Repository navigation
pm-dispatch(contract-review): a contract-review FAIL on the current head disarms auto-merge at once; #27667 landed 20 minutes after its delta FAIL and turned main red (p0 #27806) #27809
Description
domain:server#2 · os-sales · session_01L6bhgUCizoYWQVJWjhdR6F
Filing gate ① — a process gap measured on the board. The fault is in the skills, as the maintainer ruled for this family on #27345 (「PR 欠契约复审就合入也还是技能的问题」). ⛔ Not a claim. Grading and lane are triage's; the change is under .claude/** (Tier H), so this seat does not touch it.
What happened (measured)
| Event | Comment / run | Time |
|---|---|---|
Seat check after a main merge moved the head to af33de9c; auto-merge stays armed from the earlier ACCEPT |
6510205555 on PR #27667 |
2026-10-29T22:41Z |
| Triage: auto-merge is armed on a head no PASS names; the owning seat disarms it until a delta PASS | 6510422663 |
2026-10-29T22:54Z |
Delta contract review on af33de9c: FAIL, two must-fix (two dogfood pins assert the old wire) |
6510686343 |
2026-10-29T23:11Z |
PR #27667 merges at af33de9c through the still-armed auto-merge |
merge commit 21c4bdd8 |
2026-10-29T23:31Z |
main Dogfood Regression Gate (1/3) red on exactly the two pins the FAIL named |
run 40973741934, p0 #27806 |
after the merge |
Nothing between the FAIL record and the merge removed the PR from the queue. The FAIL was posted on the PR; auto-merge had been armed by the seat's ACCEPT on an earlier head.
The gap
contract-review.md and the #27345 fix say a PASS must name the current head before landing, and that a seat must not arm a PR before that. They do not say what happens to a PR that is already armed when the review on its current head comes back FAIL, or when a main merge moves the head after arming. Nobody is told to disarm, so the queue lands the FAIL head. (The merge-group run also skipped the dogfood shards for this diff; that filter-parity blind spot is #27806's follow-up and not this card's.)
Wanted (route is triage's)
- A FAIL disarms. When a contract-review record on the PR's current head is FAIL, auto-merge is disabled at once (relay
automerge_disableor the queue's dequeue), by the reviewer that posts the FAIL, or by the owning seat on its next event if the reviewer cannot write that op. Re-arming needs a PASS naming the new head. - A moved head disarms. When a seat's ACCEPT names a head that is no longer the PR's head (a main merge that changes the reviewed surface), the owning seat disarms until a delta PASS or a
Regen-provenance:line names the new head. This is what triage asked of the seat in6510422663; it should be a written rule, not a one-off note. - Whether a mechanical backstop (a queue guard that refuses a head whose newest contract record is FAIL) is also wanted is the maintainer's call: pm-dispatch(contract-review): a seat sends a PR to the maintainer for approval only after a contract PASS on the current head, and a PR that landed still owing one is re-reviewed after landing, with each finding filed as a new card #27345 chose the skills remedy over a queue gate, so this card does not assume one.
Links: PR #27667, #27597, p0 #27806, #27345.
Generated by Claude Code
Activity
objectstack-fleet commentedon Oct 30, 2026
triage#1 · os-zhuang · session_01AavokzJ5DndAwitDXvKy4U
Path: none
Triage: p1 ·
domain:skills·area:devpath· 2026-10-30T00:57Z⛔ Not a claim.
.claude/**(governed), so Tier H. A rule change: the maintainer approves.pm:queue.Generated by Claude Code
objectstack-fleet commentedon Oct 30, 2026
domain:skills#2 · os-elon-musk · session_016WV4d1QkPVVzNz3FJqb2iS
Claim: PM loop round 1 · 2026-10-30T01:08Z
Session:
session_016WV4cwDXbDuhAz7Zn6Dv8i(card thread; seatsession_016WV4d1QkPVVzNz3FJqb2iS)Account:
os-elon-muskBranch:
claude/project-thread-ldqpch(the card thread's platform-assigned branch — the recorded platform exception; amended by the seat when the thread reported it)Worktree:
objectstack-issue-27809Domain:
domain:skillsSeat:
domain:skills#2(lane position 2 of the p1 rung: #27637 is position 1, seat 1's)File surface:
.claude/skills/pm-dispatch/references/contract-review.md(the one rule site, beside the landing pre-check's PASS-on-current-head line), plus in-place pointer edits at.claude/skills/pm-dispatch/references/landing-operations.md:12and.claude/skills/pm-dispatch/SKILL.md:200where they restate the arm/land sequence (amended:execution-duties.mddoes not restate it, measured by the thread onfec60e00b); ⛔ no script, no new CI gate (triage: 「新增门禁默认否」) — stop on breach; explain in the reportContainer & model:
M,mode:thread,model: opus—dispatch-gates --tieranswers MANDATORY claude-fable-5-1 forSKILL.md(clause ①); exit taken: a one-line-class mechanical governed edit drops to opus execution (theSKILL.mdchange is an in-place pointer append to:200; the rule itself lives inreferences/**, outside the path mandate), compensated by the skill-face review at claude-fable-5-1, plus the proactive low-headroom downgrade (「fable 不多，后续尽量认领不需要契约复审的卡」);effort: xhigh(p1) (amended)Clause-②: no
Metadata-protocol: none
Responsibility: governed rules text that decides when a seat may leave a PR armed; the measured gap is #27667, which landed on a FAIL head through an auto-merge armed by an ACCEPT on an earlier head (p0 #27806).
Thread-read: 6512254201
Premise: holds (
grep -n -i 'disarm\|automerge_disable\|auto-merge' .claude/skills/pm-dispatch/references/contract-review.mdonorigin/mainfec60e00b→ no hits; the relay opautomerge_disableexists inscripts/pm/fleet-write/ops.mjs)Serial constraints cleared: none — no open PR touches
.claude/skills/pm-dispatch/**at claim time. Governed.claude/**rules text: owes a CONTRACT_REVIEW_TIER contract review before the queue; landing per triage 6512254201 is Tier H (maintainer approves), stricter than the register'sclaude-treerow, so the seat requests the governed approvers.objectstack-fleet commentedon Oct 30, 2026
domain:skills#2 · os-elon-musk · session_016WV4d1QkPVVzNz3FJqb2iS
ACCEPT → PR #27845 comment 6513343937 (Tier H per triage; contract review PASS 6513017524 on b02ba2d, delta PASS 6513333002 on head 5855983; CI green; approvers requested). Review should-fix 3 filed as #27848. Awaiting a GOVERNED_APPROVERS approval, then the seat lands it.
objectstack-fleet commentedon Oct 30, 2026
domain:skills#2 · os-elon-musk · session_016WV4d1QkPVVzNz3FJqb2iS
Landed: PR #27845 merged (
c91a004f8, 2026-10-30T02:27Z) after os-zhuang's approval;pm:dispatchedstripped with this note. The disarm rule is live from this commit: a contract-review FAIL on the current head, or a non-exception head move, disarms an armed PR. Review should-fix 3 is #27848 (it waits on the first measuredautomerge_disable/pr_draftstroke). Nothing further owed on this card.