if !activeFound {
-		return nil, ProviderProfile{}, fmt.Errorf("active provider %q not found", activeName)
+		// Return the successfully normalized list alongside the error (rather than
+		// nil) so a caller like the interactive TUI can still fall back to an
+		// already-configured, usable provider instead of forcing a full
+		// re-onboarding wizard just because none was marked active.
+		return normalized, ProviderProfile{}, fmt.Errorf("%w: active provider %q not found", ErrNoActiveProvider, activeName)
 	}
 	if active.Model == "" {
-		return nil, ProviderProfile{}, providerError(active, "provider %s requires model", active.Name)
+		return nil, ProviderProfile{}, &setupFixableError{
+			err:      providerError(active, "provider %s requires model — add \"model\" to its entry in config.json, or re-run: zero setup <catalog-id> --model <model>", active.Name),
+			sentinel: ErrProviderRequiresModel,
+		}
 	}
 
 	return normalized, active, nil
@@ -778,7 +925,7 @@ func normalizeProvider(profile ProviderProfile, env map[string]string, options n
 	case ProviderKindOpenAI:
 		if profile.BaseURL == "" || isOfficialOpenAIBaseURL(profile.BaseURL) {
 			profile.BaseURL = OpenAIBaseURL
-			if options.defaultOpenAIModel && profile.Model == "" {
+			if options.defaultModels && profile.Model == "" {
 				profile.Model = modelregistry.DefaultModelID
 			}
 			return profile, nil
@@ -795,6 +942,9 @@ func normalizeProvider(profile ProviderProfile, env map[string]string, options n
 	case ProviderKindAnthropic:
 		if profile.BaseURL == "" || isOfficialAnthropicBaseURL(profile.BaseURL) {
 			profile.BaseURL = AnthropicBaseURL
+			if options.defaultModels && profile.Model == "" {
+				profile.Model = catalogDefaultModel("anthropic")
+			}
 			return profile, nil
 		}
 		return ProviderProfile{}, providerError(profile, "anthropic provider %s requires official baseURL %s", profile.Name, AnthropicBaseURL)
@@ -809,6 +959,9 @@ func normalizeProvider(profile ProviderProfile, env map[string]string, options n
 	case ProviderKindGoogle:
 		if profile.BaseURL == "" || isOfficialGoogleBaseURL(profile.BaseURL) {
 			profile.BaseURL = GoogleBaseURL
+			if options.defaultModels && profile.Model == "" {
+				profile.Model = catalogDefaultModel("google")
+			}
 			return profile, nil
 		}
 		return ProviderProfile{}, providerError(profile, "google provider %s requires official baseURL %s", profile.Name, GoogleBaseURL)

internal/config/resolver_test.go

@@ -1,6 +1,7 @@
 package config
 
 import (
+	"errors"
 	"os"
 	"path/filepath"
 	"reflect"
@@ -616,6 +617,35 @@ func TestResolveUsesOpenAIEnvFallback(t *testing.T) {
 	}
 }
 
+func TestResolveUsesAnthropicEnvBaseURLAsCompatible(t *testing.T) {
+	// ANTHROPIC_BASE_URL pointing at a proxy/gateway must resolve to an
+	// anthropic-compatible provider (mirroring OPENAI_BASE_URL), not the "requires
+	// official baseURL" rejection — issue #479. The env is user-controlled, so a
+	// custom URL there is a deliberate gateway choice, unlike a project config.json.
+	resolved, err := Resolve(ResolveOptions{
+		Env: map[string]string{
+			"ANTHROPIC_API_KEY":  "sk-ant-env",
+			"ANTHROPIC_BASE_URL": "https://gateway.example/anthropic",
+			"ANTHROPIC_MODEL":    "claude-custom",
+		},
+	})
+	if err != nil {
+		t.Fatalf("Resolve() error = %v", err)
+	}
+	if resolved.ActiveProvider != "anthropic" {
+		t.Fatalf("ActiveProvider = %q, want anthropic", resolved.ActiveProvider)
+	}
+	if resolved.Provider.ProviderKind != ProviderKindAnthropicCompat {
+		t.Fatalf("ProviderKind = %q, want anthropic-compatible", resolved.Provider.ProviderKind)
+	}
+	if resolved.Provider.BaseURL != "https://gateway.example/anthropic" {
+		t.Fatalf("BaseURL = %q, want gateway URL", resolved.Provider.BaseURL)
+	}
+	if resolved.Provider.APIKey != "sk-ant-env" || resolved.Provider.Model != "claude-custom" {
+		t.Fatalf("Provider = %#v, want env credentials/model", resolved.Provider)
+	}
+}
+
 func TestResolveUsesOpenAIAPIKeyOnlyWithDefaultModel(t *testing.T) {
 	resolved, err := Resolve(ResolveOptions{
 		Env: map[string]string{
@@ -883,8 +913,8 @@ func TestResolveAllowsNoConfiguredProviders(t *testing.T) {
 	if HasProviderProfile(resolved.Provider) {
 		t.Fatalf("Provider = %#v, want zero value", resolved.Provider)
 	}
-	if resolved.MaxTurns != 30 {
-		t.Fatalf("MaxTurns = %d, want 30", resolved.MaxTurns)
+	if resolved.MaxTurns != defaultMaxTurns {
+		t.Fatalf("MaxTurns = %d, want default %d", resolved.MaxTurns, defaultMaxTurns)
 	}
 }
 
@@ -898,6 +928,9 @@ func TestResolveRejectsActiveProviderWithoutConfiguredProfiles(t *testing.T) {
 	if !strings.Contains(err.Error(), `active provider "ghost" not found`) {
 		t.Fatalf("error = %q, want active provider missing message", err.Error())
 	}
+	if !errors.Is(err, ErrNoActiveProvider) {
+		t.Fatalf("error = %v, want errors.Is(err, ErrNoActiveProvider)", err)
+	}
 	if resolved.ActiveProvider != "" {
 		t.Fatalf("ActiveProvider = %q, want empty", resolved.ActiveProvider)
 	}
@@ -912,6 +945,25 @@ func TestResolveRejectsActiveProviderWithoutConfiguredProfiles(t *testing.T) {
 	}
 }
 
+func TestResolveKeepsNormalizedProvidersWhenNoneMarkedActive(t *testing.T) {
+	// Multiple providers configured (e.g. via `zero provider add`) but
+	// activeProvider is blank/stale — a caller like the interactive TUI still
+	// needs the normalized list to fall back to an already-usable provider
+	// instead of forcing a full re-onboarding wizard.
+	path := writeConfig(t, `{"providers":[
+		{"name":"work","provider_kind":"openai","apiKey":"sk-test","model":"gpt-test"},
+		{"name":"other","provider_kind":"openai","apiKey":"sk-other","model":"gpt-test"}
+	]}`)
+
+	resolved, err := Resolve(ResolveOptions{ProjectConfigPath: path, Env: map[string]string{}})
+	if !errors.Is(err, ErrNoActiveProvider) {
+		t.Fatalf("error = %v, want errors.Is(err, ErrNoActiveProvider)", err)
+	}
+	if len(resolved.Providers) != 2 {
+		t.Fatalf("Providers = %#v, want the 2 normalized profiles preserved despite the error", resolved.Providers)
+	}
+}
+
 func TestResolveTrimsProviderProfileAliasesBeforeFallback(t *testing.T) {
 	path := writeConfig(t, `{
 		"activeProvider": "custom",
@@ -1055,6 +1107,130 @@ func TestResolveAppliesProviderCatalogDefaults(t *testing.T) {
 	}
 }
 
+func TestResolveAppliesMiniMaxCNCatalogDefaults(t *testing.T) {
+	path := writeConfig(t, `{
+		"activeProvider": "mini-cn",
+		"providers": [{
+			"name": "mini-cn",
+			"catalog_id": "minimaxi-cn"
+		}]
+	}`)
+
+	resolved, err := Resolve(ResolveOptions{
+		ProjectConfigPath: path,
+		Env: map[string]string{
+			"MINIMAXI_API_KEY": "sk-cn-mini",
+		},
+	})
+	if err != nil {
+		t.Fatalf("Resolve() error = %v", err)
+	}
+
+	if resolved.Provider.CatalogID != "minimaxi-cn" {
+		t.Fatalf("CatalogID = %q, want minimaxi-cn", resolved.Provider.CatalogID)
+	}
+	if resolved.Provider.ProviderKind != ProviderKindAnthropicCompat {
+		t.Fatalf("ProviderKind = %q, want %q", resolved.Provider.ProviderKind, ProviderKindAnthropicCompat)
+	}
+	if resolved.Provider.BaseURL != "https://api.minimaxi.com/anthropic" {
+		t.Fatalf("BaseURL = %q, want MiniMax CN default", resolved.Provider.BaseURL)
+	}
+	if resolved.Provider.Model != "MiniMax-M3" {
+		t.Fatalf("Model = %q, want MiniMax-M3", resolved.Provider.Model)
+	}
+	if resolved.Provider.APIKeyEnv != "MINIMAXI_API_KEY" {
+		t.Fatalf("APIKeyEnv = %q, want MINIMAXI_API_KEY", resolved.Provider.APIKeyEnv)
+	}
+	if resolved.Provider.APIKey != "sk-cn-mini" {
+		t.Fatalf("APIKey = %q, want env-resolved secret", resolved.Provider.APIKey)
+	}
+}
+
+func TestResolveAppliesZaiCNCatalogDefaults(t *testing.T) {
+	// The "zai-cn" catalog entry preserves the legacy open.bigmodel.cn endpoint
+	// for users who still target the China hosting; "zai" itself now points at
+	// api.z.ai (international). The two env-vars are intentionally distinct
+	// (ZHIPU_API_KEY for China, ZAI_API_KEY for international) so an account on
+	// one side cannot be silently used against the other.
+	path := writeConfig(t, `{
+		"activeProvider": "zai-cn",
+		"providers": [{
+			"name": "zai-cn",
+			"catalog_id": "zai-cn"
+		}]
+	}`)
+
+	resolved, err := Resolve(ResolveOptions{
+		ProjectConfigPath: path,
+		Env: map[string]string{
+			"ZHIPU_API_KEY": "sk-zhipu-cn",
+		},
+	})
+	if err != nil {
+		t.Fatalf("Resolve() error = %v", err)
+	}
+
+	if resolved.Provider.CatalogID != "zai-cn" {
+		t.Fatalf("CatalogID = %q, want zai-cn", resolved.Provider.CatalogID)
+	}
+	if resolved.Provider.ProviderKind != ProviderKindOpenAICompatible {
+		t.Fatalf("ProviderKind = %q, want %q", resolved.Provider.ProviderKind, ProviderKindOpenAICompatible)
+	}
+	if resolved.Provider.BaseURL != "https://open.bigmodel.cn/api/paas/v4" {
+		t.Fatalf("BaseURL = %q, want Z.ai China default", resolved.Provider.BaseURL)
+	}
+	if resolved.Provider.Model != "glm-4.5" {
+		t.Fatalf("Model = %q, want glm-4.5 (China default)", resolved.Provider.Model)
+	}
+	if resolved.Provider.APIKeyEnv != "ZHIPU_API_KEY" {
+		t.Fatalf("APIKeyEnv = %q, want ZHIPU_API_KEY", resolved.Provider.APIKeyEnv)
+	}
+	if resolved.Provider.APIKey != "sk-zhipu-cn" {
+		t.Fatalf("APIKey = %q, want env-resolved secret", resolved.Provider.APIKey)
+	}
+}
+
+func TestResolveAppliesZaiInternationalCatalogDefaults(t *testing.T) {
+	// "zai" now resolves to the international endpoint; the catalog default
+	// model stays glm-4.5 (live discovery upgrades to glm-5.2 at runtime).
+	path := writeConfig(t, `{
+		"activeProvider": "zai-intl",
+		"providers": [{
+			"name": "zai-intl",
+			"catalog_id": "zai"
+		}]
+	}`)
+
+	resolved, err := Resolve(ResolveOptions{
+		ProjectConfigPath: path,
+		Env: map[string]string{
+			"ZAI_API_KEY": "sk-zai-intl",
+		},
+	})
+	if err != nil {
+		t.Fatalf("Resolve() error = %v", err)
+	}
+
+	if resolved.Provider.CatalogID != "zai" {
+		t.Fatalf("CatalogID = %q, want zai", resolved.Provider.CatalogID)
+	}
+	if resolved.Provider.ProviderKind != ProviderKindOpenAICompatible {
+		t.Fatalf("ProviderKind = %q, want %q", resolved.Provider.ProviderKind, ProviderKindOpenAICompatible)
+	}
+	if resolved.Provider.BaseURL != "https://api.z.ai/api/paas/v4" {
+		t.Fatalf("BaseURL = %q, want Z.ai international default", resolved.Provider.BaseURL)
+	}
+	if resolved.Provider.Model != "glm-4.5" {
+		t.Fatalf("Model = %q, want glm-4.5 (preserved default; live discovery upgrades to glm-5.2 at runtime)", resolved.Provider.Model)
+	}
+	if resolved.Provider.APIKeyEnv != "ZAI_API_KEY" {
+		t.Fatalf("APIKeyEnv = %q, want ZAI_API_KEY", resolved.Provider.APIKeyEnv)
+	}
+	if resolved.Provider.APIKey != "sk-zai-intl" {
+		t.Fatalf("APIKey = %q, want env-resolved secret", resolved.Provider.APIKey)
+	}
+}
+
 func TestResolveProviderCatalogAliasAndLocalNoAuth(t *testing.T) {
 	path := writeConfig(t, `{
 		"activeProvider": "local",
@@ -1269,6 +1445,45 @@ func TestResolveSandboxBlockUnixSocketsFromFile(t *testing.T) {
 	}
 }
 
