ce-code-review (4 personas) on Phase B. No P0; the review conjunction was
confirmed safe (no false-done — empty map and per-repo throws both route to
UNAVAILABLE, which blocks). Applied:
P1: the fn_task_done scope-leak guard now fails CLOSED in workspace mode — a
per-repo capture throw blocks completion ("refusing as a precaution") instead of
the outer .catch returning {blocked:false} and letting an incomplete check pass.
A scoped task that acquired ZERO sub-repo worktrees is now blocked rather than
silently passing scope enforcement.
P2: reviewWorkspacePerRepo breaks on the first non-APPROVE repo so a later repo's
throw can't discard an already-determined REVISE (callers were seeing UNAVAILABLE
instead). captureWorkspaceModifiedFiles isolates each per-repo capture in
try/catch so one repo's throw can't skip the modifiedFiles write. The .changeset
always-allowed carve-out is honored in workspace mode: the scope-leak branch now
filters repo-LOCAL paths via the (previously dead) workspace-paths.ts
deriveRepoScopeSubset helper through the same filter as the singular path, so a
sub-repo .changeset/* no longer falsely blocks fn_task_done. All four per-repo
loops iterate sorted keys for deterministic offending-repo reporting; the dead
repoRel callback param and the duplicate path-normalizer are removed.
Verified safe (no change): the reviewer semaphore releases on throw (try/finally),
and per-repo reviewers inherit the task abort via session disposal.
Deferred to Phase C: extracting a workspace-executor.ts module (before the merge
loop lands). Gate green: typecheck, lint, build, test:gate (649+58).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
In workspace mode both review entry points and the completion guards now iterate
every acquired sub-repo. A shared reviewWorkspacePerRepo loops task.workspaceWorktrees
and invokes the existing single-cwd reviewStep once per repo (cwd = the sub-repo —
the reviewer agent runs its own git diff there), aggregating repo-tagged verdicts
as a conjunction: the task is reviewed only if every repo APPROVEs; the first
non-APPROVE repo's verdict becomes the aggregate. Both call sites loop — the
in-session fn_review_step tool AND the step-inversion seam (createReviewStepTool
and the stepReview workflow seam) — so no review surface silently scopes to the
non-git root (FN-5893). reviewStep itself stays single-cwd; the callers loop.
fn_task_done completion verification iterates per repo: verifyWorktreeInvariants
(from U1) already covers all worktrees, and evaluateTaskDoneScopeLeak now loops
each sub-repo (cwd + repo.baseCommitSha, repo-prefixed touched files vs the
repo-prefixed declared File Scope), blocking on the first repo with off-scope
files and naming it. Both return shapes preserved (ReviewResult; {blocked,message}).
New workspace-paths.ts repo-prefix helper (deriveRepoForPath/splitRepoScopedPath/
deriveRepoScopeSubset; segment-wise longest-prefix match, unscoped fallback) —
master U5 reuses it. Singular non-workspace path unchanged. 16 new fixture tests.
Gate green: typecheck, lint, build, test:gate (649+58).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
In workspace mode the executor now captures changes and verifies worktree
invariants per acquired sub-repo instead of degrading to empty against the
non-git root. Post-session capture (:7898) gains a workspace branch that loops
task.workspaceWorktrees and reuses captureModifiedFiles(repo.worktreePath,
repo.baseCommitSha, …) per repo — inheriting resolveDiffBaseRef's merge-base
fallback (repo baseCommitSha may be undefined) and the filterFilesToOwnTaskCommits
contamination/divergence audit — then prefixes each repo's files with the repo
path into task.modifiedFiles. Branch attribution runs per sub-repo (cwd), never
against the root. The no-op assertCleanBranchAtBase is not iterated.
verifyWorktreeInvariants is un-stubbed for workspace mode: it iterates every
workspaceWorktrees entry asserting toplevel match + HEAD on fusion/<id>, and
returns the FIRST failing repo while preserving the exact discriminated union
{ok:true} | {ok:false; reason:'wrong_toplevel'|'wrong_branch'|'no_commits';
observed; expected} (the :10889 consumer switches on reason for requeue/handoff)
— the new repo field is additive. Singular non-workspace path unchanged.
