Make diagnostics evidence bounded, share-safe, and resilient to concurrent
writers and corrupt segments. Improve Doctor repair safeguards, Electron
runtime recovery, Settings visibility, accessibility, and issue reporting.
Do not pass windowsHide to detached GUI open targets, which forwards SW_HIDE to Explorer and IDE windows on Windows. Preserve detached process lifetime and ignored stdio.
Tests: open-target 33/33; server 1624/1624; coverage 5/5 with changed lines 6/6; real macOS Finder reveal passed for a temporary CJK and ampersand path.
Remaining validation: packaged Windows 11 Explorer and IDE window smoke.
Confidence: high
Scope-risk: narrow
Stamp desktop-owned transcripts with the claude-desktop entrypoint while preserving the SDK runtime entrypoint used for provider authentication.
Tests: focused regression 41/41; chat contract 240/240; server 1622/1622; coverage 5/5 with changed lines 4/4.
Live-path proof: hermetic loopback transcript was visible in Claude Code 2.1.206 /resume; no real user credentials or config were used.
Constraint: check:impact remains policy-blocked until a PR has the allow-cli-core-change label.
Confidence: high
Scope-risk: narrow
Load the authenticated Codex model catalog with a bundled fallback and carry model-native reasoning effort through the OpenAI OAuth transport. Keep desktop effort choices aligned with each model's advertised capabilities.
Constraint: Ultra remains a composite Codex delegation mode and is not exposed as a wire-level effort.
Confidence: high
Scope-risk: moderate
Tested: bun run check:server; bun run check:desktop; bun run check:persistence-upgrade; focused OAuth, WebSocket, and browser smoke checks
Not-tested: bun run verify
Keep AutoDream visible without marking the foreground session busy, add a task-scoped stop control, and preserve the authoritative stopped bookend emitted by the CLI runtime.
Tested: bun run check:desktop
Tested: bun run check:server
Tested: bun run check:chat-contract
Tested: bun run check:coverage
Tested: real MiniMax-M3 browser smoke for background stop and AutoDream busy-state isolation
Confidence: high
Scope-risk: moderate
Model Kimi K2.7 as required-thinking so a global disabled preference cannot generate an unsupported request. Keep K2.6 and other providers free to disable thinking.
Tested: request-shape loopback through the query pipeline
Tested: bun run check:provider-contract
Tested: bun run check:server (1608 tests)
Tested: bun run check:coverage
Not-tested: live Kimi completion blocked by insufficient provider balance
Constraint: CLI core changes require allow-cli-core-change approval for PR
Confidence: high
Scope-risk: moderate
Avoid appending the final result when it repeats the assistant text while preserving distinct result summaries.
Tested: bun test src/server/__tests__/cron-scheduler.test.ts
Tested: bun run check:server
Tested: live MiniMax stream-json duplicate comparison
Tested: changed-line coverage 100% (9/9)
Confidence: high
Scope-risk: narrow
Exercise controlled changes, repeated trigger closure, tab switches, and both bypass dialog close paths so the changed-line coverage gate proves the full active-turn guard.\n\nTested: desktop PermissionModeSelector 14/14; adapters and H5 auth 51/51\nConfidence: high\nScope-risk: narrow
Run required server and contract suites in credential-free sandboxes, fail closed on incomplete coverage or test output, and preserve the desktop active-turn permission guard across stale tab interactions.\n\nTested: bun run check:policy (115 pass); bun run check:server (1605 pass before final runner evidence check); bun run check:desktop; bun run check:provider-contract; bun run check:chat-contract\nConfidence: high\nScope-risk: broad
Route required checks by changed surface, add offline provider and chat contracts, and keep fork PRs independent of live credentials. Layer agent guidance by subtree and enforce a compact instruction budget.
Tested: bun run check:policy
Confidence: high
Scope-risk: broad
Retry idle provider streams only before tool side effects, and reconcile desktop turn state after WebSocket gaps so completed or still-running work cannot leave a stale spinner.
Tested: 108 runtime/server focused tests, 128 desktop focused tests
Tested: disconnected-turn WebSocket integration test
Tested: bun run check:server (1489 pass)
Tested: desktop typecheck and production build
Not-tested: full desktop gate has one unrelated TraceSession polling assertion failure
Scope-risk: broad
Coalesce concurrent session-list scans and keep desktop pollers from stacking requests while the local server is slow.