+func TestResolveSandboxNetworkProjectConfigCannotWeaken(t *testing.T) {
+	// 1. Project config tries to set network to "allow" (should be ignored).
+	userPath := writeConfig(t, `{}`)
+	projectPath := writeConfig(t, `{
+		"sandbox": {"network": "allow"}
+	}`)
+
+	resolved, err := Resolve(ResolveOptions{
+		UserConfigPath:    userPath,
+		ProjectConfigPath: projectPath,
+		Env:               map[string]string{},
+	})
+	if err != nil {
+		t.Fatalf("Resolve() error = %v", err)
+	}
+	if resolved.Sandbox.Network == "allow" {
+		t.Fatal("resolved.Sandbox.Network = allow, want empty/deny to prevent project config from weakening sandbox")
+	}
+
+	// 2. User config sets network to "allow", project config sets it to "deny" (tightening is allowed).
+	userPath2 := writeConfig(t, `{
+		"sandbox": {"network": "allow"}
+	}`)
+	projectPath2 := writeConfig(t, `{
+		"sandbox": {"network": "deny"}
+	}`)
+	resolved2, err := Resolve(ResolveOptions{
+		UserConfigPath:    userPath2,
+		ProjectConfigPath: projectPath2,
+		Env:               map[string]string{},
+	})
+	if err != nil {
+		t.Fatalf("Resolve() error = %v", err)
+	}
+	if resolved2.Sandbox.Network != "deny" {
+		t.Fatalf("resolved.Sandbox.Network = %q, want deny (project config allowed to tighten)", resolved2.Sandbox.Network)
+	}
+}
+
 func TestResolveNotifyValid(t *testing.T) {
 	path := writeConfig(t, `{"notify":{"mode":"both","focusMode":"always"}}`)
 	resolved, err := Resolve(ResolveOptions{UserConfigPath: path, Env: map[string]string{}})
@@ -1551,3 +1766,97 @@ func TestResolveMaxTurnsZeroFallsBackToDefault(t *testing.T) {
 		t.Fatalf("MaxTurns = %d, want default %d", resolved.MaxTurns, defaultMaxTurns)
 	}
 }
+
+// The reported brick: a hand-written google profile with an apiKey but no
+// model made EVERY resolving command fail ("provider google requires model"),
+// including zero config and bare zero setup — the only commands that could
+// have fixed it. Official-API kinds now fall back to their catalog default
+// model, exactly like the openai kind always has.
+func TestResolveDefaultsGoogleModelFromCatalog(t *testing.T) {
+	path := writeConfig(t, `{
+		"activeProvider": "google",
+		"providers": [{"name": "google", "provider_kind": "google", "apiKey": "AIza-x"}]
+	}`)
+	resolved, err := Resolve(ResolveOptions{UserConfigPath: path, Env: map[string]string{}})
+	if err != nil {
+		t.Fatalf("Resolve() error = %v, want model defaulted from the google catalog", err)
+	}
+	if resolved.Provider.Model != "gemini-2.5-pro" {
+		t.Fatalf("Model = %q, want the google catalog default gemini-2.5-pro", resolved.Provider.Model)
+	}
+	if resolved.Provider.BaseURL != GoogleBaseURL {
+		t.Fatalf("BaseURL = %q, want %q", resolved.Provider.BaseURL, GoogleBaseURL)
+	}
+}
+
+func TestResolveDefaultsAnthropicModelFromCatalog(t *testing.T) {
+	path := writeConfig(t, `{
+		"activeProvider": "anthropic",
+		"providers": [{"name": "anthropic", "provider_kind": "anthropic", "apiKey": "sk-ant-x"}]
+	}`)
+	resolved, err := Resolve(ResolveOptions{UserConfigPath: path, Env: map[string]string{}})
+	if err != nil {
+		t.Fatalf("Resolve() error = %v, want model defaulted from the anthropic catalog", err)
+	}
+	if resolved.Provider.Model != "claude-sonnet-4.5" {
+		t.Fatalf("Model = %q, want the anthropic catalog default claude-sonnet-4.5", resolved.Provider.Model)
+	}
+}
+
+// An explicitly configured model must always win over the catalog default.
+func TestResolveKeepsExplicitGoogleModel(t *testing.T) {
+	path := writeConfig(t, `{
+		"activeProvider": "google",
+		"providers": [{"name": "google", "provider_kind": "google", "apiKey": "AIza-x", "model": "gemini-2.5-flash"}]
+	}`)
+	resolved, err := Resolve(ResolveOptions{UserConfigPath: path, Env: map[string]string{}})
+	if err != nil {
+		t.Fatalf("Resolve() error = %v", err)
+	}
+	if resolved.Provider.Model != "gemini-2.5-flash" {
+		t.Fatalf("Model = %q, want the explicitly configured gemini-2.5-flash", resolved.Provider.Model)
+	}
+}
+
+// Provider-command configs must NOT get invented models: the without-defaults
+// path surfaces exactly what the external command returned.
+func TestNormalizeWithoutModelDefaultsStillRequiresGoogleModel(t *testing.T) {
+	_, _, err := normalizeProvidersWithoutModelDefaults(
+		[]ProviderProfile{{Name: "google", ProviderKind: ProviderKindGoogle, APIKey: "AIza-x"}},
+		"google", map[string]string{})
+	if err == nil {
+		t.Fatal("provider-command path must still require an explicit model")
+	}
+	if !strings.Contains(err.Error(), "requires model") {
+		t.Fatalf("error = %q, want requires-model", err.Error())
+	}
+	// The residual error must tell the user how to fix it.
+	if !strings.Contains(err.Error(), "config.json") {
+		t.Fatalf("error = %q, want an actionable config.json hint", err.Error())
+	}
+}
+
+// The residual requires-model failure (custom endpoint, no catalog default)
+// must be tagged setup-fixable WITHOUT changing its message: the interactive
+// TUI matches the sentinel to drop into the wizard, while headless commands
+// print the exact actionable text.
+func TestResolveRequiresModelErrorIsSetupFixable(t *testing.T) {
+	path := writeConfig(t, `{
+		"activeProvider": "gw",
+		"providers": [{"name": "gw", "provider_kind": "openai-compatible", "baseURL": "https://gw.example/v1", "apiKey": "sk-x"}]
+	}`)
+	_, err := Resolve(ResolveOptions{UserConfigPath: path, Env: map[string]string{}})
+	if err == nil {
+		t.Fatal("expected requires-model error for a custom endpoint without model")
+	}
+	if !errors.Is(err, ErrProviderRequiresModel) {
+		t.Fatalf("error = %v, want errors.Is(err, ErrProviderRequiresModel)", err)
+	}
+	// The sentinel must NOT prefix the user-facing message.
+	if !strings.HasPrefix(err.Error(), "provider gw requires model") {
+		t.Fatalf("message = %q, want it to start with the provider error, not the sentinel text", err.Error())
+	}
+	if strings.Contains(err.Error(), "sk-x") {
+		t.Fatalf("error leaked API key: %q", err.Error())
+	}
+}

internal/config/types.go

@@ -97,6 +97,10 @@ type ToolsConfig struct {
 
 type PreferencesConfig struct {
 	FavoriteModels []string `json:"favoriteModels,omitempty"`
+	// Theme is the persisted TUI palette preference — "auto" or a registered theme
+	// name (e.g. "dracula"). Applied at startup below the --theme flag and
+	// ZERO_THEME, so a /theme choice survives restart. Empty = unset (defaults auto).
+	Theme string `json:"theme,omitempty"`
 	// Recaps is a tri-state: nil (unset) defaults to ON; an explicit false means
 	// the user turned post-turn recaps off. A *bool is its own tri-state, so no
 	// custom unmarshal is needed (unlike ToolsConfig.DeferThreshold's int).
@@ -108,6 +112,25 @@ func (p PreferencesConfig) RecapsEnabled() bool {
 	return p.Recaps == nil || *p.Recaps
 }
 
+// KeyBindingDef defines one key binding string (e.g. "ctrl+o") that the TUI
+// can remap. An empty string means "use the built-in default" for that action.
+type KeyBindingDef string
+
+// KeyBindingsConfig holds the subset of TUI keybindings that users may remap
+// via config.json. Each field defaults to a sensible built-in when empty.
+type KeyBindingsConfig struct {
+	// ToggleDetailed toggles the detailed transcript view (default: ctrl+o).
+	ToggleDetailed KeyBindingDef `json:"toggleDetailed,omitempty"`
+	// ToggleMouse toggles mouse capture release (default: ctrl+e).
+	ToggleMouse KeyBindingDef `json:"toggleMouse,omitempty"`
+	// CycleReasoning cycles through reasoning effort levels (default: ctrl+t).
+	CycleReasoning KeyBindingDef `json:"cycleReasoning,omitempty"`
+	// TogglePlan toggles the plan panel expansion (default: ctrl+p).
+	TogglePlan KeyBindingDef `json:"togglePlan,omitempty"`
+	// ToggleSidebar toggles the right context sidebar (default: ctrl+b).
+	ToggleSidebar KeyBindingDef `json:"toggleSidebar,omitempty"`
+}
+
 // LocalControlConfig controls local browser/desktop/terminal automation helpers.
 // Helpers are discovered lazily by the tool that needs them; no setup command or
 // background probe is required during startup.
@@ -202,6 +225,7 @@ type FileConfig struct {
 	Tools          ToolsConfig        `json:"tools,omitempty"`
 	Swarm          SwarmConfig        `json:"swarm,omitempty"`
 	Preferences    PreferencesConfig  `json:"preferences,omitempty"`
+	KeyBindings    KeyBindingsConfig  `json:"keybindings,omitempty"`
 	LocalControl   LocalControlConfig `json:"localControl,omitempty"`
 }
 