Real two-repo fixture tests (capture A+B repo-prefixed vs own base, undefined-base
fallback, foreign-commit contamination audit, wrong_branch verify failure,
single-repo regression). Gate green: typecheck, lint, test:gate (649+58).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
ce-code-review (5 personas) on Phase A. No P0; the workspace-root-removal path
was ruled out and the contract changes verified additive. Applied:
P1: fn_acquire_repo_worktree now catches WorkspaceRepoAcquireBusyError (and
generic failures) and returns a sanitized retryable isError instead of an
uncaught throw into the agent loop; runContext is forwarded so audit/log keep
attribution. Per-repo acquired worktree paths are now registered into the
executor's activeWorktrees Set (via an onAcquired callback) — previously the Set
only held the browse-only root, making the U1 per-repo liveness invariant hollow.
Post-acquire identity-guard install and base-SHA capture are now non-fatal
(log-and-continue): a hook/branch failure no longer strands the on-disk worktree
(the worktree is usable without the guard; an undefined baseCommitSha is already
an accepted state).
P2: the KTD3 settings-strip also strips settings.baseBranch (resolveFromSettings
falls back integrationBranch → baseBranch, so a shared baseBranch leaked); the
workspaceWorktrees write re-reads the task fresh before merging to avoid a
sibling-repo clobber on sequential acquires (store-level atomic merge deferred to
Phase B); the busy-path logging is wrapped so it can't mask the busy error; the
TaskCard memo compares key-sets not counts; the stuck-kill no-op for workspace
tasks is now logged; the exclusivity check-then-act synchrony is documented.
Residuals (Phase B): per-repo worktree teardown, orphan-scan coverage, reaper
dedup, store-level atomic merge.
Gate green: typecheck, lint, build, test:gate (649+58), affected (25 + TaskCard 251).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
acquireWorkspaceRepoWorktree now hardens each sub-repo worktree at acquisition:
(1) installs the identity guard with the executor's settings args
(commitMsgHookEnabled/taskPrefix/taskAttributionTrailerName) for single-repo
parity — it was installing no guard before; (2) captures a per-repo
baseCommitSha local-first against the repo's resolved integration branch via
resolveIntegrationBranch(repoAbsPath, {...settings, integrationBranch: undefined})
— stripping the shared override so each sub-repo falls through to its own
origin/HEAD, not a project-wide branch; (3) persists baseCommitSha into the
workspaceWorktrees[repo] entry (Task type extended); (4) registers same-sub-repo
exclusivity on the sub-repo path via activeSessionRegistry under a distinct
"workspace-repo-acquire" kind (released in finally), so two concurrent workspace
tasks contending for the same sub-repo are serialized (throws
WorkspaceRepoAcquireBusyError). Idempotent re-acquire short-circuits.
resolveCapturedBaseCommitSha gains an optional trailing integrationBranch param
defaulting to "main", so existing single-repo callers + base-commit-capture
real-git tests stay green. New audit events worktree:workspace-repo-acquire-busy
/-failed. 6 new real-fixture tests.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
In workspace mode (loadWorkspaceConfig present), the executor now skips the
root acquireTaskWorktree({rootDir}) and every intervening rootDir git preflight
(base-commit capture, contamination, identity-guard, verifyWorktreeInvariants),
runs the agent session rooted at the non-git workspace root (cwd=rootDir,
browse-only; task.worktree never set), and tracks activeWorktrees as a per-task
Set<path>. scopePromptToWorktree is a no-op in workspace mode. The non-workspace
path is unchanged (every change branches on this.workspaceConfig; a single-repo
task holds a one-element Set).
Converted every activeWorktrees consumer to membership semantics (feasibility-
verified list): findActiveWorktreeOwner, hasActiveWorktreeBinding, the FN-6736
phantom-binding reclaim, listWorktreeHolders (flat-maps a Set into N holder rows
— verified the FN-6782 reaper keys off taskId only, so slot accounting is
unaffected), the conflict-set iteration, the three deleteActive* unregister
resolvers (loop every path), cleanup, getWorktreePath (undefined for a
multi-worktree workspace task), and the verifyWorktreeInvariants singular
resolution (gated off in workspace mode — per-repo verify returns in Phase B).
