When verify_task_against_plan returns EMPTY because the requested changes
already live in origin/main (parallel work, earlier PR, race between
siblings), the worker had no clean exit: update_job_status only accepted
running|done|failed. 'failed' triggered the PBI fail-cascade which then
overwrote the error column with 'cancelled_by_self' and cancelled all
sibling tasks of the PBI — see Scrum4Me job cmovkur8 / T-695 for the
reference incident.
This change introduces a fourth status and tightens the cascade:
ST-1273 — 'skipped' exit in update_job_status (T-706 + T-707)
- src/tools/update-job-status.ts: status enum + DB_STATUS_MAP +
resolveNextAction now include 'skipped'. cleanupWorktreeForTerminalStatus
signature widened to ('done'|'failed'|'skipped'); SKIPPED uses keepBranch
semantics identical to FAILED (no push, no branch keep). New input guard:
'skipped' is only valid for TASK_IMPLEMENTATION jobs and requires a
non-empty error (≥10 chars) explaining the reason — it bypasses the
verify-gate, the auto-PR, the SprintRun finalize/fail paths and the
PBI fail-cascade. Locks are still released on terminal exit.
- Tool description spells out when to pick 'skipped' so MCP clients see it.
- New __tests__/update-job-status-skipped.test.ts: resolveNextAction with
'skipped' (wait_for_job_again / queue_empty), and cleanupWorktreeForTerminalStatus
with status='skipped' (keepBranch=false even with a branch reported,
defers cleanup with active siblings).
ST-1274 — cascade ignores SKIPPED + appends trace (T-708 + T-709)
- src/cancel/pbi-cascade.ts: runCascade reads job.status, returns EMPTY
when status === 'SKIPPED' (no sibling cancel). Trace persistence now
reads the current error first and writes `${original}\n---\n${trace}`
(truncated at 1900 chars), so the original failure cause is preserved
for forensics instead of being overwritten.
- New cases in __tests__/cancel-pbi-cascade.test.ts: SKIPPED entry-guard
(no findMany / updateMany / update), original error preserved with
trace appended after '---', trace-only fallback when no original
error, 1900-char truncation keeps the head of the original.
All 282 scrum4me-mcp tests pass; tsc build clean.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Three findings from PBI-47 review:
P1 — primary_worktree_path scheiden van lock-volgorde
setupProductWorktrees acquired locks in alphabetical order (deadlock prevention)
but also returned worktrees in that order, so worktrees[0] could point at a
secondary product when its id sorted before the primary's. Lock-acquire stays
sorted; output now preserves caller's input order so worktrees[0] is always
the primary.
P1 — Idea-claim rollback bij worktree setup failure
setupProductWorktrees runs after tryClaimJob has already flipped the job to
CLAIMED. A failure in lock-acquire/git-fetch/reset/sync left the job hanging
until the 30-min stale-reset and the lock-map populated. Wrapped in try/catch
with releaseLocksOnTerminal + rollbackClaim mirror of the task-pad behaviour.
P2 — SPRINT mark-ready fallback when last task didn't push
The mark-ready path used updated.pr_url, which is null when the closing task
was verify-only or had no diff. Now falls back to a Prisma findFirst on the
SprintRun's earliest job with pr_url IS NOT NULL.
Tests: 31 files, 243 passing (incl. new input-order regression for setupProductWorktrees).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Schema sync vanaf upstream Scrum4Me (v77617e8): FAILED toegevoegd aan
Task/Story/Pbi/SprintStatus, nieuw SprintRunStatus + PrStrategy enums,
SprintRun model, ClaudeJob.sprint_run_id, Product.pr_strategy.
T-18 — propagateStatusUpwards in src/lib/tasks-status-update.ts.
Real-time cascade Task → Story → PBI → Sprint → SprintRun bij elke
task-statuswijziging. Bij FAILED cancelt sibling-jobs in dezelfde
SprintRun. PBI-status BLOCKED blijft handmatig. Houd deze helper bit-
voor-bit synchroon met Scrum4Me/lib/tasks-status-update.ts.
updateTaskStatusWithStoryPromotion blijft als BC-wrapper.
