roadmap(#229): P1 precompact hook stale-handoff auto-expire by BaseInfinity · Pull Request #227 · BaseInfinity/claude-sdlc-harness

BaseInfinity · 2026-04-23T23:23:22Z

Summary

P1 wizard UX bug found during 2026-04-23 live-fire shepherd session. The precompact-seam-check hook blocked a manual /compact because .reviews/handoff.json was stuck at PENDING_RECHECK from the original review that SHIPPED the hook (review_id: precompact-seam-001).

That review was written BEFORE the self-heal pr_number feature (ROADMAP #209) existed, so there's no PR to auto-query. The handoff is stuck forever until manually flipped.

Every wizard user who did a seam review before #209 has this landmine in their repo.

Fix (shipped in 58d1a86 / `d761028`)

hooks/precompact-seam-check.sh now has two self-heal paths:

(a) docs(cc-2.1.118): use /usage as canonical command (aliases /cost, /stats) #209 PR-linked self-heal (unchanged) — pr_number present → query gh pr view, if MERGED treat as implicit CERTIFIED, silent unblock
(b) roadmap(#212): local-max e2e shepherd foundation #229 stale-handoff auto-expire (new) — no pr_number AND mtime older than SDLC_HANDOFF_STALE_DAYS (default 14) → implicit CERTIFIED with one-line WARN

Paths are mutually exclusive: if [ -n "$PR_NUMBER" ] / else structure ensures an OPEN PR with an old mtime is still treated as live and blocks.

Tests

4 new tests in tests/test-hooks.sh, all passing (19/19 precompact tests green):

test_precompact_unblocks_stale_pending_without_pr_number — stale + no pr → unblock + WARN
test_precompact_still_blocks_fresh_pending_without_pr_number — fresh + no pr → still block (regression guard)
test_precompact_stale_with_pr_number_prefers_self_heal — stale + pr OPEN → still block (pr_number wins)
test_precompact_stale_threshold_override — SDLC_HANDOFF_STALE_DAYS=0 → all PENDING stale

Rationale for 14-day default

Typical Codex review round-trip: <1 day
PR rebase cycles rarely exceed 7 days
14 is ~2x the 99th-percentile review lifetime — wide enough to not fire on legitimate in-flight work, tight enough to rescue forgotten artifacts

Power users can tune via SDLC_HANDOFF_STALE_DAYS env var.

Shepherd loop

TDD RED: 2 new tests fail against old hook
TDD GREEN: all 4 new + 15 existing precompact tests pass
Full test-hooks.sh: 114 passed (1 pre-existing failure on main, unrelated)
Cross-model Codex review (in flight — see .reviews/latest-review.md when complete)
CI green
Merge after Codex CERTIFIED + CI green

CI caught a real portability bug: `stat -f %m` on Linux is `--file-system` (dumps FS info to stdout, exit 0), not format. My `||` fallback never fired, $MTIME got garbage, and arithmetic blew up on every PENDING handoff path — regressing the existing PENDING_REVIEW/RECHECK blocks. Fix: try GNU `stat -c %Y` first (BSD errors → falls through). Add numeric case-guard so any junk in $MTIME skips the stale check entirely and falls through to HOLD (safe default). Locally macOS green. Pushing to verify Linux CI passes now.

github-actions · 2026-04-23T23:23:35Z

Claude Code Review Summary

Status: Skipped (trivial changes)

This PR only contains documentation or configuration changes. Code review skipped to save API costs.

To force a review, add the needs-review label.

Codex review of PR #227 caught: `SDLC_HANDOFF_STALE_DAYS=foo` leaked `[: foo: integer expression expected` to stderr before the normal HOLD text. A typo in the env var should silently fall back, not emit shell diagnostics. Fix: non-negative-integer case-pattern guard. Empty, non-numeric, negative, or decimal values fall back to default 14. Test: test_precompact_stale_threshold_invalid_falls_back asserts rc=2, PENDING_REVIEW in stderr, and NO `integer expression` noise. 20/20 precompact tests green. Codex score: 8/10 CERTIFIED (this addresses the sole P2 finding).