@@ -216,6 +240,7 @@ func (cfg FileConfig) MarshalJSON() ([]byte, error) {
 		Tools          ToolsConfig         `json:"tools,omitempty"`
 		Swarm          SwarmConfig         `json:"swarm,omitempty"`
 		Preferences    PreferencesConfig   `json:"preferences,omitempty"`
+		KeyBindings    KeyBindingsConfig   `json:"keybindings,omitempty"`
 		LocalControl   *LocalControlConfig `json:"localControl,omitempty"`
 	}
 	raw := rawConfig{
@@ -228,6 +253,7 @@ func (cfg FileConfig) MarshalJSON() ([]byte, error) {
 		Tools:          cfg.Tools,
 		Swarm:          cfg.Swarm,
 		Preferences:    cfg.Preferences,
+		KeyBindings:    cfg.KeyBindings,
 	}
 	if !cfg.LocalControl.Empty() {
 		raw.LocalControl = &cfg.LocalControl
@@ -252,6 +278,7 @@ type Overrides struct {
 	Sandbox        SandboxConfig
 	Notify         NotifyConfig
 	Tools          ToolsConfig
+	KeyBindings    KeyBindingsConfig
 	LocalControl   LocalControlConfig
 }
 
@@ -266,6 +293,7 @@ type ResolvedConfig struct {
 	Tools          ToolsConfig
 	Swarm          SwarmConfig
 	Preferences    PreferencesConfig
+	KeyBindings    KeyBindingsConfig
 	LocalControl   LocalControlConfig
 }
 
@@ -312,6 +340,7 @@ func (cfg *FileConfig) UnmarshalJSON(data []byte) error {
 		Tools           ToolsConfig                `json:"tools"`
 		Swarm           SwarmConfig                `json:"swarm"`
 		Preferences     PreferencesConfig          `json:"preferences"`
+		KeyBindings     KeyBindingsConfig          `json:"keybindings"`
 		LocalControl    LocalControlConfig         `json:"localControl"`
 		MCPServers      map[string]MCPServerConfig `json:"mcpServers"`
 		MCPServersSnake map[string]MCPServerConfig `json:"mcp_servers"`
@@ -338,6 +367,7 @@ func (cfg *FileConfig) UnmarshalJSON(data []byte) error {
 	cfg.Tools = raw.Tools
 	cfg.Swarm = raw.Swarm
 	cfg.Preferences = raw.Preferences
+	cfg.KeyBindings = raw.KeyBindings
 	cfg.LocalControl = raw.LocalControl
 	if cfg.MCP.Servers == nil && (len(raw.MCPServers) > 0 || len(raw.MCPServersSnake) > 0) {
 		cfg.MCP.Servers = map[string]MCPServerConfig{}

internal/config/validate_test.go

@@ -66,10 +66,13 @@ func TestValidateFileRedactsSecretInIssue(t *testing.T) {
 }
 
 func TestValidateFileMissingModelWarns(t *testing.T) {
+	// An openai-compatible CUSTOM endpoint has no catalog default to fall back
+	// on, so a missing model is still a real issue — and the message must tell
+	// the user how to fix it.
 	path := writeValidateFixture(t, `{
 		"activeProvider": "main",
 		"providers": [
-			{"name": "main", "provider_kind": "anthropic"}
+			{"name": "main", "provider_kind": "openai-compatible", "baseURL": "https://gateway.example/v1"}
 		]
 	}`)
 
@@ -80,6 +83,30 @@ func TestValidateFileMissingModelWarns(t *testing.T) {
 	if !strings.Contains(issues[0].Message, "requires model") {
 		t.Fatalf("expected requires-model issue, got %#v", issues)
 	}
+	if !strings.Contains(issues[0].Message, "config.json") {
+		t.Fatalf("requires-model issue should carry an actionable hint, got %#v", issues)
+	}
+}
+
+func TestValidateFileDefaultsOfficialKindModels(t *testing.T) {
+	// Official-API kinds (anthropic/google) fall back to their catalog default
+	// model, so a hand-written model-less profile validates clean instead of
+	// bricking zero config / bare zero setup — the only commands that could
+	// have fixed it (the reported google case).
+	path := writeValidateFixture(t, `{
+		"activeProvider": "google",
+		"providers": [
+			{"name": "google", "provider_kind": "google", "apiKey": "AIza-x"},
+			{"name": "anthropic", "provider_kind": "anthropic", "apiKey": "sk-ant-x"}
+		]
+	}`)
+
+	_, issues := ValidateFile(path)
+	for _, issue := range issues {
+		if strings.Contains(issue.Message, "requires model") {
+			t.Fatalf("official-kind profiles must default their model, got issue %#v", issue)
+		}
+	}
 }
 
 func TestValidateFileValidConfigHasNoIssues(t *testing.T) {

internal/config/writer.go

@@ -154,6 +154,28 @@ func SetRecapsEnabled(path string, enabled bool) (FileConfig, error) {
 	return cfg, nil
 }
 
+// SetTheme persists the TUI theme preference, mirroring SetFavoriteModels
+// (read-modify-atomic-write). A blank theme clears the stored preference.
+func SetTheme(path string, theme string) (FileConfig, error) {
+	path = strings.TrimSpace(path)
+	if path == "" {
+		return FileConfig{}, fmt.Errorf("config path is required")
+	}
+	cfg := FileConfig{}
+	if data, err := os.ReadFile(path); err == nil {
+		if err := json.Unmarshal(data, &cfg); err != nil {
+			return FileConfig{}, fmt.Errorf("invalid config JSON %s: %w", path, err)
+		}
+	} else if !os.IsNotExist(err) {
+		return FileConfig{}, fmt.Errorf("read config %s: %w", path, err)
+	}
+	cfg.Preferences.Theme = strings.TrimSpace(theme)
+	if err := writeConfigFile(path, cfg); err != nil {
+		return FileConfig{}, err
+	}
+	return cfg, nil
+}
+
 func normalizeFavoriteModels(models []string) []string {
 	seen := map[string]bool{}
 	favorites := make([]string, 0, len(models))

internal/config/writer_test.go

@@ -272,6 +272,39 @@ func TestSetFavoriteModelsPersistsUserPreferences(t *testing.T) {
 	}
 }
 
+func TestSetThemePersistsUserPreference(t *testing.T) {
+	path := filepath.Join(t.TempDir(), "zero.json")
+	writeConfigFixture(t, path, FileConfig{
+		ActiveProvider: "openai",
+		Providers: []ProviderProfile{
+			{Name: "openai", ProviderKind: ProviderKindOpenAI, Model: "gpt-4.1"},
+		},
+	}, 0o600)
+
+	cfg, err := SetTheme(path, "  dracula  ")
+	if err != nil {
+		t.Fatalf("SetTheme() error = %v", err)
+	}
+	if cfg.Preferences.Theme != "dracula" {
+		t.Fatalf("Theme = %q, want dracula (trimmed)", cfg.Preferences.Theme)
+	}
+	persisted := readConfigFixture(t, path)
+	if persisted.Preferences.Theme != "dracula" {
+		t.Fatalf("persisted Theme = %q, want dracula", persisted.Preferences.Theme)
+	}
+	if persisted.ActiveProvider != "openai" || len(persisted.Providers) != 1 {
+		t.Fatalf("provider config was not preserved by SetTheme: %#v", persisted)
+	}
+
+	// A blank value clears the stored preference.
+	if cfg, err = SetTheme(path, ""); err != nil {
+		t.Fatalf("SetTheme(\"\") error = %v", err)
+	}
+	if cfg.Preferences.Theme != "" {
+		t.Fatalf("SetTheme(\"\") should clear the theme, got %q", cfg.Preferences.Theme)
+	}
+}
+
 func TestRecapsPreferenceRoundTrips(t *testing.T) {
 	// Default (unset) is ON.
 	if !(PreferencesConfig{}).RecapsEnabled() {

internal/doctor/hardening.go

@@ -131,29 +131,45 @@ func doctorSandboxPolicy(cfg config.SandboxConfig) sandbox.Policy {
 // lspServersCheck reports which language servers ZERO would use are present on
 // PATH. Missing servers are not a failure — ZERO degrades to text-only edits for
 // those languages — so the worst status is WARN, and each missing server gets an
-// actionable install command keyed by its binary name.
+// actionable install command keyed by its binary name. Only the tier-1 servers
+// (lsp.CoreServerBinaries) drive the status: the long-tail servers configured
+// for breadth are listed informationally under missingOptional, because warning
+// about a missing zls on a machine with no Zig code would be permanent noise.
 func lspServersCheck(lookup func(string) (string, error)) Check {
 	if lookup == nil {
 		lookup = exec.LookPath
 	}
+	core := map[string]bool{}
+	for _, binary := range lsp.CoreServerBinaries() {
+		core[binary] = true
+	}
 	present := map[string]any{}
 	missing := map[string]any{}
+	missingOptional := map[string]any{}
 	for _, binary := range lsp.ServerBinaries() {
 		if _, err := lookup(binary); err == nil {
 			present[binary] = "on PATH"
 			continue
 		}
-		missing[binary] = lspRemedy(binary)
+		if core[binary] {
+			missing[binary] = lspRemedy(binary)
+		} else {
+			missingOptional[binary] = lspRemedy(binary)
+		}
+	}
+	details := map[string]any{"present": present}
+	if len(missingOptional) > 0 {
+		details["missingOptional"] = missingOptional
 	}
 	if len(missing) == 0 {
-		return check("lsp.servers", "LSP servers", StatusPass, "All known language servers are available on PATH.", map[string]any{
-			"present": present,
-		})
+		message := "All core language servers are available on PATH."
+		if len(missingOptional) > 0 {
+			message = fmt.Sprintf("All core language servers are available on PATH (%d optional server(s) not installed).", len(missingOptional))
+		}
+		return check("lsp.servers", "LSP servers", StatusPass, message, details)
 	}
-	return check("lsp.servers", "LSP servers", StatusWarn, fmt.Sprintf("%d language server(s) missing from PATH; affected files degrade to text-only edits.", len(missing)), map[string]any{
-		"present": present,
-		"missing": missing,
-	})
+	details["missing"] = missing
+	return check("lsp.servers", "LSP servers", StatusWarn, fmt.Sprintf("%d language server(s) missing from PATH; affected files degrade to text-only edits.", len(missing)), details)
 }
 