T-19 — wait-for-job.ts claim-filter. Task-jobs worden alleen geclaimd
als hun SprintRun status QUEUED of RUNNING heeft. Idea-jobs blijven
ongefilterd. Bij eerste claim van een QUEUED SprintRun → RUNNING
binnen dezelfde tx (race-safe).
T-20 — update-job-status.ts roept propagateStatusUpwards aan na elke
task DONE/FAILED. Bestaande cancelPbiOnFailure-aanroep blijft voor
PR-cleanup; sibling-cancellation overlap is harmless (idempotent).
T-21 — classify.ts (verifier) leest nu ook "--- a/<path>" zodat
delete-only commits niet meer als EMPTY worden geclassificeerd.
Bug had eerder geleid tot ten onrechte FAILED-status op cmotto5h en
cmotto5i (06-05-2026); zou met cascade-flow een hele sprint laten
falen.
Cleanup: create-todo.ts en open_todos in get-claude-context.ts
verwijderd (Todo-model is op main gedropt). Endpoint geeft nu
open_ideas terug — ideeën die niet PLANNED zijn.
Status-mappers (src/status.ts) uitgebreid met failed.
Tests: 184/184 groen (180 → 184; vier nieuwe delete-only classify-tests
en herwerkte propagate-status tests).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Wanneer een TASK_IMPLEMENTATION-job FAILED wordt, cancelt
cancelPbiOnFailure alle queued/claimed/running siblings binnen
dezelfde PBI (over alle stories heen) en draait gepushte commits
ongedaan:
- Open PR → gh pr close --delete-branch (PR-close + remote-branch-
delete in één).
- Gemergde PR → revert-PR via git revert -m 1 <mergeSha> in een
korte worktree, gepusht naar revert/<orig>-<jobid>, gh pr create
zonder auto-merge (mens reviewed).
- Branch zonder PR → best-effort git push origin --delete.
Race-protectie: update_job_status weigert nu een statuswijziging op
een job die al CANCELLED is met een specifieke JOB_CANCELLED-error,
zodat een parallelle worker zijn lokale werk weggooit ipv een DONE
te forceren. Idempotent — een tweede cascade voor dezelfde PBI is
een no-op. Non-blocking — alle fouten worden warnings in de trace
op de oorspronkelijke failed job zijn error-veld; cascade throwt
nooit naar de caller.
Niet in scope: per-product opt-out, sprint-niveau cascade,
idea-job cascade.
11 nieuwe vitest-cases dekken DB-cascade, branch-grouping, open/
merged/no-PR paden, repo-root-mismatch en de never-throws-garantie.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
update_job_status accepts optionele model_id + 4 token-velden conform het
runbook-contract (mcp-integration.md:42). De waarden komen niet van de agent
zelf maar van scripts/persist-job-usage.ts, een PostToolUse-hook die het
lokale Claude Code transcript (~/.claude/projects/.../*.jsonl) leest en de
usage tussen de laatste wait_for_job en update_job_status optelt.
Geen Anthropic API-key nodig — alle data staat al lokaal op disk omdat
Claude Code per assistant-message het API usage-blok logt
(input_tokens, output_tokens, cache_creation_input_tokens,
cache_read_input_tokens + message.model).
Robustness:
- Subagent (isSidechain: true) lines worden geskipt om double-counting
te voorkomen tegen subagents/-subdirectory transcripts.
- Lines worden gededupliceerd op uuid (branching/resumption).
- model_id wordt genormaliseerd: claude-opus-4-7[1m] -> claude-opus-4-7-1m
zodat de [1m]-variant op een aparte model_prices-rij kan matchen.
- Hook is non-blocking: elke fout logt een warning en exit 0.
Hook-config in .claude/settings.json met SCRUM4ME_MCP_DIR-fallback zodat
de agent vanuit een product-worktree (andere cwd) ook werkt mits de user
de hook in ~/.claude/settings.json kopieert.
16 nieuwe vitest-cases voor parseTranscript, computeUsageFromTranscript,
normalizeModelId en persistJobUsage.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Synchronous, non-blocking count of active ClaudeJobs per product or across
all accessible products. Registers check_queue_empty MCP tool with optional
product_id scope, productAccessFilter AuthZ, tests, and README docs.