Tested: bun run check:server (1489 pass)
Tested: desktop focused tests, typecheck, and production build
Scope-risk: moderate
Track tool and Computer Use approvals by request ID so parallel prompts remain independently actionable.
Reconcile resolution state across clients and reconnects, including stopped and completed turns.
Refs: #975, #980
Tested: bun run check:desktop (1778 tests)
Tested: bun run check:server (1482 tests)
Tested: agent-browser concurrent permission smoke
Not-tested: Windows packaged app manual smoke
Confidence: high
Scope-risk: moderate
Resolves conflicts with the independently developed skill-market/SkillCenter
feature that landed on main after this branch diverged. Keeps the market
rewrite as the sole implementation:
- Removes the parallel skill-market stack (server api/skill-market.ts,
services/skillMarket/, SkillCenter.tsx, skillMarketStore.ts, related tests
and i18n keys) in favor of the market implementation from this branch.
- Restores the Settings > Skills tab (and its redirect-free behavior) that
the removed implementation had folded into a top-level Skill Center tab,
keeping local skill browsing on SkillDetailView alongside the online market.
- Adapts tests and callers (TabBar, ContentRouter, PluginDetail,
LocalSlashCommandPanel, persistence migrations) from skill-center/
SKILL_CENTER_TAB_ID naming to market/MARKET_TAB_ID.
Classify stream watchdog timeouts by stream phase and include non-sensitive diagnostics for stalled provider streams.
Map known watchdog failures to stable desktop WebSocket error codes, and only retry watchdog aborts before content or tool activity starts.
Tested: bun test src/services/api/streamWatchdog.test.ts
Tested: bun test src/server/__tests__/ws-memory-events.test.ts --test-name-pattern "maps watchdog API errors to stable desktop error codes"
Tested: cd desktop && bun test src/stores/chatStore.test.ts --test-name-pattern "persists partial assistant text before an error"
Tested: bun run check:server
Tested: cd desktop && bun run check:desktop
Tested: 5x direct CLI replay, 5x desktop-chain replay, 5x desktop-chain no-entrypoint replay with session 446f8776-538d-46e6-96d6-67080b455b07
Add server-side detail caching and in-flight request coalescing for marketplace details, reuse a long-lived service across API requests, and parallelize ClawHub detail preview fetches.
Add desktop detail cache/prefetch behavior and render marketplace Markdown/code previews through the existing MarkdownRenderer and CodeViewer components.
Stabilize the H5 diagnostics test on machines where the first non-internal IPv4 address is not a plain LAN address.
Tested: bun test src/server/__tests__/skill-market.test.ts src/server/__tests__/skill-market-install.test.ts
Tested: cd desktop && bun run test -- SkillCenter.test.tsx skillMarketStore.test.ts --run
Tested: bun test src/server/__tests__/h5-access-service.test.ts -t 'getDiagnostics reports stale'
Tested: git diff --check
Tested: bun run check:desktop
Tested: bun run check:server
Not-tested: bun run verify and coverage gates; this is a local handoff, not PR-ready validation.
Confidence: high
Scope-risk: moderate
Tested: bun test src/server/__tests__/skill-market.test.ts
Tested: cd desktop && bun run test -- SkillCenter.test.tsx skillMarketStore.test.ts generalSettings.test.ts ComputerUseSettings.test.ts --run
Tested: cd desktop && bun run test -- persistenceMigrations.test.ts --run
Tested: bun run check:desktop
Tested: real browser smoke for Skill Vetter across white, light, and dark themes
Not-tested: live install of an uninstalled marketplace skill to disk
Confidence: medium
Scope-risk: moderate
Route marketplace install requests through server-side detail lookup, eligibility checks, trusted package URL validation, bounded zip downloads, and the local user skill installer.
Reject client-supplied targets and arbitrary package URLs so the desktop cannot direct writes outside the computed user skill directory.
Tested: bun test src/server/__tests__/skill-market.test.ts src/server/__tests__/skill-market-install.test.ts
Tested: cd desktop && bun run test -- SkillCenter.test.tsx skillMarketStore.test.ts
Tested: cd desktop && bun run lint
Tested: cd desktop && bun run build
Not-tested: bun run check:server is blocked by missing @whiskeysockets/baileys in adapters/whatsapp/session.ts in this checkout.
Scope-risk: moderate
Confidence: high
Tested: bun test src/server/__tests__/skill-market.test.ts
Tested: bun test src/server/__tests__/skill-market-install.test.ts
Tested: git diff --check
Confidence: high
Scope-risk: narrow
