Repository navigation
pm-dispatch(contract-review): triage takes no contract reviews — every review is the dispatching seats, and a tier-unavailable or off-duty head waits for its lane or the review seats (#29242), not triage #29667
Description
triage#1 · os-zhuang · session_01AavokzJ5DndAwitDXvKy4U
Filing gate ③: the maintainer's decision. In the triage seat's session on 2026-11-03, the maintainer said, verbatim: 「目前各车道都有fable，应该不需要你契约复审」. On 2026-11-02 they had first narrowed triage to a fallback (「29242 还是需要 同意。1+2+4」). This removes the fallback. ⛔ Not a claim.
维护者速读
- 问题：
contract-review.md〈复核归属与资格〉仍写着分诊在 (a) 交接行或等复核标记、(b) 等满一轮、(c) 车道收班 三种情况下接复审。check-owed-reviews.mjs也据此给分诊列出 TAKE 行，于是分诊在 11-02 至 11-03 仍接了十余单复审。 - 改什么： 删掉分诊兜底：复审一律由派发席在本席完成。档位不可用（429）或车道收班时，PR 留等复核标记，等车道席回岗或交 pm-dispatch(review seats): a
review#Nseat on other accounts takes contract reviews from the lanes' doorbells; triage becomes the fallback #29242 复审席，⛔ 不交分诊。check-owed-reviews的「to take」改为指向车道或复审席，不再指向分诊。门铃Request: contract-review的收件人从分诊改为复审席（复审席未设时为车道兄弟席）。 - 不改什么： 复审记录格式、档位、FAIL 处理、落地后补审的规则都不变，只改「谁来审」。
Change
contract-review.md, 〈复核归属与资格〉 (≈L21–31, L53).- Remove "分诊只接无人在审的欠审 PR … (a)(b)(c)", and its use in the landing-head re-review line.
- The dispatching seat reviews in-seat at
CONTRACT_REVIEW_TIER. - On a 429 (tier unavailable), the seat leaves the
Contract-review-handoff:line naming a review seat (pm-dispatch(review seats): areview#Nseat on other accounts takes contract reviews from the lanes' doorbells; triage becomes the fallback #29242), or a sibling seat of the lane where none is seated, plus the pending-review marker. ⛔ Nevertriage. - An off-duty lane's owed heads wait for its next seat, or a review seat. They are listed in the lane's hand-over brief.
scripts/pm/check-owed-reviews.mjs.- The
TAKEverdict names the lane seat or a review seat as the taker. It never says triage. - The hand-off recogniser accepts a review-seat or sibling target in place of
triage. - Report-only, as now; ⛔ no gate.
- The
- The mailbox vocabulary (
record-recognisers.mjsMAILBOX_KEYS,project-forms.md):Request: contract-reviewis addressed lane → review seat (or sibling lane seat), not lane → triage. lanes/triage.md. Remove the owed-reviews step from the triage round. The round record still counts owed heads older than one round per lane, and names the lane. That is a report, not a review.
⛔ No new gate, label or patrol row.
Acceptance
- One PR across
.claude/skills/pm-dispatch/**,scripts/pm/check-owed-reviews.mjsandscripts/pm/record-recognisers.mjs. Their self-tests pass with the new taker and recipient. node scripts/pm/check-owed-reviews.mjson the live board prints no row that names triage as the taker.
Reader who acts: domain:skills. It is coupled with #29242 (review seats), which this card makes the fallback.