Live-fire bug 2026-04-23: at session end, /compact was blocked by the precompact-seam hook because .reviews/handoff.json is stuck at PENDING_RECHECK from the review that SHIPPED the hook itself. That original review was written BEFORE the self-heal pr_number feature (#209) existed, so there's no PR to auto-query. Handoff is stuck forever until manually flipped. Every wizard user who did a seam review before #209 has this landmine in their repo. Fix: add stale-handoff auto-expire. PENDING_* + no pr_number + mtime > N days → treat as CERTIFIED with WARN (not HOLD). Rescues stale handoffs without bypassing live reviews. P1 because it silently blocks a core UX (manual /compact) and the user hit it mid-session.

Fix UX landmine where handoff.json reviews predating the #209 pr_number self-heal feature would block manual /compact forever. Hook now heals via two paths: - (a) #209: pr_number + gh says MERGED → silent unblock - (b) #229: no pr_number + mtime > SDLC_HANDOFF_STALE_DAYS (default 14) → unblock with one-line WARN PR-linked handoffs still ignore mtime — an OPEN PR with old mtime is still a live review. Tests: 4 new in test-hooks.sh covering stale-unblocks, fresh-still-blocks, stale-with-pr_number-prefers-self-heal, threshold-override. 19/19 precompact tests green. Caught live-fire 2026-04-23 when session-end /compact blocked on a review that shipped the hook itself.

…pdate (#271) Two `cusum.sh --add` calls in weekly-update.yml (version-test + community- e2e-test jobs) appended scores to tests/e2e/score-history.txt, which is tracked by git. Result: every weekly run opened a PR whose only diff was the legacy side-file append — pure noise (Codex finding PR213-02 on auto- PR #213, 2026-04-23). The canonical history is score-history.jsonl, written via the CI Tier 2 persist step in ci.yml. The PR bodies already cite Phase A / baseline scores directly, so removing the standalone CUSUM update steps loses nothing functional. Per Codex consult this session: prefer "stop writing the .txt" over "close PRs with only that diff" (avoids GitHub API edge cases, simpler). New regression test test_weekly_update_does_not_call_legacy_cusum_add in tests/test-workflow-triggers.sh greps for `cusum.sh --add ` (legacy .txt write) and fails if reintroduced. --add-json (canonical .jsonl) is unaffected. Closes #227

Closes #227 (Codex finding PR213-02). The PR-noise source was the scan-community job appending to legacy score-history.txt via `cusum.sh --add`. #231 Phase 3c (v1.53.0) deleted that whole job, so the noise source is already gone — Test 227 in test-workflow-triggers.sh:2445 is the regression check. Cleanup pass: - Delete empty tests/e2e/score-history.txt from the repo (was a stale runtime artifact, never appended-to since Phase 3c) - Add tests/e2e/score-history.txt to .gitignore (test-cusum.sh recreates it on runs that exercise legacy --add mode) - Drop the stale .txt mention in plans/AUTO_SELF_UPDATE.md (was a Files Structure block listing both .txt and .jsonl; .jsonl is the canonical store now) cusum.sh keeps both modes (--add legacy plain-text + --add-json canonical JSONL) for users who want plain-text mode for ad-hoc CLI work. No interface change. No version bump — zero user-facing change. Tests still 169/169 + 50/50 (cusum) green.

Codex review of PR #227 caught: `SDLC_HANDOFF_STALE_DAYS=foo` leaked `[: foo: integer expression expected` to stderr before the normal HOLD text. A typo in the env var should silently fall back, not emit shell diagnostics. Fix: non-negative-integer case-pattern guard. Empty, non-numeric, negative, or decimal values fall back to default 14. Test: test_precompact_stale_threshold_invalid_falls_back asserts rc=2, PENDING_REVIEW in stderr, and NO `integer expression` noise. 20/20 precompact tests green. Codex score: 8/10 CERTIFIED (this addresses the sole P2 finding).

