Triage announced specification completion before its finally block marked the
plan work item terminal, so the Plan Review seeder saw its own still-running
predecessor as an "active continuation", bailed, and the discarded result
silently stranded the card until FN-8592 self-healing re-seeded it ~10 minutes
later (529 occurrences in 18 days).
- seedStrandedPlanReviewContinuation gains retirePredecessorId: idle check
excludes the named predecessor, then retires it and installs the successor in
ONE transaction under the task lock; a bailed seed mutates nothing.
- triage threads planningWorkItemId through PlanningHandoffReport; the runtime
reaction passes it as retirePredecessorId.
- reactToSpecificationComplete consumes the seed result: typed quiet parks
(incl. new "no-pre-release-plan-review"), bounded retries with a fresh
task/IR snapshot per attempt (mid-retry pause/needs-replan honored), loud
warning naming self-healing on exhaustion.
- Tests: PG both-orderings/no-mutation-on-bail/cross-task cases, direct engine
seeder handoff cases, reaction retry/park/pause/replan cases.
- docs/solutions: new planning-handoff-race writeup; graph-entry-contract doc
reclassifies the FN-8592 sweep as backstop-only.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Replace the fixed ~2.1s real time-wait in the agent-activity SSE durable
integration test with a bounded FUSION_AGENT_ACTIVITY_POLL_MS env test-seam
(default 2000ms unchanged in production) and poll for the delivered frame.
Removes an FN-5048 real time-wait anti-pattern; production behavior unchanged.
Refresh the Grok runtime README with the current flagship model invocation.
- Update the optional Grok ACP command to use grok-4.6.
- Document why the illustrative model identifier tracks the provider default.
Files changed:
plugins/fusion-plugin-grok-runtime/README.md | 3 ++-
1 file changed, 2 insertions(+), 1 deletion(-)
Fusion-Task-Id: FN-9015
Fusion-Task-Lineage: 6dd04810-fe06-4462-82b9-5abb38e229c0
Co-authored-by: Fusion (runfusion.ai) <noreply@runfusion.ai>
Make the Quick Add merger model row consistent with the other role selectors.
- Label the top-level merger selector as Merger
- Align the merger icon spacing with other model menu rows
- Cover role labels, spacing, and override navigation
Files changed:
.changeset/fn-9009-quick-add-merger-label.md | 7 ++++
.../dashboard/app/components/QuickEntryBox.tsx | 25 ++++++++++---
.../components/__tests__/QuickEntryBox.test.tsx | 41 +++++++++++++++++++---
3 files changed, 65 insertions(+), 8 deletions(-)
Fusion-Task-Id: FN-9009
Fusion-Task-Lineage: bfc915d6-7a8d-4f08-86e8-c270b5783458
Co-authored-by: Fusion (runfusion.ai) <noreply@runfusion.ai>
Keep the Quick Add model filter focused and functional within its portaled menu.
- Exempt form controls and the nested model dropdown from the composer focus guard.
- Add desktop and mobile regression coverage for filtering model choices.
- Add a patch changeset for the Quick Add filter fix.
Files changed:
.changeset/fix-quick-add-model-filter.md | 7 ++
.../dashboard/app/components/QuickEntryBox.tsx | 14 ++-
.../__tests__/QuickEntryBox-model-filter.test.tsx | 113 +++++++++++++++++++++
3 files changed, 133 insertions(+), 1 deletion(-)
Fusion-Task-Id: FN-9010
Fusion-Task-Lineage: da86e2d8-8722-4598-b25e-786b5850971c
Co-authored-by: Fusion (runfusion.ai) <noreply@runfusion.ai>
The check-workspace-package-graph validator globs the filesystem for package
manifests but only excluded node_modules/dist. On any machine that had run a
desktop packaging build (pnpm deploy), the gitignored electron-builder staging
dir packages/desktop/deploy/ carries a copy of desktop's package.json and
tripped the unglobbed-package violation, breaking pretest/gate locally while CI
(clean checkout, no deploy dir) stayed green. Exclude the staging path alongside
node_modules/dist since a gitignored dir is absent from the isolated worktree
this validator guards.
## Summary
Starts **U5** of the package code-organization program after wave 18
(executor peels) landed.
Peels pure free-function clusters out of `self-healing.ts` into
`packages/engine/src/self-healing/` without behavior changes. Public
imports from `./self-healing.js` remain stable via re-exports.
### Peels
| Symbol | New home |
|--------|----------|
| `autoRecoverWorktreeSessionStartFailure` |
`self-healing/auto-recover-worktree-session.ts` |
| `archiveAsGhostBug` | `self-healing/archive-ghost-bug.ts` |
| `hasStepProgress` / work-complete helpers |
`self-healing/step-progress.ts` |
### Line count
- `self-healing.ts`: ~15456 → ~15231 (baseline ratcheted to post-peel
live; main had already drifted past the prior grandfathered ceiling via
organic growth)
- New modules each well under 2,000 lines
## Test plan
- [x] `pnpm --filter @fusion/engine exec tsc --noEmit`
- [x] `self-healing-trait-rekey.test.ts` (autoRecover requeue)
- [x] `self-healing-paused-abort-recovery.test.ts`
- [x] `self-healing-model-unavailable-recovery.test.ts`
- [ ] CI gate
## Follow-ups
U5 Slice B: domain method clusters (startup, in-review, merge-status,
workspace, surfacing) into additional `self-healing/*.ts` modules.