 // lspRemedy returns an actionable install command for a missing language-server

internal/errhint/errhint.go

@@ -0,0 +1,167 @@
+// Package errhint classifies provider/model failures into a few user-actionable
+// categories and turns them into a one-line "next step" hint.
+//
+// Provider errors already arrive with a classified string prefix from
+// providerio.ClassifiedError ("auth error:", "rate limit error:", …); lower-level
+// failures (DNS, TLS, timeouts, context-length) arrive as raw driver or library
+// messages. Classify matches both so the interactive error row (TUI) and the
+// `zero exec` provider-error path (CLI) can append one concrete next step instead
+// of dumping an identical red blob for every failure mode.
+package errhint
+
+import "strings"
+
+// Category buckets a provider/model failure into a small set of classes that each
+// map to a distinct recovery action.
+type Category int
+
+const (
+	// Unknown means the error didn't match any known signature; callers should
+	// emit no hint rather than guess.
+	Unknown Category = iota
+	Auth
+	RateLimit
+	Connectivity
+	ModelNotFound
+	ContextOverflow
+)
+
+// providerMarkers are the prefixes the provider layer attaches to every
+// provider-originated failure: providerio.ClassifiedError emits "auth error:",
+// "rate limit error:", "provider error:", and "provider request error:", and the
+// streaming paths wrap transport/decoding failures as "provider stream error:".
+// A UI surface's error can also be a *local* failure (a tool's "permission
+// denied", a "file does not exist", a config error), so Classify only proceeds
+// past this gate for messages that are recognizably from the provider — otherwise
+// a broad substring like "does not exist" would attach a bogus /model hint to an
+// unrelated local error.
+var providerMarkers = []string{
+	"auth error:",
+	"rate limit error:",
+	"provider error:",
+	"provider request error:",
+	"provider stream error:",
+}
+
+// Classify buckets err by scanning its message for known signatures. It is a
+// deliberately conservative string heuristic — the provider layer has already
+// discarded the numeric HTTP status by the time the error reaches a UI surface,
+// so the message is all we have. It first gates on a provider-origin marker (see
+// providerMarkers) so local failures never draw a provider hint, then sub-classifies.
+// Order matters: more specific signatures are tested before broader ones (e.g.
+// "context length" as overflow before the generic "timeout" as connectivity).
+func Classify(err error) Category {
+	if err == nil {
+		return Unknown
+	}
+	m := strings.ToLower(err.Error())
+	if !containsAny(m, providerMarkers...) {
+		return Unknown
+	}
+	switch {
+	case containsAny(m, "auth error:", "unauthorized", "api key", "api_key", "invalid_api_key",
+		"authentication", "permission denied", "forbidden") || containsStatusCode(m, "401", "403"):
+		return Auth
+	case containsAny(m, "rate limit", "rate_limit", "too many requests", "quota",
+		"resource_exhausted", "overloaded") || containsStatusCode(m, "429", "529"):
+		return RateLimit
+	case containsAny(m, "context length", "context window", "maximum context", "context_length_exceeded",
+		"too many tokens", "prompt is too long", "reduce the length", "maximum context length"):
+		return ContextOverflow
+	case containsAny(m, "model not found", "model_not_found", "does not exist", "unknown model",
+		"no such model", "unsupported model", "invalid model", "model is not"):
+		return ModelNotFound
+	case containsAny(m, "dial tcp", "no such host", "connection refused", "network is unreachable",
+		"i/o timeout", "context deadline exceeded", "tls handshake", "connection reset",
+		"unexpected eof", "lookup ", "timeout"):
+		return Connectivity
+	default:
+		return Unknown
+	}
+}
+
+// TUIHint returns a one-line hint referencing interactive slash commands, or ""
+// when the category is Unknown. Meant to sit under the raw error in the live
+// error row.
+func TUIHint(err error) string {
+	switch Classify(err) {
+	case Auth:
+		return "API key rejected — run /provider to re-check your credentials"
+	case RateLimit:
+		return "Rate limited — wait a moment, or switch model with /model"
+	case Connectivity:
+		return "Can't reach the provider — run /doctor --connectivity"
+	case ModelNotFound:
+		return "Model unavailable — pick another with /model"
+	case ContextOverflow:
+		return "Context window full — run /compact to free space"
+	default:
+		return ""
+	}
+}
+
+// CLIHint returns a one-line hint referencing `zero …` subcommands, or "" when the
+// category is Unknown. Meant for the non-interactive `zero exec` error path, where
+// slash commands don't apply.
+func CLIHint(err error) string {
+	switch Classify(err) {
+	case Auth:
+		return "API key rejected — run `zero auth` or set the provider's API key, then retry"
+	case RateLimit:
+		return "Rate limited — wait a moment, or switch model with --model"
+	case Connectivity:
+		return "Can't reach the provider — run `zero doctor`"
+	case ModelNotFound:
+		return "Model unavailable — run `zero doctor` or pick another with --model"
+	case ContextOverflow:
+		return "Context window full — shorten the prompt or start a fresh session"
+	default:
+		return ""
+	}
+}
+
+func containsAny(haystack string, needles ...string) bool {
+	for _, n := range needles {
+		if strings.Contains(haystack, n) {
+			return true
+		}
+	}
+	return false
+}
+
+// containsStatusCode reports whether haystack contains any of the given HTTP
+// status codes as a standalone number — not embedded in a longer digit run like
+// "completed in 4290ms" or "request id 14015" — so an incidental number can't be
+// mis-bucketed as an auth/rate-limit failure.
+func containsStatusCode(haystack string, codes ...string) bool {
+	return HasStatusCode(haystack, codes...)
+}
+
+// HasStatusCode reports whether haystack contains any of the given HTTP status
+// codes as a standalone number — not embedded in a longer digit run like
+// "completed in 4290ms" or "request id 14015". Exported so other packages (e.g.
+// the reconnect classifier) can gate on a status code without re-implementing
+// the digit-boundary check.
+func HasStatusCode(haystack string, codes ...string) bool {
+	for _, code := range codes {
+		for from := 0; ; {
+			rel := strings.Index(haystack[from:], code)
+			if rel < 0 {
+				break
+			}
+			pos := from + rel
+			beforeOK := pos == 0 || !isASCIIDigit(haystack[pos-1])
+			end := pos + len(code)
+			afterOK := end >= len(haystack) || !isASCIIDigit(haystack[end])
+			if beforeOK && afterOK {
+				return true
+			}
+			from = pos + 1
+		}
+	}
+	return false
+}
+
+func isASCIIDigit(b byte) bool {
+	return b >= '0' && b <= '9'
+}

internal/errhint/errhint_test.go

@@ -0,0 +1,94 @@
+package errhint
+
+import (
+	"errors"
+	"strings"
+	"testing"
+)
+
+func TestClassify(t *testing.T) {
+	cases := []struct {
+		name string
+		msg  string
+		want Category
+	}{
+		{"nil-ish empty", "", Unknown},
+		{"providerio auth prefix", "auth error: your API key is missing or invalid — run `zero auth`", Auth},
+		{"raw 401", "provider request error: 401 Unauthorized", Auth},
+		{"invalid api key", "provider request error: invalid_api_key: incorrect key provided", Auth},
+		{"rate limit prefix", "rate limit error: 429 too many requests", RateLimit},
+		{"overloaded", "provider error: model is overloaded, please retry", RateLimit},
+		{"resource exhausted gemini", "provider stream error: rpc error: code = ResourceExhausted desc = quota exceeded", RateLimit},
+		{"context length openai", "provider request error: this model's maximum context length is 128000 tokens", ContextOverflow},
+		{"context_length_exceeded", "provider request error: context_length_exceeded", ContextOverflow},
+		{"model not found", "provider request error: 404 the model `gpt-9` does not exist", ModelNotFound},
+		{"unknown model", "provider request error: unknown model: sonnet-99", ModelNotFound},
+		{"dns failure", "provider stream error: Post \"https://api...\": dial tcp: lookup api.foo.com: no such host", Connectivity},
+		{"connection refused", "provider stream error: dial tcp 127.0.0.1:443: connection refused", Connectivity},
+		{"deadline", "provider stream error: context deadline exceeded (Client.Timeout exceeded)", Connectivity},
+		{"provider-marked but no sub-signature", "provider error: something totally unexpected happened", Unknown},
+
+		// An incidental number that merely contains a status-code digit run must NOT
+		// be mis-bucketed (digit-boundary check), even behind a provider marker.
+		{"incidental latency number", "provider error: request completed in 4290ms then failed", Unknown},
+		{"incidental request id", "provider error: request id 14015 failed unexpectedly", Unknown},
+		{"standalone 401 still auth", "provider request error: got 401 from upstream", Auth},
+		{"standalone 429 still rate limit", "provider error: 429 returned", RateLimit},
+
+		// Local (non-provider) failures must NOT be classified — no provider marker,
+		// so no bogus recovery hint attaches to them.
+		{"local fs permission denied", "open /etc/shadow: permission denied", Unknown},
+		{"local file missing", "stat foo.go: no such file or directory", Unknown},
+		{"local model config typo", "unknown model: sonnet-99", Unknown},
+		{"tool timeout local", "tool bash timed out after 600s", Unknown},
+	}
+	for _, tc := range cases {
+		t.Run(tc.name, func(t *testing.T) {
+			if got := Classify(errors.New(tc.msg)); got != tc.want {
+				t.Fatalf("Classify(%q) = %v, want %v", tc.msg, got, tc.want)
+			}
+		})
+	}
+}
+
+func TestClassifyNil(t *testing.T) {
+	if got := Classify(nil); got != Unknown {
+		t.Fatalf("Classify(nil) = %v, want Unknown", got)
+	}
+}
+
+// Context-overflow must win over the connectivity "timeout" catch-all: a message
+// mentioning both "context length" and a timeout is an overflow, not a network
+// problem.
+func TestContextOverflowBeatsConnectivity(t *testing.T) {
+	err := errors.New("provider request error: maximum context length exceeded; request timeout")
+	if got := Classify(err); got != ContextOverflow {
+		t.Fatalf("Classify = %v, want ContextOverflow", got)
+	}
+}
+
+func TestHintsPresentForKnownCategories(t *testing.T) {
+	known := []error{
+		errors.New("auth error: bad key"),
+		errors.New("rate limit error: 429"),
+		errors.New("provider stream error: dial tcp: no such host"),
+		errors.New("provider request error: model does not exist"),
+		errors.New("provider request error: maximum context length is 128000 tokens"),
+	}
+	for _, err := range known {
+		if h := TUIHint(err); strings.TrimSpace(h) == "" {
+			t.Fatalf("TUIHint(%q) empty, want a hint", err)
+		}
+		if h := CLIHint(err); strings.TrimSpace(h) == "" {
+			t.Fatalf("CLIHint(%q) empty, want a hint", err)
+		}
+	}
+	// Unknown yields no hint on either surface.
+	unknown := errors.New("provider error: mystery")
+	if h := TUIHint(unknown); h != "" {
+		t.Fatalf("TUIHint(unknown) = %q, want empty", h)
+	}
+	if h := CLIHint(unknown); h != "" {
+		t.Fatalf("CLIHint(unknown) = %q, want empty", h)
+	}
+}

