Filed by the domain:cli execution PM seat #6024 (session_01QCdUBjM47SxioST9z5Zwdf). ⛔ Filed bare — ⛔ no domain:*, ⛔ no priority:*, ⛔ no type: triage grades and routes. ⛔ Not a claim. ⛔ Not a ruling.
⏱️ Every rule line below was read first-hand on origin/main eec56c37df; every comment was read first-hand from the REST API. ⛔ Nothing is quoted from a report or from recall.
The collision
references/contract-review.md defines a clause-② review of record and what discriminates one:
:27 复核记录 = 一条评论落 PR 或卡,席内与子代理同形;散文、dev 自评、os-dev-report 恒不算。
:28 同形 = ## Contract review 题头、所审 head sha 独占码段、①②③ 逐项、独立性对、PASS/FAIL。
:29 同形含首行 Served-tier::值写常量名 CONTRACT_REVIEW_TIER;无此行不成裁决,模板见 --template。
That is a complete rule for a reader who opens the comment. It says nothing about what a comment may be titled, and the discriminator it names — Served-tier: on the first line — sits below the heading.
So a seat may correctly write a comment headed ## Contract review of record: **FAIL** that is not a record: a pointer at the real one. Two such comments exist on card #18431, both written by this seat:
| comment |
heading |
what it actually is |
5747212041 |
## Contract review of record: **FAIL** — patch round opening… |
names record 5747199582 on PR #18962, its head, its verdict |
5747574143 |
## Contract review of record round 2: **FAIL** — … |
names record 5747563471, its head, its verdict |
Neither carries a Served-tier: line. Neither carries a **VERDICT:** line. Both are correct as written — :27 already excludes 散文 from counting, so nothing machine-read was misled. The defect is entirely at skim depth, and skim depth is where a seat re-reads its own thread months later.
What the collision cost
The domain:cli seat's patrol prompt carried this standing item across many rounds, marked 「NOT YET FILED, and it is this seat's own defect」:
PR #18962 / card #18431 (domain:cli, Clause-②: yes) was cleared against in-seat review-of-record comments, but e15870f7a0 landed the lane rule before that — so that merged pair did what this lane may not. Record it on #18431; ⛔ do not quietly reuse it as precedent.
It was measured for the first time this round and withdrawn — every limb of it is false. The four real records live on the PR thread (:27 sanctions 「PR 或卡」), all four open "Isolated at-tier review of record", all four carry the full :28/:29 shape, and the clearing PASS 5749558880 names Head-sha: b8e08ec2fb6b0af4818b839766be95672d4897ed — the exact sha PR #18962 merged at. The withdrawal, with the full measurement, is on #18431.
⭐ The charge was built from the headings of 5747212041 and 5747574143. Nothing else on either thread reads as an in-seat review, because nothing else is one.
⇒ the cost was not a wrong merge. It was a false non-conformance charge standing against a conforming clearance, carried forward round after round in the one artefact a successor seat inherits verbatim. A successor acting on it would have refused a precedent that is sound.
What this card asks
A naming discipline, ⛔ not a gate.
- The heading
## Contract review is reserved for the record. A comment that points at a record, relays one, or opens a patch round on one is headed something else — and names the record's comment id in its own opening line, which both of the two above already do.
- One line beside
:27 saying so, so the exclusion covers the title and not only what counts.
⛔ Deliberately not asked for: a new gate, a new --pair row, or a ratchet. Ruling batch #208 item 1 (record 5755284987, maintainer accepted) is explicit — R6: 「hard gates only for definite questions; guessing instruments report-only; a new gate / row / ratchet needs the maintainer's sentence」. "Is this heading a record or a pointer?" is answerable by opening the comment, so no reader is guessing; the failure was a seat that did not open it. Text is the proportionate remedy.
⚠️ One measurement trap, recorded because this seat fell into it first
Opening the investigation, this seat measured 5747212041 and 5747574143 against the record template and found they carry neither Served-tier: nor **VERDICT:** — and briefly took that for a third and sharper finding ("the records are malformed"). It is not a finding at all: they are not records, so the template does not apply to them. ⇒ ⭐ measuring an artefact against a contract it was never under produces a clean-looking failure, and the heading is exactly what invites the wrong contract to be picked.
Generated by Claude Code
Filed by the
domain:cliexecution PM seat #6024 (session_01QCdUBjM47SxioST9z5Zwdf). ⛔ Filed bare — ⛔ nodomain:*, ⛔ nopriority:*, ⛔ no type: triage grades and routes. ⛔ Not a claim. ⛔ Not a ruling.⏱️ Every rule line below was read first-hand on
origin/maineec56c37df; every comment was read first-hand from the REST API. ⛔ Nothing is quoted from a report or from recall.The collision
references/contract-review.mddefines a clause-② review of record and what discriminates one:That is a complete rule for a reader who opens the comment. It says nothing about what a comment may be titled, and the discriminator it names —
Served-tier:on the first line — sits below the heading.So a seat may correctly write a comment headed
## Contract review of record: **FAIL**that is not a record: a pointer at the real one. Two such comments exist on card #18431, both written by this seat:5747212041## Contract review of record: **FAIL** — patch round opening…5747199582on PR #18962, its head, its verdict5747574143## Contract review of record round 2: **FAIL** — …5747563471, its head, its verdictNeither carries a
Served-tier:line. Neither carries a**VERDICT:**line. Both are correct as written —:27already excludes 散文 from counting, so nothing machine-read was misled. The defect is entirely at skim depth, and skim depth is where a seat re-reads its own thread months later.What the collision cost
The
domain:cliseat's patrol prompt carried this standing item across many rounds, marked 「NOT YET FILED, and it is this seat's own defect」:It was measured for the first time this round and withdrawn — every limb of it is false. The four real records live on the PR thread (
:27sanctions 「PR 或卡」), all four open "Isolated at-tier review of record", all four carry the full:28/:29shape, and the clearing PASS5749558880namesHead-sha: b8e08ec2fb6b0af4818b839766be95672d4897ed— the exact sha PR #18962 merged at. The withdrawal, with the full measurement, is on #18431.⭐ The charge was built from the headings of
5747212041and5747574143. Nothing else on either thread reads as an in-seat review, because nothing else is one.⇒ the cost was not a wrong merge. It was a false non-conformance charge standing against a conforming clearance, carried forward round after round in the one artefact a successor seat inherits verbatim. A successor acting on it would have refused a precedent that is sound.
What this card asks
A naming discipline, ⛔ not a gate.
## Contract reviewis reserved for the record. A comment that points at a record, relays one, or opens a patch round on one is headed something else — and names the record's comment id in its own opening line, which both of the two above already do.:27saying so, so the exclusion covers the title and not only what counts.⛔ Deliberately not asked for: a new gate, a new
--pairrow, or a ratchet. Ruling batch #208 item 1 (record5755284987, maintainer accepted) is explicit — R6: 「hard gates only for definite questions; guessing instruments report-only; a new gate / row / ratchet needs the maintainer's sentence」. "Is this heading a record or a pointer?" is answerable by opening the comment, so no reader is guessing; the failure was a seat that did not open it. Text is the proportionate remedy.Opening the investigation, this seat measured
5747212041and5747574143against the record template and found they carry neitherServed-tier:nor**VERDICT:**— and briefly took that for a third and sharper finding ("the records are malformed"). It is not a finding at all: they are not records, so the template does not apply to them. ⇒ ⭐ measuring an artefact against a contract it was never under produces a clean-looking failure, and the heading is exactly what invites the wrong contract to be picked.Generated by Claude Code