* feat(ST-mhj9f2la): add set_pbi_pr MCP tool
- Add pr_url and pr_merged_at fields to Pbi model in schema
- Implement set_pbi_pr tool: writes pr_url, clears pr_merged_at (idempotent)
- AuthZ via requireWriteAccess + userCanAccessProduct through pbi.product_id
- 10 tests: happy path, not-found, no-access, demo-denied, schema validation
- Update README tools table and bump version to 0.2.0
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* feat(ST-mhj9f2la): add mark_pbi_pr_merged MCP tool
- Implement mark_pbi_pr_merged: sets pr_merged_at = now() on a PBI
- Requires pr_url to be set; returns error if not (geen gekoppelde PR)
- Idempotent: re-calling overwrites the timestamp
- AuthZ via requireWriteAccess + userCanAccessProduct through pbi.product_id
- 6 tests: happy path, no-pr_url, idempotent, no-access, not-found, demo-denied
- Update README tools table with mark_pbi_pr_merged entry
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* docs(ST-mhj9f2la): expand README with set_pbi_pr + mark_pbi_pr_merged docs
Add full signature/input/output/error documentation sections for both
new tools, following the verify_task_against_plan pattern.
Version already bumped to 0.2.0 in earlier commit.
Tag + MCP_GIT_REF pin in scrum4me-docker to be done by maintainer after merge.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
---------
Co-authored-by: Claude Sonnet 4.6 <noreply@anthropic.com>
Two related fixes for the agent-workflow defects exposed by the
2-May-2026 batch:
1. **Cross-repo task routing** (`task.repo_url` override).
`resolveRepoRoot` now consults `task.repo_url` first; matches against
per-repo env-var (`SCRUM4ME_REPO_ROOT_REPO_<name>`),
`~/.scrum4me-agent-config.json` `repoRoots[<name>]`, and finally
`~/Projects/<name>/.git`. Falls back to product-level resolution
when null. Tasks tracked under one product but targeting another
repo (e.g. MCP-server tasks under the main product's PBI) now work.
`getFullJobContext` exposes `task.repo_url` to the agent.
`attachWorktreeToJob` accepts and forwards it.
2. **Orphan-branch cleanup** in `createWorktreeForJob`.
Previously a name-collision suffixed with a timestamp, leaving the
agent on an unpredictable `feat/story-XXX-<ms>`-name. Worse, in the
2-May batch the agent ended up reusing an orphan branch from an
earlier story (`feat/story-x35n155c`) and pushed to a remote ref
that did not exist, causing 'src refspec does not match any'.
Now: detect orphan, attempt to remove its (stale) worktree if any,
delete the local branch, and recreate with the predictable name.
Timestamp-suffix is the last resort.
Vendor submodule bumped to pick up `Task.repo_url` from Scrum4Me #54.
Tests: 129/129 — `suffixes branch name with timestamp` updated to
`removes orphan branch and reuses the predictable name`.
Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Sluit story 'Verify-gate uitbreiden' in PBI 'Agent verify-flow hardening' af.
The previous gate weighed only EMPTY: any PARTIAL or DIVERGENT verify
slipped through. The Insights batch (2 May 2026) showed why that's
weak — agent-jobs claiming DONE while only delivering helpers, not
the requested UI components, with verify=DIVERGENT/PARTIAL accepted.
New decision matrix:
null → reject (run verify_task_against_plan)
EMPTY + !verify_only → reject
EMPTY + verify_only → allowed
ALIGNED → always allowed
PARTIAL/DIVERGENT
required=ALIGNED → reject (strict task)
required=ALIGNED_OR_PARTIAL (default) → allowed only if summary
≥20 chars (acknowledge drift)
required=ANY → allowed (refactor escape hatch)
`update_job_status('done')` now reads `task.verify_required` from the DB
(field added in Scrum4Me PR #53) and passes it + `summary` to the gate.
Tool description updated with the new rules.
Vendor submodule synced to pick up the schema enum.
Tests: 129/129 (was 120 + 9 new combinatorial gate tests).
Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Implementeert vier open stories uit PBI 'Veilige Claude-agent-workflow':
**Branch per story (cmon11tbe001zbortx35n155c)**
- `resolveBranchForJob`: zoek sibling-job in dezelfde story; reuse z'n
branch (1 PR per story i.p.v. per task).
- Branch-naam: `feat/story-<8-char>` voor nieuwe stories.
- `createWorktreeForJob` kent nu `reuseBranch=true`: detecteert stale
sibling-worktree die de branch nog vasthoudt en verwijdert die eerst.
- `attachWorktreeToJob` neemt `storyId` mee.
**PR-hergebruik (zelfde story)**
- `maybeCreateAutoPr`: als sibling-job in story al een pr_url heeft,
hergebruik die zonder nieuwe `gh pr create`-call. PR-titel komt nu
van de story (was task) zodat het als 'story-PR' leest.
**Worktree-cleanup uitgesteld bij actieve siblings**
- `cleanupWorktreeForTerminalStatus`: count active sibling-jobs in
dezelfde story; defer als > 0 (volgende sub-task gebruikt branch).
**Worktree-cleanup logging (cmon0jc14001ubortjxf2a2ck)**
- Warning bij ontbrekende repoRoot, met productId + jobId in message.
- Warning bij removeWorktreeForJob-failure met keepBranch in message.
**resolveRepoRoot fallback (cmon0jc14001ubortjxf2a2ck)**
- Convention-based fallback: `~/Projects/<repo-name>` afgeleid uit
`product.repo_url` als noch env-var noch config-bestand iets oplevert.
- `repoNameFromUrl` helper geëxporteerd voor herbruikbaarheid.
**Verify EMPTY-detection edge-case (cmon0kdq6001xbort2kgbcqmr)**
- `classifyDiffAgainstPlan`: na file-paths-check ook content-lines
checken; als alle +/- regels alleen headers of whitespace zijn,
return EMPTY met duidelijke reasoning.
Tests: 120/120 groen (3 nieuwe), tsc clean, build clean.
Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
New src/git/pr.ts helper wraps 'gh pr create'; returns { url } or { error }.
maybeCreateAutoPr() in update-job-status checks product.auto_pr, builds title
from story.code + task.title, writes pr_url to DB. Non-fatal: gh failure logs
a warning and leaves DONE status intact. Also syncs schema: auto_pr on Product,
pr_url on ClaudeJob.
resetStaleClaimedJobs now uses $queryRaw with RETURNING so it can send pg_notify
claude_job_status events per transitioned job. Jobs under the retry limit are
re-queued with retry_count incremented; jobs at ≥2 retries are marked FAILED.
Runs git push -u origin <branch> in the worktree. Detects no-changes
(HEAD = origin/main) before pushing. Classifies push failures into
no-credentials, conflict, or unknown via stderr pattern matching.
5 unit tests covering all paths.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
On DONE/FAILED, resolves repoRoot and calls removeWorktreeForJob (best-effort).
keepBranch=true when status=done and agent reported a branch (push assumed);
false otherwise. Cleanup failures are logged as warnings — DB status is preserved.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
After claiming a job, resolves repoRoot (env SCRUM4ME_REPO_ROOT_<productId>
or ~/.scrum4me-agent-config.json), creates a git worktree, and returns
worktree_path + branch_name in the response. Rolls back claim to QUEUED
on failure. Tool description updated to instruct agent to work in worktree.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Removes worktree dir via `git worktree remove --force` and deletes
the local branch by default; keepBranch=true preserves the branch.
Returns { removed: false } when the worktree path doesn't exist.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
23 unit tests covering parseAcceptanceCriteria, extractKeywords,
checkACStatus, computeDriftScore, lineDiff, and 4 end-to-end scenarios
(plan unchanged, edited, AC missed, no baseline). README documents the
tool with example output and heuristic limitations.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
tasks-status-update.test.ts: 7 tests covering promote, idempotent
promote, partial-siblings no-promote, demote, no-demote-when-not-done,
task-always-updated, and custom tx client pass-through.
get-claude-context-filter.test.ts: 3 tests verifying the OR tasks
filter is passed, sprint_id/status filter is present, and no story
query fires when there is no active sprint.