Rewrote executor-workspace.test.ts from vi.mock-the-subject to a real two-repo
git fixture harness (_workspace-fixture.ts, shared with later units), 13 tests.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Post-fix verification review (correctness + adversarial + reliability, unanimous
P0) found that the earlier retry-burn fix introduced an infinite loop: parking a
WorkspaceTaskMergeError task with status:null + mergeRetries:0 passes every
auto-merge eligibility gate (canMergeTask short-circuits only on status==='failed'),
so the cooldown sweep re-enqueues it every tick → guard re-throws → re-park, forever.
- Park with status:'failed' (keep mergeRetries:0). canMergeTask now blocks the
auto-sweep; a human's manual merge still works because it flows through the
manual-resolver branch (rejectMergeResolvers), which bypasses canMergeTask — so
'failed' does not block manual retry (the original comment's worry was wrong).
- Detect the error via `err instanceof Error && err.name === "WorkspaceTaskMergeError"`,
matching the VerificationError/MergeAbortedError convention and bundle-safe across
the @fusion/core→@fusion/engine boundary (drops the now-unused class import).
- Document that the dispatch door guard is a fast-fail only; the unconditional
chokepoint guard inside runAiMerge is the authoritative enforcement.
- Add a regression test asserting the auto-merge park sets status:'failed' (not null).
Gate green: lint, typecheck, build, test:gate (649+58), project-engine (81).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Applies ce-code-review (autofix) feedback — 5 reviewers, P1s corroborated.
F1 [P1, ×4 reviewers] Guard the merge chokepoint, not just the 4 doors. The
per-caller `getTask().catch(()=>null); if(t) assert` pattern failed open on a
transient read, and runAiMerge re-read the task unguarded — so a workspace
task could reach git work against the non-git root. Added a named
WorkspaceTaskMergeError and call assertNotWorkspaceTaskMerge inside runAiMerge
(the sole merge path) and the deprecated aiMergeTask body; door guards remain
as fast-fail defense-in-depth.
F2 [P1] The dispatch catch treated the guard throw as a merge failure and set
mergeRetries=MAX, permanently blocking manual retry. It now recognizes
WorkspaceTaskMergeError and parks without burning retries.
F3 [P2] Deprecation-warning test asserted toBeLessThanOrEqual(1) — vacuously
true on zero emissions. Now resets the per-project flag and asserts the
warning fires exactly once and not again on a second deterministic merge.
F6 [P2] The once-per-process warning flag suppressed the notice for all other
projects in a multi-project host; now keyed per project (Set by cwd).
F5/F7/F8 [P3] @deprecated propagated to the aiMergeTask barrel re-export; CLI
runTaskMerge guard moved inside the formatted try/catch; FNXC placeholder
timestamps corrected; test .at(-1) -> length index.
Documented as residual (deferred to master-plan U8, not bugs in U0's window):
self-healing auto-finalize + store.mergeTask are additional merge-completing
paths not hardened here — workspace tasks are not end-to-end runnable until
master-plan Phase A, and U8 makes self-healing workspace-aware.
Gate green: typecheck (29 projects), lint, build, test:gate (649+58),
affected tests (206+4).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Make runAiMerge (the FN-5633 clean-room AI merge path, already the default)
the sole merge path; soft-deprecate the legacy aiMergeTask pipeline. Phase 0
of the workspace-mode master plan — a standalone merge-consolidation refactor
that lands first so downstream workspace work targets one merge function.
U1 — collapse the dispatch: project-engine.ts now calls runAiMerge
unconditionally; the two direct callers that bypassed the dispatch
(dashboard.ts onMergeImpl in --no-engine mode, task.ts runTaskMerge /
`fn task merge`) now route to runAiMerge too. Export runAiMerge from
@fusion/engine.
U2 — soft-deprecate: aiMergeTask is @deprecated (body retained for a later
deletion pass; shared helpers runAiMerge imports, e.g.
captureSingleCommitLandedMetadata, left intact). merger.mode "deterministic"
is annotated deprecated and made inert (type + field kept — published
@runfusion/fusion surface); the dispatch logs a one-time deprecation warning
and routes to runAiMerge. Changeset added (minor).