internal/lsp/registry.go

@@ -9,26 +9,134 @@ import (
 
 // serverCommands maps a file extension to the language-server command (argv) ZERO
 // will spawn. The first element is the binary looked up on PATH; missing binaries
-// are not an error, the agent just degrades to text-only for that file.
+// are not an error, the agent just degrades to text-only for that file. Every
+// command is the language's community-standard server invoked in stdio mode, so
+// a wrong guess can never spawn something surprising — only fail the PATH check.
 var serverCommands = map[string][]string{
-	".go":  {"gopls", "serve"},
-	".ts":  {"typescript-language-server", "--stdio"},
-	".tsx": {"typescript-language-server", "--stdio"},
-	".js":  {"typescript-language-server", "--stdio"},
-	".jsx": {"typescript-language-server", "--stdio"},
-	".py":  {"pyright-langserver", "--stdio"},
-	".rs":  {"rust-analyzer"},
+	".go":     {"gopls", "serve"},
+	".ts":     {"typescript-language-server", "--stdio"},
+	".tsx":    {"typescript-language-server", "--stdio"},
+	".mts":    {"typescript-language-server", "--stdio"},
+	".cts":    {"typescript-language-server", "--stdio"},
+	".js":     {"typescript-language-server", "--stdio"},
+	".jsx":    {"typescript-language-server", "--stdio"},
+	".mjs":    {"typescript-language-server", "--stdio"},
+	".cjs":    {"typescript-language-server", "--stdio"},
+	".py":     {"pyright-langserver", "--stdio"},
+	".pyi":    {"pyright-langserver", "--stdio"},
+	".rs":     {"rust-analyzer"},
+	".c":      {"clangd"},
+	".h":      {"clangd"},
+	".cpp":    {"clangd"},
+	".cc":     {"clangd"},
+	".cxx":    {"clangd"},
+	".hpp":    {"clangd"},
+	".java":   {"jdtls"},
+	".kt":     {"kotlin-language-server"},
+	".kts":    {"kotlin-language-server"},
+	".rb":     {"ruby-lsp"},
+	".php":    {"intelephense", "--stdio"},
+	".zig":    {"zls"},
+	".lua":    {"lua-language-server"},
+	".ex":     {"elixir-ls"},
+	".exs":    {"elixir-ls"},
+	".hs":     {"haskell-language-server-wrapper", "--lsp"},
+	".swift":  {"sourcekit-lsp"},
+	".ml":     {"ocamllsp"},
+	".mli":    {"ocamllsp"},
+	".scala":  {"metals"},
+	".clj":    {"clojure-lsp"},
+	".cljs":   {"clojure-lsp"},
+	".cljc":   {"clojure-lsp"},
+	".dart":   {"dart", "language-server"},
+	".gleam":  {"gleam", "lsp"},
+	".nix":    {"nixd"},
+	".tf":     {"terraform-ls", "serve"},
+	".sh":     {"bash-language-server", "start"},
+	".bash":   {"bash-language-server", "start"},
+	".yaml":   {"yaml-language-server", "--stdio"},
+	".yml":    {"yaml-language-server", "--stdio"},
+	".json":   {"vscode-json-language-server", "--stdio"},
+	".css":    {"vscode-css-language-server", "--stdio"},
+	".scss":   {"vscode-css-language-server", "--stdio"},
+	".less":   {"vscode-css-language-server", "--stdio"},
+	".html":   {"vscode-html-language-server", "--stdio"},
+	".svelte": {"svelteserver", "--stdio"},
+	".vue":    {"vue-language-server", "--stdio"},
+	".astro":  {"astro-ls", "--stdio"},
 }
 
 // languageIDs maps a file extension to the LSP languageId used in didOpen.
 var languageIDs = map[string]string{
-	".go":  "go",
-	".ts":  "typescript",
-	".tsx": "typescriptreact",
-	".js":  "javascript",
-	".jsx": "javascriptreact",
-	".py":  "python",
-	".rs":  "rust",
+	".go":     "go",
+	".ts":     "typescript",
+	".tsx":    "typescriptreact",
+	".mts":    "typescript",
+	".cts":    "typescript",
+	".js":     "javascript",
+	".jsx":    "javascriptreact",
+	".mjs":    "javascript",
+	".cjs":    "javascript",
+	".py":     "python",
+	".pyi":    "python",
+	".rs":     "rust",
+	".c":      "c",
+	".h":      "c",
+	".cpp":    "cpp",
+	".cc":     "cpp",
+	".cxx":    "cpp",
+	".hpp":    "cpp",
+	".java":   "java",
+	".kt":     "kotlin",
+	".kts":    "kotlin",
+	".rb":     "ruby",
+	".php":    "php",
+	".zig":    "zig",
+	".lua":    "lua",
+	".ex":     "elixir",
+	".exs":    "elixir",
+	".hs":     "haskell",
+	".swift":  "swift",
+	".ml":     "ocaml",
+	".mli":    "ocaml",
+	".scala":  "scala",
+	".clj":    "clojure",
+	".cljs":   "clojure",
+	".cljc":   "clojure",
+	".dart":   "dart",
+	".gleam":  "gleam",
+	".nix":    "nix",
+	".tf":     "terraform",
+	".sh":     "shellscript",
+	".bash":   "shellscript",
+	".yaml":   "yaml",
+	".yml":    "yaml",
+	".json":   "json",
+	".css":    "css",
+	".scss":   "scss",
+	".less":   "less",
+	".html":   "html",
+	".svelte": "svelte",
+	".vue":    "vue",
+	".astro":  "astro",
+}
+
+// coreServerBinaries are the tier-1 servers for the languages agents hit most;
+// `zero doctor` treats their absence as warn-worthy. The long tail of servers
+// configured above for breadth is reported informationally only — warning about
+// a missing zls on a machine with no Zig code would be permanent noise.
+var coreServerBinaries = []string{
+	"gopls",
+	"typescript-language-server",
+	"pyright-langserver",
+	"rust-analyzer",
+}
+
+// CoreServerBinaries returns the tier-1 server binaries (sorted copy).
+func CoreServerBinaries() []string {
+	binaries := append([]string(nil), coreServerBinaries...)
+	sort.Strings(binaries)
+	return binaries
 }
 
 // ServerBinaries returns the unique set of language-server binaries ZERO may

internal/mcp/client_test.go

@@ -1,16 +1,17 @@
 package mcp
 
 import (
-	"bytes"
 	"context"
 	"encoding/json"
 	"errors"
 	"fmt"
+	"io"
 	"net/http"
 	"net/http/httptest"
 	"os"
 	"strings"
 	"sync"
+	"sync/atomic"
 	"testing"
 	"time"
 
@@ -211,6 +212,85 @@ func TestHTTPClientListsAndCallsTools(t *testing.T) {
 	}
 }
 
+func TestHTTPClientFollowsSameOriginRedirect(t *testing.T) {
+	ctx, cancel := context.WithTimeout(context.Background(), 5*time.Second)
+	defer cancel()
+
+	testServer := httptest.NewServer(http.HandlerFunc(func(response http.ResponseWriter, request *http.Request) {
+		if request.URL.Path == "/mcp" {
+			http.Redirect(response, request, "/redirected", http.StatusTemporaryRedirect)
+			return
+		}
+		if request.Method != http.MethodPost || request.URL.Path != "/redirected" {
+			t.Errorf("request = %s %s, want POST /redirected", request.Method, request.URL.Path)
+			http.Error(response, "bad request", http.StatusNotFound)
+			return
+		}
+		if got := request.Header.Get("X-Api-Key"); got != "secret" {
+			t.Errorf("X-Api-Key = %q, want secret", got)
+			http.Error(response, "missing auth", http.StatusUnauthorized)
+			return
+		}
+
+		message := readHTTPRPCMessage(t, request)
+		switch message.Method {
+		case "initialize":
+			writeHTTPRPCResponse(t, response, message.ID, map[string]any{"protocolVersion": "2024-11-05"})
+		case "notifications/initialized":
+			response.WriteHeader(http.StatusAccepted)
+		default:
+			writeHTTPRPCResponse(t, response, message.ID, map[string]any{})
+		}
+	}))
+	defer testServer.Close()
+
+	client, err := Connect(ctx, Server{
+		Name:    "docs",
+		Type:    ServerTypeHTTP,
+		URL:     testServer.URL + "/mcp",
+		Headers: map[string]string{"X-Api-Key": "secret"},
+	})
+	if err != nil {
+		t.Fatalf("Connect() error = %v", err)
+	}
+	if err := client.Close(); err != nil {
+		t.Fatalf("Close() error = %v", err)
+	}
+}
+
+func TestHTTPClientRejectsCrossOriginRedirectBeforeSendingHeaders(t *testing.T) {
+	ctx, cancel := context.WithTimeout(context.Background(), 5*time.Second)
+	defer cancel()
+
+	var targetHits int32
+	target := httptest.NewServer(http.HandlerFunc(func(response http.ResponseWriter, request *http.Request) {
+		atomic.AddInt32(&targetHits, 1)
+		if got := request.Header.Get("X-Api-Key"); got != "" {
+			t.Errorf("redirect target received X-Api-Key = %q", got)
+		}
+		response.WriteHeader(http.StatusOK)
+	}))
+	defer target.Close()
+
+	redirector := httptest.NewServer(http.HandlerFunc(func(response http.ResponseWriter, request *http.Request) {
+		http.Redirect(response, request, target.URL+"/mcp", http.StatusTemporaryRedirect)
+	}))
+	defer redirector.Close()
+
+	_, err := Connect(ctx, Server{
+		Name:    "docs",
+		Type:    ServerTypeHTTP,
+		URL:     redirector.URL + "/mcp",
+		Headers: map[string]string{"X-Api-Key": "secret"},
+	})
+	if err == nil || !strings.Contains(err.Error(), "cross-origin redirect") {
+		t.Fatalf("Connect() error = %v, want cross-origin redirect error", err)
+	}
+	if got := atomic.LoadInt32(&targetHits); got != 0 {
+		t.Fatalf("redirect target hits = %d, want 0", got)
+	}
+}
+
 func TestSSEClientListsAndCallsToolsFromRemoteStream(t *testing.T) {
 	ctx, cancel := context.WithTimeout(context.Background(), 5*time.Second)
 	defer cancel()
@@ -340,6 +420,53 @@ func TestSSEClientListsAndCallsToolsFromRemoteStream(t *testing.T) {
 	}
 }
 
+func TestSSEClientRejectsCrossOriginEndpointBeforeSendingHeaders(t *testing.T) {
+	ctx, cancel := context.WithTimeout(context.Background(), 5*time.Second)
+	defer cancel()
+
+	var targetHits int32
+	target := httptest.NewServer(http.HandlerFunc(func(response http.ResponseWriter, request *http.Request) {
+		atomic.AddInt32(&targetHits, 1)
+		if got := request.Header.Get("X-Api-Key"); got != "" {
+			t.Errorf("SSE endpoint target received X-Api-Key = %q", got)
+		}
+		response.WriteHeader(http.StatusAccepted)
+	}))
+	defer target.Close()
+
+	stream := httptest.NewServer(http.HandlerFunc(func(response http.ResponseWriter, request *http.Request) {
+		if got := request.Header.Get("X-Api-Key"); got != "secret" {
+			t.Errorf("X-Api-Key = %q, want secret on configured SSE server", got)
+			http.Error(response, "missing auth", http.StatusUnauthorized)
+			return
+		}
+		if request.Method != http.MethodGet || request.URL.Path != "/sse" {
+			t.Errorf("request = %s %s, want GET /sse", request.Method, request.URL.Path)
+			http.Error(response, "bad request", http.StatusNotFound)
+			return
+		}
+		response.Header().Set("Content-Type", "text/event-stream")
+		_, _ = fmt.Fprintf(response, "event: endpoint\ndata: %s/messages\n\n", target.URL)
+		if flusher, ok := response.(http.Flusher); ok {
+			flusher.Flush()
+		}
+	}))
+	defer stream.Close()
+
+	_, err := Connect(ctx, Server{
+		Name:    "docs",
+		Type:    ServerTypeSSE,
+		URL:     stream.URL + "/sse",
+		Headers: map[string]string{"X-Api-Key": "secret"},
+	})
+	if err == nil || !strings.Contains(err.Error(), "endpoint origin") {
+		t.Fatalf("Connect() error = %v, want cross-origin endpoint error", err)
+	}
+	if got := atomic.LoadInt32(&targetHits); got != 0 {
+		t.Fatalf("SSE endpoint target hits = %d, want 0", got)
+	}
+}
+
 func TestHTTPClientReportsNonOKStatus(t *testing.T) {
 	ctx, cancel := context.WithTimeout(context.Background(), 5*time.Second)
 	defer cancel()
@@ -393,28 +520,43 @@ func TestConnectRejectsUnsupportedTransport(t *testing.T) {
 }
 