<!-- This is an auto-generated comment: release notes by coderabbit.ai
-->
## Summary by CodeRabbit
* **Bug Fixes**
* Improved automatic recovery when worktree sessions fail to start,
including stale or incomplete session data.
* Tasks can be safely requeued while preserving progress, or escalated
after retry limits are reached.
* Improved handling of completed work and failures where task completion
was not recorded.
* Preserved valid task branches during recovery and provided more
reliable fallback requeue behavior.
* Ghost bugs are automatically archived with recovery details and
activity history.
<!-- end of auto-generated comment: release notes by coderabbit.ai -->
## Summary
Supersedes #3421 — fork head not writable for main-merge. Same change
rebased onto current main so lifecycle-column lint stays green.
## Test plan
- [x] merges cleanly onto main
- [ ] CI green
Co-authored-by: BESA-Franz <49682134+BESA-Franz@users.noreply.github.com>
## Summary
- Classify workflow work-item `workflowRole` comparisons as role
vocabulary in the lifecycle-column census.
- Add a regression test so triage role comparisons cannot raise a
phantom lifecycle-column guard.
## Test Plan
- `node --test scripts/__tests__/lifecycle-census*.test.mjs`
- `corepack pnpm check:lifecycle-columns`
- `corepack pnpm lint`
- `corepack pnpm check:changesets --strict`
<!-- This is an auto-generated comment: release notes by coderabbit.ai
-->
## Summary by CodeRabbit
* **Bug Fixes**
* Improved classification of workflow role comparisons, including
`workflowRole === "triage"`, so they are recognized separately from
lifecycle-column comparisons.
* Ensured workflow role values are correctly identified as role
vocabulary rather than lifecycle-column values.
* **Tests**
* Added automated coverage to verify accurate workflow role and column
identification across comparison patterns.
<!-- end of auto-generated comment: release notes by coderabbit.ai -->
## Problem
`app/components/__tests__/navigation-history.test.tsx`,
`TaskDetail.swipe-back.test.tsx`, and
`TaskDetail.mobile-transition.board-panel.test.tsx` fail to load.
FN-8796 (`132026545`, "stabilize task-detail lifecycle snapshots") added
`mergeTaskSnapshot` to `hooks/useTasks.ts`, which
`App.tsx`/`MainContent.tsx` now import. The curated test suites that
`vi.mock("../../hooks/useTasks")` still only expose `useTasks`, so the
fixture throws at import time:
```
No "mergeTaskSnapshot" export is defined on the useTasks mock
```
Confirmed on `origin/main`: `mergeTaskSnapshot` is present in
`useTasks.ts` and imported by `App.tsx`, while these tests' `vi.mock`
shims don't surface it.
## Fix
Update the mocks to surface every export the rendered components import,
using a partial `importOriginal` mock so `mergeTaskSnapshot` keeps its
real implementation (a `vi.fn()` returning `undefined` would make
`App.tsx:2179` `liveTask.id` throw):
```ts
vi.mock("../../hooks/useTasks", async (importOriginal) => {
const actual = await importOriginal();
return { ...actual, useTasks: (_options?: any) => mockUseTasks() };
});
```
## Verification
- `FUSION_DASHBOARD_DEEP=1 vitest run navigation-history swipe-back
board-panel` → **14 files, 116 tests passed**
- Pre-existing on `origin/main`, fixes a regression introduced by
FN-8796.
## Scope
Tests + a one-line docs note only. No production behavior change.
<!-- This is an auto-generated comment: release notes by coderabbit.ai
-->
## Summary by CodeRabbit
- **Documentation**
- Documented improved worktree recovery behavior, including safer
handling of incomplete, inactive, and unregistered directories.
- Added guidance for recovery across different storage locations and
bounded cleanup of recognized entries.
- Clarified protections that prevent internal recovery containers from
being treated as worktrees.
- **Tests**
- Improved dashboard test coverage and reliability for task detail
rendering, mobile transitions, swipe-back navigation, and navigation
history.