U3 — R7 workspace merge-boundary guard: shared @fusion/core predicate
assertNotWorkspaceTaskMerge(task) rejects tasks with populated
workspaceWorktrees at all four merge entry points (dispatch, store.mergeTask,
onMergeImpl, runTaskMerge) with an error naming master-plan U6. Covers the
window until per-repo merge support lands; U6 removes it.
U4 — deterministic-mode blast-radius audit: no production project, CI config,
or seeded setting pins merger.mode "deterministic"; only four engine tests
used it to drive the dispatch to aiMergeTask as a mockable seam — migrated to
mock runAiMerge instead. Other module-level aiMergeTask mocks were dead under
the default "ai" mode or test aiMergeTask directly (body retained), so they
are unaffected.
Also removes an unused acquireWorkspaceRepoWorktree import inherited from the
foundation branch (executor.ts) that was failing lint; master-plan U1 re-adds
it with its per-repo usage.
Merge gate green: lint, typecheck (29 projects), build, test:gate
(649 + 58), plus the migrated (114) and new predicate (4) tests.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Allow registering a non-git parent directory that contains multiple git
repositories as a single Fusion project. The agent acquires per-repo
worktrees on demand via a new `fn_acquire_repo_worktree` tool as it
discovers it needs to work in each sub-repo.
This commit lays the foundation:
- detectWorkspaceRepos / loadWorkspaceConfig / saveWorkspaceConfig in
@fusion/core (config persisted to .fusion/workspace.json)
- Task.workspaceWorktrees data model + store plumbing (per-repo
worktree/branch map, distinct from the singular task.worktree)
- acquireWorkspaceRepoWorktree wraps acquireTaskWorktree per sub-repo,
clearing the singular worktree/branch fields so each sub-repo gets a
fresh worktree instead of resuming a sibling repo's worktree
- fn_acquire_repo_worktree agent tool + workspace prompt injection
- executor git-repository validation bypassed when a workspace config
is present
- CLI `fn init` detects a non-git dir containing sub-repos and writes a
workspace config
Known gap (intentionally left for design discussion, see PR): the
executor's main worktree-acquisition path still assumes a single git
root and is not yet workspace-aware. End-to-end execution (skipping the
root acquisition, per-repo merge, per-repo session scoping) is a
follow-on once the execution model is agreed.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Tighten the executor agent guidance and the fn_run_verification tool guidance so
verification stays scoped to changed files instead of running the full workspace
test suite, which (for a foundational-package edit) reverse-expands across the
whole workspace and stalls the task.
- agent-prompts.ts: remove the "during final integration" blanket permission to
run workspace-wide suites; name the forbidden full-run commands explicitly.
- run-verification-tool.ts: strengthen BOUNDED_VERIFICATION_GUIDANCE to forbid
`pnpm test:full` / `pnpm verify:workspace` / whole-package runs as verification.
Engine + core typecheck pass; run-verification-command and executor-review-verdicts
prompt-assertion suites stay green (93 tests).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
- Read hasResumableProgress from latestTask (the store snapshot fetched
at ~9226), not the frozen dispatch-time task param, so a freshly
dispatched task that commits step progress mid-session is preserved on
a hard-cancel teardown — not just the re-dispatch case (greptile P1)
- Add companion regression test for the fresh-task-with-in-session-progress
case, which fails against the stale-task snapshot (greptile P1)
- Reformat the requirement comment block to the canonical FNXC heading
(FNXC:WorkflowLifecycle yyyy-MM-dd-hh:mm:) convention (CodeRabbit)
Fusion-Task-Id: FN-6722
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
When the engine aborted in-flight work mid-execution and bounced the task
back to `todo`, the single-session teardown cleared the task `branch` and
re-queued without `preserveResumeState` — resetting every step to `pending`
(store.moveTaskInternal resetAllStepsToPending) and dropping the pointer to
the commits already on the task branch. The next dispatch then re-planned
from Step 0 and the committed work was stranded, observed as FN-6722 losing
all its progress and getting stuck in todo.
The teardown now keeps the branch and moves with `preserveResumeState`
whenever the task has resumable step progress, matching the sibling
step-session (executor ~8065) and pause-park (executor ~1826) paths, so
execute() resumes onto the existing branch from the first incomplete step.
The worktree is still removed to free its concurrency slot (FN-6782) — only
the durable pointers (branch + step state) are kept.