 func TestClientRequestWaitsForMatchingResponseID(t *testing.T) {
-	var incoming bytes.Buffer
-	incomingWriter := newMessageWriter(&incoming)
-	if err := incomingWriter.write(rpcMessage{Method: "notifications/progress"}); err != nil {
-		t.Fatal(err)
-	}
-	if err := incomingWriter.write(rpcMessage{
-		ID:    99,
-		Error: &rpcError{Code: -32000, Message: "wrong response"},
-	}); err != nil {
-		t.Fatal(err)
-	}
-	if err := incomingWriter.write(rpcMessage{
-		ID:     1,
-		Result: mustRaw(map[string]any{"value": "matched"}),
-	}); err != nil {
-		t.Fatal(err)
-	}
+	// Pipes, not a pre-filled bytes.Buffer: the fake server responds only AFTER
+	// reading the outgoing request, matching production ordering. With responses
+	// pre-loaded, the client's reader goroutine could consume the matching
+	// response — and hit EOF — before request() registered its pending id,
+	// flaking as "request() error = EOF" on fast runners.
+	serverReader, clientWriter := io.Pipe() // client → server
+	clientReader, serverWriter := io.Pipe() // server → client
+	t.Cleanup(func() {
+		_ = clientWriter.Close()
+		_ = serverWriter.Close()
+	})
+
+	serverErr := make(chan error, 1)
+	go func() {
+		requests := newMessageReader(serverReader)
+		request, err := requests.read()
+		if err != nil {
+			serverErr <- err
+			return
+		}
+		responses := newMessageWriter(serverWriter)
+		for _, message := range []rpcMessage{
+			{Method: "notifications/progress"},
+			{ID: 99, Error: &rpcError{Code: -32000, Message: "wrong response"}},
+			{ID: request.ID, Result: mustRaw(map[string]any{"value": "matched"})},
+		} {
+			if err := responses.write(message); err != nil {
+				serverErr <- err
+				return
+			}
+		}
+		serverErr <- nil
+	}()
 
-	var outgoing bytes.Buffer
 	client := &Client{
-		reader: newMessageReader(&incoming),
-		writer: newMessageWriter(&outgoing),
+		reader: newMessageReader(clientReader),
+		writer: newMessageWriter(clientWriter),
 		nextID: 1,
 	}
 	var result struct {
@@ -426,6 +568,9 @@ func TestClientRequestWaitsForMatchingResponseID(t *testing.T) {
 	if result.Value != "matched" {
 		t.Fatalf("result.Value = %q, want matched response", result.Value)
 	}
+	if err := <-serverErr; err != nil {
+		t.Fatalf("fake server error: %v", err)
+	}
 }
 
 type errorCloser struct {

internal/mcp/network_client.go

@@ -733,7 +733,11 @@ func resolveSSEEndpointURL(baseValue string, endpointValue string) (string, erro
 	if err != nil {
 		return "", fmt.Errorf("parse MCP SSE endpoint URL: %w", err)
 	}
-	return baseURL.ResolveReference(endpointURL).String(), nil
+	resolvedURL := baseURL.ResolveReference(endpointURL)
+	if !sameMCPOrigin(baseURL, resolvedURL) {
+		return "", fmt.Errorf("MCP SSE endpoint origin %s differs from configured server origin %s", mcpOrigin(resolvedURL), mcpOrigin(baseURL))
+	}
+	return resolvedURL.String(), nil
 }
 
 func rpcResponseKey(id any) string {
@@ -873,10 +877,10 @@ func (source *storeTokenSource) Refresh(ctx context.Context) (string, error) {
 
 // oauthHTTPClient returns an HTTP client whose transport attaches OAuth bearer
 // tokens and refreshes them on 401 for OAuth-configured servers. Servers that do
-// not declare OAuth get the default client unchanged.
+// not declare OAuth use the default transport with the same MCP redirect guard.
 func oauthHTTPClient(server Server) (*http.Client, error) {
 	if !strings.EqualFold(strings.TrimSpace(server.Auth), ServerAuthOAuth) {
-		return http.DefaultClient, nil
+		return mcpHTTPClient(server, nil), nil
 	}
 	store, err := NewTokenStore(TokenStoreOptions{})
 	if err != nil {
@@ -888,5 +892,5 @@ func oauthHTTPClient(server Server) (*http.Client, error) {
 		httpClient: http.DefaultClient,
 		now:        time.Now,
 	}
-	return &http.Client{Transport: newOAuthRoundTripper(http.DefaultTransport, source, server.Name)}, nil
+	return mcpHTTPClient(server, newOAuthRoundTripper(http.DefaultTransport, source, server.Name)), nil
 }

internal/mcp/network_client_test.go

@@ -4,6 +4,7 @@ import (
 	"context"
 	"net/http"
 	"net/http/httptest"
+	"net/url"
 	"strings"
 	"sync/atomic"
 	"testing"
@@ -149,6 +150,141 @@ func TestOAuthRoundTripperDoesNotRetryAfterSuccessfulNon401(t *testing.T) {
 	}
 }
 
+func TestOAuthRoundTripperRejectsCrossOriginRedirectBeforeBearerLeak(t *testing.T) {
+	source := &fakeTokenSource{}
+	source.access.Store("access-secret")
+	source.refreshFunc = func() (string, error) { return "access-fresh", nil }
+
+	var targetHits int32
+	target := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+		atomic.AddInt32(&targetHits, 1)
+		if got := r.Header.Get("Authorization"); got != "" {
+			t.Errorf("redirect target received Authorization = %q", got)
+		}
+		w.WriteHeader(http.StatusOK)
+	}))
+	defer target.Close()
+
+	redirector := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+		http.Redirect(w, r, target.URL+"/mcp", http.StatusTemporaryRedirect)
+	}))
+	defer redirector.Close()
+
+	client := mcpHTTPClient(
+		Server{Name: "oauth-docs", Type: ServerTypeHTTP},
+		newOAuthRoundTripper(http.DefaultTransport, source, "oauth-docs"),
+	)
+	resp, err := client.Get(redirector.URL + "/mcp")
+	if resp != nil {
+		_ = resp.Body.Close()
+	}
+	if err == nil || !strings.Contains(err.Error(), "cross-origin redirect") {
+		t.Fatalf("Get() error = %v, want cross-origin redirect error", err)
+	}
+	if got := atomic.LoadInt32(&targetHits); got != 0 {
+		t.Fatalf("redirect target hits = %d, want 0", got)
+	}
+}
+
+func TestSameMCPOriginNormalizesDefaultPorts(t *testing.T) {
+	for _, tc := range []struct {
+		name  string
+		left  string
+		right string
+		want  bool
+	}{
+		{
+			name:  "https default port",
+			left:  "https://example.test/mcp",
+			right: "https://EXAMPLE.test:443/other",
+			want:  true,
+		},
+		{
+			name:  "http default port",
+			left:  "http://example.test/mcp",
+			right: "http://example.test:80/other",
+			want:  true,
+		},
+		{
+			name:  "different scheme",
+			left:  "https://example.test/mcp",
+			right: "http://example.test:443/other",
+			want:  false,
+		},
+		{
+			name:  "different port",
+			left:  "https://example.test/mcp",
+			right: "https://example.test:8443/other",
+			want:  false,
+		},
+	} {
+		t.Run(tc.name, func(t *testing.T) {
+			left, err := url.Parse(tc.left)
+			if err != nil {
+				t.Fatal(err)
+			}
+			right, err := url.Parse(tc.right)
+			if err != nil {
+				t.Fatal(err)
+			}
+			if got := sameMCPOrigin(left, right); got != tc.want {
+				t.Fatalf("sameMCPOrigin(%q, %q) = %v, want %v", tc.left, tc.right, got, tc.want)
+			}
+		})
+	}
+}
+
+func TestResolveSSEEndpointURLRestrictsOrigin(t *testing.T) {
+	for _, tc := range []struct {
+		name     string
+		base     string
+		endpoint string
+		want     string
+		wantErr  string
+	}{
+		{
+			name:     "relative endpoint",
+			base:     "https://example.test/sse",
+			endpoint: "/messages",
+			want:     "https://example.test/messages",
+		},
+		{
+			name:     "absolute same origin",
+			base:     "https://example.test/sse",
+			endpoint: "https://example.test/messages",
+			want:     "https://example.test/messages",
+		},
+		{
+			name:     "default port equivalent",
+			base:     "https://example.test/sse",
+			endpoint: "https://example.test:443/messages",
+			want:     "https://example.test:443/messages",
+		},
+		{
+			name:     "cross origin",
+			base:     "https://example.test/sse",
+			endpoint: "https://evil.test/messages",
+			wantErr:  "differs from configured server origin",
+		},
+	} {
+		t.Run(tc.name, func(t *testing.T) {
+			got, err := resolveSSEEndpointURL(tc.base, tc.endpoint)
+			if tc.wantErr != "" {
+				if err == nil || !strings.Contains(err.Error(), tc.wantErr) {
+					t.Fatalf("resolveSSEEndpointURL() error = %v, want %q", err, tc.wantErr)
+				}
+				return
+			}
+			if err != nil {
+				t.Fatalf("resolveSSEEndpointURL() error = %v", err)
+			}
+			if got != tc.want {
+				t.Fatalf("resolveSSEEndpointURL() = %q, want %q", got, tc.want)
+			}
+		})
+	}
+}
+
 func TestNonOAuthServerIsUnaffected(t *testing.T) {
 	// Regression: a server without auth must not get a bearer header or any
 	// OAuth round tripper, and must connect normally.

internal/mcp/network_redirect.go

@@ -0,0 +1,74 @@
+package mcp
+
+import (
+	"fmt"
+	"net"
+	"net/http"
+	"net/url"
+	"strings"
+)
+
+const maxMCPRedirects = 10
+
+func mcpHTTPClient(server Server, transport http.RoundTripper) *http.Client {
+	return &http.Client{
+		Transport:     transport,
+		CheckRedirect: checkMCPRedirect(server),
+	}
+}
+
+func checkMCPRedirect(server Server) func(*http.Request, []*http.Request) error {
+	return func(request *http.Request, via []*http.Request) error {
+		if len(via) >= maxMCPRedirects {
+			return fmt.Errorf("MCP %s server %s stopped after %d redirects", server.Type, server.Name, maxMCPRedirects)
+		}
+		if len(via) == 0 {
+			return nil
+		}
+		if !sameMCPOrigin(via[0].URL, request.URL) {
+			return fmt.Errorf("MCP %s server %s refused cross-origin redirect to %s", server.Type, server.Name, mcpOrigin(request.URL))
+		}
+		return nil
+	}
+}
+
+func sameMCPOrigin(left *url.URL, right *url.URL) bool {
+	if left == nil || right == nil {
+		return false
+	}
+	return strings.EqualFold(left.Scheme, right.Scheme) &&
+		strings.EqualFold(left.Hostname(), right.Hostname()) &&
+		effectiveMCPPort(left) == effectiveMCPPort(right)
+}
+
+func effectiveMCPPort(value *url.URL) string {
+	if value == nil {
+		return ""
+	}
+	if port := value.Port(); port != "" {
+		return port
+	}
+	switch strings.ToLower(value.Scheme) {
+	case "http":
+		return "80"
+	case "https":
+		return "443"
+	default:
+		return ""
+	}
+}
+
+func mcpOrigin(value *url.URL) string {
+	if value == nil {
+		return "<nil>"
+	}
+	host := value.Hostname()
+	port := effectiveMCPPort(value)
+	if port != "" {
+		host = net.JoinHostPort(host, port)
+	}
+	if value.Scheme == "" {
+		return host
+	}
+	return strings.ToLower(value.Scheme) + "://" + host
+}