<!-- end of auto-generated comment: release notes by coderabbit.ai -->
---------
Co-authored-by: Fusion <noreply@runfusion.ai>
Co-authored-by: gsxdsm <gsxdsm@users.noreply.github.com>
## Summary
- make the already-merged recovery fixture explicit that no maintenance
sweep owns terminal-failure recovery
- preserve the fail-open `task-wedged` notification assertion when no
landed commit is found
## Test plan
- `corepack pnpm --filter @fusion/engine exec vitest run
--silent=passed-only --reporter=dot src/__tests__/self-healing.test.ts`
(460 passed)
- `corepack pnpm --filter @fusion/engine typecheck`
- `node scripts/run-static-gate-checks.mjs` (12 passed)
- `corepack pnpm --filter @fusion/engine test:core` (422 passed)
- `corepack pnpm --filter @fusion/core test:unit-gate` (184 passed)
- `corepack pnpm --filter @runfusion/fusion test:ci-shape` (71 passed)
- `corepack pnpm test:gate` reaches the PostgreSQL harness but cannot
authenticate locally (`empty password returned by client`) before
product assertions
<!-- This is an auto-generated comment: release notes by coderabbit.ai
-->
## Summary by CodeRabbit
* **Tests**
* Updated failure-notification test conditions to cover scenarios
without automatic maintenance processing.
* Added coverage confirming terminal-failure notifications fail open
when no maintenance sweep owns the task.
* Bypassed production debounce timing to make the test behavior
deterministic.
<!-- end of auto-generated comment: release notes by coderabbit.ai -->
## Summary
Supersedes #3416 — the fork head is not writable from maintainers, so
this branch carries the same fix merged onto current `main`.
## Changes
- Bind the self-owned worktree liveness probe in task-executor pure
facades (from #3416).
- Merge current main so the lifecycle-column census baseline stays
green.
## Test plan
- [x] `pnpm check:lifecycle-columns` clean after merge
- [ ] CI green
Closes context from #3416.
<!-- This is an auto-generated comment: release notes by coderabbit.ai
-->
## Summary by CodeRabbit
* **Bug Fixes**
* Fixed cleanup for self-owned worktrees so it no longer crashes before
checking whether an active session is present.
* Improved safety during worktree reconciliation by correctly verifying
that the worktree is still active before removal.
* **Documentation**
* Added a changeset documenting the worktree cleanup fix.
<!-- end of auto-generated comment: release notes by coderabbit.ai -->
Co-authored-by: BESA-Franz <49682134+BESA-Franz@users.noreply.github.com>
## Summary
- Complete the isolated `@fusion/core` mock used by the
experiment-finalize extension suite
- Classify three intentional physical/synthetic lifecycle literals
introduced on current main
- Re-record the strict lifecycle census baseline with zero unexamined
guards
## Test plan
- `pnpm --filter @runfusion/fusion exec vitest run
src/__tests__/extension-experiment-finalize.test.ts --silent=passed-only
--reporter=dot`
- `pnpm --filter @fusion/core exec vitest run
src/__tests__/task-intake-owner-resolver.test.ts --silent=passed-only
--reporter=dot`
- `pnpm --filter @fusion/engine exec vitest run --project engine-default
src/__tests__/mission-feature-sync-lanes.test.ts --silent=passed-only
--reporter=dot`
- `pnpm check:lifecycle-columns`
- `node scripts/check-mock-completeness.mjs`
<!-- This is an auto-generated comment: release notes by coderabbit.ai
-->
## Summary by CodeRabbit
- **Bug Fixes**
- Improved mission reconciliation previews for task links, specification
alignment, and lifecycle updates.
- Prevented stale or superseded validation runs from overwriting current
feature status or ownership.
- Improved blocked-feature diagnostics and archived-task handling across
workflow configurations.
- **Documentation**
- Clarified validation, assignment checks, and mission synchronization
behavior.
- **Tests**
- Expanded coverage for reconciliation previews and validator ownership
scenarios.
- **Chores**
- Updated lifecycle baseline data for known archived-task cases.
<!-- end of auto-generated comment: release notes by coderabbit.ai -->
The planning-continuation drain skipped every due `kind: "task"` row whose
`waitReason` was not "planning", on the premise that such rows "belong to a
different drain". No such drain exists: `listDueWorkflowWorkItems` has exactly
two callers, this pass and the self-healing reclaim sweep, and the sweep
deliberately leaves `runnable`/`retrying` rows alone as "the dispatcher's own
queue". A capacity-parked continuation was therefore owned by nobody — skipped
here every poll with no state change and no audit row, and passed over there by
design.
Observed on the Fusion board: eight cards sat runnable for up to 8h with the
engine unpaused, 0 tasks in progress, and 4 of 10 worktrees used. Three carried
`waitReason: "capacity"` from the capacity-suspend path; five carried NULL. The
09:04 reclaim sweep had just moved them held -> runnable, handing them to this
drain and simultaneously putting them out of its own reach, so the auto-resume
fix tightened the strand it repaired.
Dispatch stays admission-gated by `admitPlanningContinuation`, so a
capacity-parked card resumes only when a slot is genuinely free.
Also repairs two stale path allowlists in planning-claim-single-writer.ts: the
mission stores and replan-target.ts moved into subdirectories, leaving that
ratchet red on main and accusing the two modules it exists to exclude.
Verified: the patched classifier returns `actionable` for all 8 live stranded
rows; gate + lint green.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