* release: v1.36.0 See CHANGELOG for details. Bumps: - package.json: 1.35.0 -> 1.36.0 - SDLC.md metadata + visible table - CLAUDE_CODE_SDLC_WIZARD.md example version (2 places) - skills/update/SKILL.md example output Latest line - CHANGELOG new section * chore: record E2E score [skip ci] * roadmap(#224-227): follow-ups from Codex review of auto-PRs #210, #213 Four new backlog items surfaced by Codex batch review of the auto-generated PRs (2026-04-23): From PR #210 (CC 2.1.118 auto-update analysis): - #224: runtime test for sdlc-prompt-check.sh double-firing — CC 2.1.118 shipped the fix, we have no assertion it holds. - #225: document cleanupPeriodDays minimum — TodoWrite checklists now fall under that retention sweep (CC 2.1.117). From PR #213 (community patterns, closed): - #226: P1 — weekly-update workflow reports fake 5-trial CI (re-scores one transcript N times, not N simulations). Current STABLE verdicts aren't trustworthy. - #227: stop writing legacy score-history.txt in community PRs. --------- Co-authored-by: github-actions[bot] <github-actions[bot]@users.noreply.github.com>

…pdate (#271) Two `cusum.sh --add` calls in weekly-update.yml (version-test + community- e2e-test jobs) appended scores to tests/e2e/score-history.txt, which is tracked by git. Result: every weekly run opened a PR whose only diff was the legacy side-file append — pure noise (Codex finding PR213-02 on auto- PR #213, 2026-04-23). The canonical history is score-history.jsonl, written via the CI Tier 2 persist step in ci.yml. The PR bodies already cite Phase A / baseline scores directly, so removing the standalone CUSUM update steps loses nothing functional. Per Codex consult this session: prefer "stop writing the .txt" over "close PRs with only that diff" (avoids GitHub API edge cases, simpler). New regression test test_weekly_update_does_not_call_legacy_cusum_add in tests/test-workflow-triggers.sh greps for `cusum.sh --add ` (legacy .txt write) and fails if reintroduced. --add-json (canonical .jsonl) is unaffected. Closes #227

Closes #227 (Codex finding PR213-02). The PR-noise source was the scan-community job appending to legacy score-history.txt via `cusum.sh --add`. #231 Phase 3c (v1.53.0) deleted that whole job, so the noise source is already gone — Test 227 in test-workflow-triggers.sh:2445 is the regression check. Cleanup pass: - Delete empty tests/e2e/score-history.txt from the repo (was a stale runtime artifact, never appended-to since Phase 3c) - Add tests/e2e/score-history.txt to .gitignore (test-cusum.sh recreates it on runs that exercise legacy --add mode) - Drop the stale .txt mention in plans/AUTO_SELF_UPDATE.md (was a Files Structure block listing both .txt and .jsonl; .jsonl is the canonical store now) cusum.sh keeps both modes (--add legacy plain-text + --add-json canonical JSONL) for users who want plain-text mode for ad-hoc CLI work. No interface change. No version bump — zero user-facing change. Tests still 169/169 + 50/50 (cusum) green.

BaseInfinity merged commit 07807b4 into main Apr 24, 2026
4 of 5 checks passed

BaseInfinity added 2 commits April 23, 2026 22:23

This was referenced Apr 27, 2026

[enhancement] PreCompact hook should auto-CERTIFY when fixes_applied SHAs are already in HEAD's ancestry #257

Closed

fix(ci): close #227 — stop legacy score-history.txt churn in weekly-update #271

Merged

BaseInfinity mentioned this pull request Apr 29, 2026

chore: close ROADMAP #227 — delete dead score-history.txt #288

Merged

2 tasks

BaseInfinity mentioned this pull request May 13, 2026

release: v1.43.0 quick-wins bundle (#226, #227, #338, #219, #337, #339) #340

Closed

17 tasks

Navigation Menu

Uh oh!

roadmap(#229): P1 precompact hook stale-handoff auto-expire - #227

roadmap(#229): P1 precompact hook stale-handoff auto-expire#227
BaseInfinity merged 4 commits into
mainfrom
roadmap-229-stale-handoff-expire

BaseInfinity commented Apr 23, 2026 •

edited

Loading

Uh oh!

github-actions Bot commented Apr 23, 2026

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

Uh oh!

Conversation

BaseInfinity commented Apr 23, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Summary

Fix (shipped in 58d1a86 / d761028)

Tests

Rationale for 14-day default

Shepherd loop

Uh oh!

github-actions Bot commented Apr 23, 2026

Claude Code Review Summary

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

BaseInfinity commented Apr 23, 2026 •

edited

Loading

Fix (shipped in 58d1a86 / `d761028`)