internal/mcp/oauth_test.go

@@ -370,7 +370,14 @@ func TestFullFlowStoresTokens(t *testing.T) {
 		redirect := parsed.Query().Get("redirect_uri")
 		go func() {
 			cbURL := redirect + "?code=auth-code&state=" + url.QueryEscape(state)
-			_, _ = http.Get(cbURL)
+			for i := 0; i < 20; i++ {
+				resp, err := http.Get(cbURL)
+				if err == nil {
+					resp.Body.Close()
+					return
+				}
+				time.Sleep(50 * time.Millisecond)
+			}
 		}()
 		return nil
 	}

internal/modelregistry/catalog.go

@@ -63,6 +63,10 @@ func DefaultModelEntries() []ModelEntry {
 		googleModel("gemini-2.5-flash-lite", "Gemini 2.5 Flash-Lite", "gemini-2.5-flash-lite", ModelStatusActive, []string{"google:gemini-2.5-flash-lite", "gemini-flash-lite"}, ContextLimits{ContextWindow: 1_048_576, MaxOutputTokens: 65_536}, ModelCost{InputPerMillion: 0.1, CachedInputPerMillion: 0.01, OutputPerMillion: 0.4}, []ModelCapability{ModelCapabilityVision, ModelCapabilityJSONMode, ModelCapabilityReasoning, ModelCapabilityLongContext}, standardReasoningEfforts(), "Google low-cost Flash model for background routing and summaries."),
 	}
 	decorateModelDepth(entries)
+	// Overlay volatile facts (context limits, base pricing) from a cached
+	// models.dev snapshot when one is present and fresh — see modelsdev.go.
+	// Identity fields (ids, aliases, patterns, deprecations) stay curated.
+	entries = applyModelsDevOverrides(entries, cachedModelsDevProviders())
 	return cloneModelEntries(entries)
 }
 

internal/modelregistry/modelsdev.go

@@ -0,0 +1,269 @@
+package modelregistry
+
+import (
+	"context"
+	"encoding/json"
+	"fmt"
+	"io"
+	"net/http"
+	"os"
+	"path/filepath"
+	"strings"
+	"sync"
+	"sync/atomic"
+	"time"
+)
+
+// Live models.dev overlay for the curated catalog. The hand-maintained
+// DefaultModelEntries list is the source of truth for identity — ids, aliases,
+// match patterns, deprecations, escalation targets — but its VOLATILE facts
+// (context window, max output tokens, per-million pricing) go stale between
+// releases. When a cached snapshot of https://models.dev/api.json is present,
+// those fields are refreshed from it at registry construction; everything else
+// stays curated. The overlay never adds models (an auto-added entry would lack
+// aliases, match patterns, and provider wiring) and never touches the network
+// on the registry hot path — fetching happens only in the explicit background
+// refresh, cached to disk with a TTL.
+
+const (
+	modelsDevDefaultURL = "https://models.dev/api.json"
+	// modelsDevRefreshAfter is how old the cache may get before a background
+	// refresh re-fetches it.
+	modelsDevRefreshAfter = 24 * time.Hour
+	// modelsDevMaxAge is the oldest cache still applied as an overlay. Beyond
+	// this the curated catalog (updated with the binary) is likely fresher than
+	// the snapshot, so a stale file is ignored rather than trusted.
+	modelsDevMaxAge      = 7 * 24 * time.Hour
+	modelsDevFetchLimit  = 32 << 20 // 32MiB guard on the response body
+	modelsDevFetchWindow = 15 * time.Second
+)
+
+// modelsDevModel is the subset of a models.dev model record the overlay uses.
+type modelsDevModel struct {
+	Limit struct {
+		Context int `json:"context"`
+		Output  int `json:"output"`
+	} `json:"limit"`
+	Cost struct {
+		Input      float64 `json:"input"`
+		Output     float64 `json:"output"`
+		CacheRead  float64 `json:"cache_read"`
+		CacheWrite float64 `json:"cache_write"`
+	} `json:"cost"`
+}
+
+// modelsDevProvider matches one provider object in api.json.
+type modelsDevProvider struct {
+	Models map[string]modelsDevModel `json:"models"`
+}
+
+// parseModelsDev decodes an api.json document into provider-slug -> api-model
+// -> record. api.json is a top-level object keyed by provider id.
+func parseModelsDev(data []byte) (map[string]map[string]modelsDevModel, error) {
+	var doc map[string]modelsDevProvider
+	if err := json.Unmarshal(data, &doc); err != nil {
+		return nil, fmt.Errorf("modelregistry: parse models.dev: %w", err)
+	}
+	providers := make(map[string]map[string]modelsDevModel, len(doc))
+	for slug, provider := range doc {
+		if len(provider.Models) == 0 {
+			continue
+		}
+		providers[slug] = provider.Models
+	}
+	if len(providers) == 0 {
+		return nil, fmt.Errorf("modelregistry: models.dev document has no providers")
+	}
+	return providers, nil
+}
+
+// modelsDevSlugs maps an entry's provider kind to the models.dev provider ids
+// worth checking. Only first-party slugs are used: router entries on models.dev
+// carry generic or stale numbers.
+func modelsDevSlugs(kind ProviderKind) []string {
+	switch kind {
+	case ProviderAnthropic:
+		return []string{"anthropic"}
+	case ProviderOpenAI:
+		return []string{"openai"}
+	case ProviderGoogle:
+		return []string{"google", "google-vertex"}
+	default:
+		return nil
+	}
+}
+
+// applyModelsDevOverrides refreshes each curated entry's context limits and
+// base pricing from the snapshot, when the snapshot knows the model. Tiered
+// pricing is never touched: models.dev has no tier data, and mixing a live
+// base rate with curated tiers would misprice the tier boundaries.
+func applyModelsDevOverrides(entries []ModelEntry, providers map[string]map[string]modelsDevModel) []ModelEntry {
+	if len(providers) == 0 {
+		return entries
+	}
+	for i := range entries {
+		entry := &entries[i]
+		var record modelsDevModel
+		found := false
+		for _, slug := range modelsDevSlugs(entry.Provider) {
+			if models, ok := providers[slug]; ok {
+				if candidate, ok := models[strings.TrimSpace(entry.APIModel)]; ok {
+					record = candidate
+					found = true
+					break
+				}
+			}
+		}
+		if !found {
+			continue
+		}
+		if record.Limit.Context > 0 {
+			entry.ContextLimits.ContextWindow = record.Limit.Context
+		}
+		if record.Limit.Output > 0 {
+			entry.ContextLimits.MaxOutputTokens = record.Limit.Output
+		}
+		if len(entry.Cost.Tiers) == 0 && record.Cost.Input > 0 && record.Cost.Output > 0 {
+			entry.Cost.InputPerMillion = record.Cost.Input
+			entry.Cost.OutputPerMillion = record.Cost.Output
+			if record.Cost.CacheRead > 0 {
+				entry.Cost.CachedInputPerMillion = record.Cost.CacheRead
+			}
+			if record.Cost.CacheWrite > 0 {
+				entry.Cost.CacheWritePerMillion = record.Cost.CacheWrite
+			}
+			entry.Cost.Source = "models.dev/api.json (cached)"
+		}
+	}
+	return entries
+}
+
+// modelsDevCachePath returns the on-disk cache location. ZERO_MODELS_CACHE_PATH
+// overrides it (used by tests and unusual setups).
+func modelsDevCachePath() (string, error) {
+	if override := strings.TrimSpace(os.Getenv("ZERO_MODELS_CACHE_PATH")); override != "" {
+		return override, nil
+	}
+	base, err := os.UserCacheDir()
+	if err != nil {
+		return "", err
+	}
+	return filepath.Join(base, "zero", "modelsdev.json"), nil
+}
+
+var (
+	modelsDevOnce    sync.Once
+	modelsDevCached  map[string]map[string]modelsDevModel
+	modelsDevEnabled atomic.Bool
+)
+
+// EnableModelsDevOverlay opts the process into applying the cached models.dev
+// snapshot on top of the curated catalog. The CLI entrypoint calls it; library
+// consumers and tests that never do get the curated catalog byte-identical to
+// before, so hermetic tests can't be perturbed by a cache file on the machine.
+// ZERO_DISABLE_MODELS_FETCH disables both the overlay and the fetch.
+func EnableModelsDevOverlay() {
+	modelsDevEnabled.Store(true)
+}
+
+// cachedModelsDevProviders loads the cached snapshot once per process. Not
+// enabled, missing, stale (> modelsDevMaxAge), or malformed all yield nil and
+// the curated catalog is used untouched. Read once deliberately:
+// DefaultRegistry is called on hot paths (pickers, cost views) and must not
+// re-stat the file every time; a background refresh benefits the NEXT process.
+func cachedModelsDevProviders() map[string]map[string]modelsDevModel {
+	if !modelsDevEnabled.Load() || strings.TrimSpace(os.Getenv("ZERO_DISABLE_MODELS_FETCH")) != "" {
+		return nil
+	}
+	modelsDevOnce.Do(func() {
+		path, err := modelsDevCachePath()
+		if err != nil {
+			return
+		}
+		info, err := os.Stat(path)
+		if err != nil || time.Since(info.ModTime()) > modelsDevMaxAge {
+			return
+		}
+		data, err := os.ReadFile(path)
+		if err != nil {
+			return
+		}
+		providers, err := parseModelsDev(data)
+		if err != nil {
+			return
+		}
+		modelsDevCached = providers
+	})
+	return modelsDevCached
+}
+
+// resetModelsDevCacheForTest clears the process-level cache memoization and
+// disables the overlay.
+func resetModelsDevCacheForTest() {
+	modelsDevOnce = sync.Once{}
+	modelsDevCached = nil
+	modelsDevEnabled.Store(false)
+}
+
+// RefreshModelsDevCache fetches models.dev/api.json into the on-disk cache
+// when the cache is missing or older than modelsDevRefreshAfter. It is safe to
+// call fire-and-forget from startup (use a goroutine); it never affects the
+// current process's registry (see cachedModelsDevProviders). Disabled entirely
+// by ZERO_DISABLE_MODELS_FETCH. The URL can be overridden with ZERO_MODELS_URL.
+func RefreshModelsDevCache(ctx context.Context) error {
+	if strings.TrimSpace(os.Getenv("ZERO_DISABLE_MODELS_FETCH")) != "" {
+		return nil
+	}
+	path, err := modelsDevCachePath()
+	if err != nil {
+		return err
+	}
+	if info, err := os.Stat(path); err == nil && time.Since(info.ModTime()) < modelsDevRefreshAfter {
+		return nil
+	}
+
+	url := strings.TrimSpace(os.Getenv("ZERO_MODELS_URL"))
+	if url == "" {
+		url = modelsDevDefaultURL
+	}
+	fetchCtx, cancel := context.WithTimeout(ctx, modelsDevFetchWindow)
+	defer cancel()
+	request, err := http.NewRequestWithContext(fetchCtx, http.MethodGet, url, nil)
+	if err != nil {
+		return err
+	}
+	request.Header.Set("User-Agent", "zero-models-refresh/0.1")
+	response, err := http.DefaultClient.Do(request)
+	if err != nil {
+		return err
+	}
+	defer response.Body.Close()
+	if response.StatusCode != http.StatusOK {
+		return fmt.Errorf("modelregistry: models.dev fetch: HTTP %d", response.StatusCode)
+	}
+	data, err := io.ReadAll(io.LimitReader(response.Body, modelsDevFetchLimit))
+	if err != nil {
+		return err
+	}
+	// Validate before persisting: a bad body must never clobber a good cache.
+	if _, err := parseModelsDev(data); err != nil {
+		return err
+	}
+	if err := os.MkdirAll(filepath.Dir(path), 0o755); err != nil {
+		return err
+	}
+	temp, err := os.CreateTemp(filepath.Dir(path), "modelsdev-*.json")
+	if err != nil {
+		return err
+	}
+	if _, err := temp.Write(data); err != nil {
+		_ = temp.Close()
+		_ = os.Remove(temp.Name())
+		return err
+	}
+	if err := temp.Close(); err != nil {
+		_ = os.Remove(temp.Name())
+		return err
+	}
+	return os.Rename(temp.Name(), path)
+}