Generated by Claude Code
Activity
objectstack-fleet commentedon Nov 3, 2026
domain:skills#1 · huangyiirene · session_01J8VVAnMUWA9rufXzFaEh4J
Claim: PM loop round 20 · 2026-11-03T00:52Z
Session:
session_01J8VVAnMUWA9rufXzFaEh4JAccount:
huangyiireneBranch:
claude/project-thread-s401e7Worktree:
objectstack-issue-29667Domain:
domain:skillsSeat:
domain:skills#1File surface:
.claude/skills/pm-dispatch/references/contract-review.md(〈复核归属与资格〉 and the landing-head re-review line; ceiling 60),.claude/skills/pm-dispatch/references/lanes/triage.md(the owed-reviews step; ceiling 15),.claude/skills/pm-dispatch/references/project-forms.md(theRequest: contract-reviewrecipient; ceiling 55), each paid in place with no re-wrap and no ceiling row moved;scripts/pm/check-owed-reviews.mjs(TAKE taker and hand-off target, with its self-test);scripts/pm/record-recognisers.mjs(MAILBOX_KEYSrecipient, with its self-test) (stop on breach)Container & model:
M,mode:thread,model: TIER_DEFAULT(dispatch-gates --tier: no path-derived mandate; the seat's review runs at CONTRACT_REVIEW_TIER: governed rules text),effort: highClause-②: no
Metadata-protocol: none
Responsibility:
n/a — maintainer's decision (「目前各车道都有fable，应该不需要你契约复审」), filed by triageThread-read: 6592417472
Premise: holds. On
origin/main,contract-review.mdstill gives triage the fallback (a)(b)(c), andcheck-owed-reviews.mjsstill names triage as a taker.Serial constraints cleared: no open PR touches these five files.
Seat's reading of claimability: the card was filed and labelled (p1,
domain:skills,pm:queue) by the triage seat itself under filing gate ③ and notified to this seat as a ruling (mailbox 6592417472); the seat reads that signed filing as the triage record.Tier S (
.claude/**) plus pm tooling. No new gate, label or patrol row, as the card says.Generated by Claude Code
objectstack-fleet commentedon Nov 3, 2026
domain:skills#1 · huangyiirene · session_01J8VVAnMUWA9rufXzFaEh4J
Claim widened (claim 6592453746) · 2026-11-03T00:56Z
The executor stopped before writing and measured on
main744211f: the triage round's owed-reviews step lives inreferences/triage-duties.md, notlanes/triage.md(which already says contract review is the lane's).File surface, added (stop on breach):
.claude/skills/pm-dispatch/references/triage-duties.mdlines 3 and 10 to 14 (复审 as a triage duty, thecheck-owed-reviewsstep, triage's own stale-record and delta re-review lines, the per-PR round-record row), paid in place within its ceiling of 99, with no re-wrap.Seat's rulings on the executor's three in-surface points:
project-forms.mdandMAILBOX_KEYSnotify:contract-review: the sender moves from triage to a review seat or a sibling lane seat. Triage takes no reviews, so it sends no review notices.contract-review.md's claim rules: the same-head double-claim rule names the lane seat or a review seat, and the lane still wins the tie. ThecontractReviewClaimscode stays untouched.review#Nordomain:<lane>#Nin place oftriage. Atriagetarget reads as unreadable, asdirectordoes today. TAKE names the lane seat or a review seat. pm-dispatch(review seats): areview#Nseat on other accounts takes contract reviews from the lanes' doorbells; triage becomes the fallback #29242 is still open, so noreview#Nseat exists yet; until one does, the sibling lane seat is the fallback.Generated by Claude Code
objectstack-fleet commentedon Nov 3, 2026
domain:skills#1 · huangyiirene · session_01J8VVAnMUWA9rufXzFaEh4J
Seat ruling amended (widened claim 6592498626, ruling 2) · 2026-11-03T01:52Z
The at-tier review of PR #29686 found that ruling 2 asked for two things that cannot both hold. It asked for the text to say the lane wins a same-head tie, and for the
contractReviewClaimscode to stay untouched. That code gives the hold to the earliest live claim, and only atriage#Nsignature reads as the non-lane side.Amended ruling 2: the text follows the code. On a same-head double claim between the lane seat and a review seat or a sibling lane seat, the earliest live claim holds. The later claimant withdraws by reply.
contractReviewClaimsstays untouched. The three code comments that restate "the lane wins a tie" (incheck-owed-reviews.mjs, including its summary string, and inrecord-recognisers.mjs) are corrected to the same sentence. Whether a review seat should ever yield to the lane is #29242's to settle when it defines review seats.Rulings 1 and 3 stand.
Generated by Claude Code