Adds a regression test driving the exact single-session catch teardown.
Fusion-Task-Id: FN-6722
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Move externalEngines.delete(projectId) to immediately after acquiring the
singleton lock instead of after engine.start() succeeds. If a project was
marked external, the holder exits, acquire succeeds, but start() then throws,
the success-path delete never ran and hasRunningEngine() reported a phantom
engine forever. Added a regression test for the failed-takeover path.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
- Stop reconciliation/startAll/onProjectAccessed from warning every tick for
externally-owned engines: swallow EngineAlreadyRunningError in the outer
catches (it's expected and already logged once in createAndStart)
- Add FNXC:DashboardHealth requirement-trace comments on the externalEngines
field and the dashboard hasRunningEngine health check
- Add regression test: reconciliation stays quiet across ticks for an
externally-owned engine (inner refusal logged once, outer failure suppressed)
- Add regression test: hasDashboardEngine legacy fallback to getAllEngines when
hasRunningEngine is unavailable on the manager
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The dashboard's engine-availability health check only counted engines
this process started. A second launch (e.g. `pnpm dev dashboard`
alongside an already-running `fusion`) is correctly refused the
per-machine engine singleton lock, so its engine map stays empty and
the dashboard showed a false "engine not running" banner even though an
engine was live on the machine.
ProjectEngineManager now records projects whose singleton lock is held
by another process (via EngineAlreadyRunningError) and exposes
hasRunningEngine(), which the health endpoint consults so the banner
reflects machine-level truth. Reconciliation still retries so this
process takes over if the other exits, and the "refusing to start" log
fires once per project instead of on every 30s reconciliation tick.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
- Add JSDoc to resolveCustomProviderApiType (docstring coverage + AGENTS.md
jsdoc convention) and convert the inline rationale to FNXC format.
- FNXC-format the test rationale comment.
- Strengthen the regression test per FN-5893 (fix the invariant, not the repro):
- negative assertion that no provider is ever registered with the bare
"anthropic" api key;
- assert every api key passed to registerProvider is one pi-ai's registry
actually registers, so a typo in any arm (not just anthropic) fails here.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
resolveCustomProviderApiType mapped the `anthropic-compatible` provider type
to the api key "anthropic", but pi-ai (@earendil-works/pi-ai) registers the
Anthropic Messages API under "anthropic-messages". Any custom provider
configured as anthropic-compatible selected a model whose `api` did not match
a registered provider, throwing "No API provider registered for api: anthropic"
at stream time (the model registered fine, but failed when a task tried to
stream).
The openai-responses and default (openai-completions) arms already map to real
registry keys and work; only the anthropic arm pointed at an unregistered key.
Extend the existing custom-provider registration test (which covered
openai-compatible and openai-responses but not anthropic-compatible) with a
regression assertion that anthropic-compatible maps to "anthropic-messages".
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
- Add fire-time safety guard tests: assert auto-continue aborts when the task
becomes paused, user-paused, moved out of todo, or deleted during the backoff
window (greptile P1 / coderabbit) — previously the guard was untested.
- Split the user-pause/global-pause negative test into it.each so a failure
names the offending case (greptile P2).
- Document that the exhausted-retry fallback leaves the shared graphResumeRetryCount
budget at MAX and that it cross-drains with the transient-resume path (greptile P2).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Resolve conflicts from main's analytics schema additions (plugin
activations, per-model token buckets) against the PR's contract-assertion
type column:
- db.ts: renumber behavioral-verification migration 124 -> 126, bump
SCHEMA_VERSION to 126 so it follows main's migrations 124/125
- core/roadmap tests: adopt main's SCHEMA_VERSION-constant assertions
instead of stale literal 124
- test-quarantine.json: keep all four quarantine entries from both sides
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
- step() test helper now carries skillName, so the WS-6 round-trip fixture
actually exercises the INVERSION CONTRACT for skillName (was silently dropped).
- executeWorkflowStep now strips an inherited FUSION_HEADLESS on board runs
(unattended=false), preserving the U3 default-safe invariant — a board step
nested under a headless-env parent could otherwise skip user questions.