internal/modelregistry/modelsdev_test.go

@@ -0,0 +1,228 @@
+package modelregistry
+
+import (
+	"net/http"
+	"net/http/httptest"
+	"os"
+	"path/filepath"
+	"testing"
+	"time"
+)
+
+const sampleModelsDev = `{
+  "anthropic": {
+    "id": "anthropic",
+    "models": {
+      "claude-sonnet-4-5-20250929": {
+        "limit": {"context": 1000000, "output": 64000},
+        "cost": {"input": 3.5, "output": 17.5, "cache_read": 0.35, "cache_write": 4.4}
+      }
+    }
+  },
+  "google": {
+    "id": "google",
+    "models": {
+      "gemini-2.5-pro": {
+        "limit": {"context": 2097152, "output": 65536},
+        "cost": {"input": 9.99, "output": 9.99}
+      }
+    }
+  }
+}`
+
+func TestParseModelsDev(t *testing.T) {
+	providers, err := parseModelsDev([]byte(sampleModelsDev))
+	if err != nil {
+		t.Fatal(err)
+	}
+	record, ok := providers["anthropic"]["claude-sonnet-4-5-20250929"]
+	if !ok {
+		t.Fatal("expected anthropic sonnet record")
+	}
+	if record.Limit.Context != 1_000_000 || record.Cost.Input != 3.5 {
+		t.Fatalf("unexpected record: %+v", record)
+	}
+	if _, err := parseModelsDev([]byte(`{}`)); err == nil {
+		t.Fatal("empty document must be rejected")
+	}
+	if _, err := parseModelsDev([]byte(`not json`)); err == nil {
+		t.Fatal("malformed document must be rejected")
+	}
+}
+
+func TestApplyModelsDevOverrides(t *testing.T) {
+	// Point the cache at a non-existent file so DefaultModelEntries returns the
+	// pure curated catalog, then apply the sample snapshot explicitly.
+	t.Setenv("ZERO_MODELS_CACHE_PATH", filepath.Join(t.TempDir(), "absent.json"))
+	resetModelsDevCacheForTest()
+	t.Cleanup(resetModelsDevCacheForTest)
+
+	providers, err := parseModelsDev([]byte(sampleModelsDev))
+	if err != nil {
+		t.Fatal(err)
+	}
+	entries := applyModelsDevOverrides(DefaultModelEntries(), providers)
+
+	var sonnet, geminiPro, opus ModelEntry
+	for _, entry := range entries {
+		switch entry.ID {
+		case "claude-sonnet-4.5":
+			sonnet = entry
+		case "gemini-2.5-pro":
+			geminiPro = entry
+		case "claude-opus-4.1":
+			opus = entry
+		}
+	}
+
+	// Known model: limits and base pricing refreshed from the snapshot.
+	if sonnet.ContextLimits.ContextWindow != 1_000_000 || sonnet.ContextLimits.MaxOutputTokens != 64_000 {
+		t.Fatalf("sonnet limits not overridden: %+v", sonnet.ContextLimits)
+	}
+	if sonnet.Cost.InputPerMillion != 3.5 || sonnet.Cost.OutputPerMillion != 17.5 || sonnet.Cost.CacheWritePerMillion != 4.4 {
+		t.Fatalf("sonnet cost not overridden: %+v", sonnet.Cost)
+	}
+	if sonnet.Cost.Source != "models.dev/api.json (cached)" {
+		t.Fatalf("sonnet cost source not marked: %q", sonnet.Cost.Source)
+	}
+
+	// Tiered pricing is curated: limits refresh, cost must NOT (gemini-2.5-pro
+	// has curated tiers and the snapshot's flat 9.99 would misprice them).
+	if geminiPro.ContextLimits.ContextWindow != 2_097_152 {
+		t.Fatalf("gemini limits not overridden: %+v", geminiPro.ContextLimits)
+	}
+	if geminiPro.Cost.InputPerMillion == 9.99 || len(geminiPro.Cost.Tiers) == 0 {
+		t.Fatalf("tiered cost must stay curated: %+v", geminiPro.Cost)
+	}
+
+	// Model absent from the snapshot: untouched.
+	if opus.ContextLimits.ContextWindow != 200_000 || opus.Cost.InputPerMillion != 15 {
+		t.Fatalf("opus must be untouched: %+v %+v", opus.ContextLimits, opus.Cost)
+	}
+}
+
+func TestRefreshModelsDevCacheFetchesAndCaches(t *testing.T) {
+	hits := 0
+	server := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+		hits++
+		_, _ = w.Write([]byte(sampleModelsDev))
+	}))
+	defer server.Close()
+
+	cachePath := filepath.Join(t.TempDir(), "modelsdev.json")
+	t.Setenv("ZERO_MODELS_CACHE_PATH", cachePath)
+	t.Setenv("ZERO_MODELS_URL", server.URL)
+	t.Setenv("ZERO_DISABLE_MODELS_FETCH", "")
+
+	if err := RefreshModelsDevCache(t.Context()); err != nil {
+		t.Fatal(err)
+	}
+	if hits != 1 {
+		t.Fatalf("expected 1 fetch, got %d", hits)
+	}
+	if _, err := os.Stat(cachePath); err != nil {
+		t.Fatalf("cache file missing: %v", err)
+	}
+	// Fresh cache: second call must not re-fetch.
+	if err := RefreshModelsDevCache(t.Context()); err != nil {
+		t.Fatal(err)
+	}
+	if hits != 1 {
+		t.Fatalf("fresh cache must skip the fetch, got %d hits", hits)
+	}
+}
+
+func TestRefreshModelsDevCacheRejectsBadBodyWithoutClobbering(t *testing.T) {
+	server := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+		_, _ = w.Write([]byte("not json"))
+	}))
+	defer server.Close()
+
+	cachePath := filepath.Join(t.TempDir(), "modelsdev.json")
+	if err := os.WriteFile(cachePath, []byte(sampleModelsDev), 0o644); err != nil {
+		t.Fatal(err)
+	}
+	stale := time.Now().Add(-48 * time.Hour)
+	if err := os.Chtimes(cachePath, stale, stale); err != nil {
+		t.Fatal(err)
+	}
+	t.Setenv("ZERO_MODELS_CACHE_PATH", cachePath)
+	t.Setenv("ZERO_MODELS_URL", server.URL)
+
+	if err := RefreshModelsDevCache(t.Context()); err == nil {
+		t.Fatal("bad body must return an error")
+	}
+	content, err := os.ReadFile(cachePath)
+	if err != nil || string(content) != sampleModelsDev {
+		t.Fatal("bad fetch must not clobber the existing cache")
+	}
+}
+
+func TestRefreshModelsDevCacheDisabledByEnv(t *testing.T) {
+	server := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+		t.Fatal("fetch must not happen when disabled")
+	}))
+	defer server.Close()
+	t.Setenv("ZERO_MODELS_CACHE_PATH", filepath.Join(t.TempDir(), "modelsdev.json"))
+	t.Setenv("ZERO_MODELS_URL", server.URL)
+	t.Setenv("ZERO_DISABLE_MODELS_FETCH", "1")
+	if err := RefreshModelsDevCache(t.Context()); err != nil {
+		t.Fatal(err)
+	}
+}
+
+func TestCachedModelsDevProvidersIgnoresStaleCache(t *testing.T) {
+	cachePath := filepath.Join(t.TempDir(), "modelsdev.json")
+	if err := os.WriteFile(cachePath, []byte(sampleModelsDev), 0o644); err != nil {
+		t.Fatal(err)
+	}
+	stale := time.Now().Add(-modelsDevMaxAge - time.Hour)
+	if err := os.Chtimes(cachePath, stale, stale); err != nil {
+		t.Fatal(err)
+	}
+	t.Setenv("ZERO_MODELS_CACHE_PATH", cachePath)
+	resetModelsDevCacheForTest()
+	t.Cleanup(resetModelsDevCacheForTest)
+	EnableModelsDevOverlay()
+
+	if providers := cachedModelsDevProviders(); providers != nil {
+		t.Fatal("stale cache must be ignored")
+	}
+}
+
+func TestCachedModelsDevProvidersRequiresOptIn(t *testing.T) {
+	cachePath := filepath.Join(t.TempDir(), "modelsdev.json")
+	if err := os.WriteFile(cachePath, []byte(sampleModelsDev), 0o644); err != nil {
+		t.Fatal(err)
+	}
+	t.Setenv("ZERO_MODELS_CACHE_PATH", cachePath)
+	resetModelsDevCacheForTest()
+	t.Cleanup(resetModelsDevCacheForTest)
+
+	// Without EnableModelsDevOverlay a fresh, valid cache must still be ignored:
+	// library consumers and hermetic tests get the pure curated catalog.
+	if providers := cachedModelsDevProviders(); providers != nil {
+		t.Fatal("overlay must be opt-in")
+	}
+}
+
+func TestDefaultModelEntriesAppliesFreshCache(t *testing.T) {
+	cachePath := filepath.Join(t.TempDir(), "modelsdev.json")
+	if err := os.WriteFile(cachePath, []byte(sampleModelsDev), 0o644); err != nil {
+		t.Fatal(err)
+	}
+	t.Setenv("ZERO_MODELS_CACHE_PATH", cachePath)
+	resetModelsDevCacheForTest()
+	t.Cleanup(resetModelsDevCacheForTest)
+	EnableModelsDevOverlay()
+
+	for _, entry := range DefaultModelEntries() {
+		if entry.ID == "claude-sonnet-4.5" {
+			if entry.ContextLimits.ContextWindow != 1_000_000 {
+				t.Fatalf("fresh cache must overlay the registry: %+v", entry.ContextLimits)
+			}
+			return
+		}
+	}
+	t.Fatal("claude-sonnet-4.5 not found")
+}

internal/oauth/encrypt.go

@@ -114,7 +114,12 @@ func createSecretFile(path string) ([]byte, error) {
 			}
 			return writeNewSecretFile(path)
 		}
-		if !errors.Is(err, os.ErrExist) {
+		// On Windows a concurrent holder's os.Remove leaves the lock file in a
+		// "delete pending" state, so an O_EXCL create races it with
+		// ERROR_ACCESS_DENIED (os.ErrPermission) rather than ErrExist. Treat that
+		// as contention and retry too — mirroring acquireFileLock in lock.go —
+		// otherwise concurrent secret creation spuriously fails on Windows.
+		if !errors.Is(err, os.ErrExist) && !errors.Is(err, os.ErrPermission) {
 			return nil, fmt.Errorf("oauth: create token secret lock: %w", err)
 		}
 		if data, rerr := readSecretFileRetry(path); rerr == nil {