Added a regression test for the inherited-env strip.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The stepwise-coding IR had no workflow-step seam node, so a per-task
enabledWorkflowSteps entry (e.g. browser verification) would never execute.
Add the seam node on the success path (steps -> workflow-step -> review,
once post-foreach) and declare browser-verification as an optional step,
matching the coding workflow. Covers the dead-toggle gap with resolver and
engine execution-divergence tests.
When the engine hard-cancels in-flight work during a pause/resume cycle and the
workflow graph run ends with the task re-queued to `todo`, the executor used to
leave it for a fresh scheduler dispatch and fire a spurious failure
notification. It now continues the agent session in place via a bounded internal
retry (reusing the graph-resume retry budget + backoff), falling back to the
benign todo re-queue only after retries are exhausted.
- Scoped strictly to engine-internal aborts via a typed `isEngineInternalAbort`
discriminant; genuine user/global/task pauses are never auto-resumed.
- Re-checks the task at retry fire time and aborts the auto-continue if it was
paused, moved, or deleted during the backoff window.
- Clears any stale `failed` status and emits an `Auto-recovered:` log so the
deferred failure notification is suppressed.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Address confirmed code-review findings on the CE workflow-step change:
- Graph-path spawn lifecycle (adversarial A-1/A-2): the graph path returns from
execute() before its outer finally that calls terminateAllChildren, so U8's new
coding-mode fn_spawn_agent children orphaned their sessions/worktrees and their
ids accumulated in the per-parent spawn budget, starving later steps' fan-out.
Call terminateAllChildren in maybeExecuteWorkflowGraph's finally (mirrors the
non-graph cleanup).
- INVERSION CONTRACT parity (api-contract AC-2 + testing TF-001): add skillName to
the workflow-steps-to-ir round-trip projections + a skill-step fixture, so the
contract the comment claims is actually asserted.
- Silent skill-load degradation (adversarial A-3 / Risk-4): warn when a step names
a skill but FUSION_CE_SKILLS_DIR is unset, instead of failing silent.
- Dead branch (maintainability M-01): drop the always-false unattendedRun guard;
keep the delete + extension-point comment.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Two engine tests for the new compound-engineering workflow-step wiring:
- conventions: assert the exported preamble carries the await-input sentinel,
FUSION_HEADLESS degrade, and path-confined persona/systemPromptOverride fan-out.
- executor: drive runGraphCustomNode + executeWorkflowStep and assert skillName is
carried onto the synthesized step, requestedSkillNames merges bare+namespaced with
additionalSkillPaths=[FUSION_CE_SKILLS_DIR], fn_spawn_agent present only in coding,
FUSION_HEADLESS only when unattended, and the verdict-JSON contract is required
only for gate/skill-less steps (relaxed for non-gate skill steps).
Session layer is mocked (asserts engine-owned wiring, not a model run); a full
model-driven e2e remains a documented residual.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The builtin compound-engineering workflow runs via runGraphCustomNode, which
never loaded the named skill or threaded the plugin-injected runtime env, and
fn_spawn_agent was registered only in the main session. This wires the real seam:
- U8: thread injected FUSION_CE_* env into skill/model graph steps (shared
buildInjectedRuntimeEnv helper); register createSpawnAgentTool for coding-mode
skill steps (readonly still strips spawn).
- U1: merge the step's skillName (namespaced + bare) into requestedSkillNames and
pass FUSION_CE_SKILLS_DIR as additionalSkillPaths so the bundled SKILL.md is
discovered and selected.
- U2: prepend the Fusion workflow-step conventions preamble (await-input sentinel,
FUSION_HEADLESS degrade, persona fan-out via systemPromptOverride).
- U3: explicit unattended opt-in sets FUSION_HEADLESS=1 (default-safe board run).
- U9: path-confined persona read documented in the preamble; accepted
write-capability posture documented at the coding-mode tool registration.
- KTD-6: verdict-JSON contract required only for gate / skill-less steps.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
- Reliability (P3): emit an `Auto-recovered:`-prefixed log on the
benign-todo stale-failure clear path so NotificationService
proactively cancels the pending failure timer (recoveredStatus path)
instead of relying only on the fire-time re-check, which is
race-contingent when failureNotificationDelayMs is near 0. Scoped to
the actual-clear path so the common no-failure re-queue isn't
mislabeled as a recovery.
- Project-standards (P3): add the required yyyy-MM-dd-hh:mm stamp to the
new FNXC comments (AGENTS.md FNXC_LOG convention).
- Maintainability (P3): extract the scheduler "queued" waiting marker to
a named SCHEDULER_WAITING_STATUS constant.
- Testing: pin the guard's skip on a clean todo row, assert the
Auto-recovered log fires on the stale-failure path, and add a
paused+unassigned in-progress idle case (paused early-return wins over
the ephemeral active-session path).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
A pause/resume abort parked status:"failed" on an earlier non-todo
observation stays dispatchable (scheduler filters on column+paused, not
status, scheduler.ts:1288) and re-enters the FN-6782 benign-todo branch.
That branch logged "benign" but left status:"failed"/error on the row,
so the board kept showing it failed and the deferred failure
notification fired (notification-service fire-time check sees
status==="failed") — contradicting the benign log. recoverPausedAbortFailures
that would clear it is suppressed during global/engine pause
(self-healing.ts:8125), so the failure survived the pause window.
Reconcile the row with the benign reclassification: clear status/error
when the live row carries them, so the board agrees it's benign and the
pending notification is suppressed at dispatch.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Addresses findings from a multi-agent review of the two prior fixes.
P0 (executor.ts): the stale-conflict recovery force-removed worktreePath with
no bounds check; that path can come from a git admin entry resolving outside
.worktrees/. Now refuses unless the path is inside the worktrees dir, not a
symlink (realpathSync), not a registered worktree, and not actively owned, and
re-verifies liveness in the catch instead of trusting the error string. Also
excludes spawn failures (spawn git ENOENT) from the stale-path classification.
worktree-pool.ts: resolveGitdirPointer -> dotGitPointerIsDangling. Reaps only
when a .git link's gitdir target is confirmed missing; a real .git dir,
unparseable pointer, or any read/stat failure is treated as NOT dangling
(conservative) so a transient read error on a live worktree can't trigger rm.
Drops the string|"directory"|null sentinel union.
core store.ts: bypass the reconcile recency window when the live task table is
empty (corruption/restore: surviving task.json keep old mtimes) and when
fusion.db was auto-recovered on startup, so .recover row loss isn't stranded.
Adds an ignoreRecencyWindow option.
Tests: executor recovery + out-of-bounds refusal, unparseable .git skip,
recency boundary, empty-DB/forced bypass. engine 135 + core 12 green.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Directories under .worktrees/ that survive with a dangling .git pointer
(present on disk, but their .git/worktrees/<name> admin entry is gone) are
invisible to `git worktree list`/`prune` yet collide with freshly generated
worktree names. The executor's conflict cleanup then fails with
"is not a working tree", failing the workflow graph at node 'execute' after
3 attempts.
- executor.ts: extend FN-4813 stale-conflict recovery to also treat
"is not a working tree" and ENOENT (not just "validation failed, cannot
remove working tree") as "no live worktree here" — prune the admin entry,
force-remove the leftover dir, and proceed with fresh creation.
- worktree-pool.ts: reapOrphanWorktrees skipped any dir on mere .git-file
presence, contradicting its own documented invariant. Resolve the .git
pointer and only skip when the gitdir target exists; reap dangling
pointers like any other orphan so they stop accumulating across runs.
- Tests for both the dangling (reaped) and valid (skipped) .git cases.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Agents stop heartbeating during long legitimate work (e.g. a verification
step blocked on a multi-minute test command). The 5-minute floor could
misread a busy agent as dead and reclaim its in-progress task mid-run.
Raise MIN_HEARTBEAT_STALENESS_MS to 10 minutes and strengthen the floor
test (7-minute-silent fast-interval agent stays healthy — would have read
stale under the old 5-minute floor). Engine typecheck clean; heartbeat
suite 148/148 pass.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The gate test asserted the OLD behavior — a paused graph exit in the `todo`
column parked `status:"failed"` with "operator action required". FN-6782
made the todo case benign (no failed park; benign log + cleared marker), so
split the parameterized test: `todo` now asserts the benign path (never
parked failed), `done` keeps the operator-action surfacing (log only, no
park). Full engine-core gate suite passes (644/644).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
