[ruled C, batch devx-cleanup] Mint a governance label that carries Clause-②: no, so the majority arm of the contract declaration stops living in free text that any body edit erases #19773

Provenance — filed by the objectstack triage seat (session_01Tw7jnJinGHvoGSi8aFkhPJ) as the domain:skills half of ruling C on objectstack-ai/objectstack#16949. The maintainer ruled letter C on 2026-09-23, in the devx-cleanup batch. ⛔ The measurement history is on #16949 and is not duplicated here.

⚠️ This card is HALF of a pair. The other half — teaching scripts/check-changeset-no-major.mjs to read this label as a no carrier — stays on #16949 (domain:devx), which is pm:blocked behind this card: the parser cannot be written until this label has a name and a mint.

What was ruled

AGENTS.md:1066-1068 requires every PR to declare Clause-②: yes|no. Read the parser, scripts/check-changeset-no-major.mjs:1226-1228:

:1226  if (carrier || (line?.kind === 'declared' && line.value === 'yes')) return { value: 'yes', … };
:1227  if (line?.kind === 'declared' && line.value === 'no')                return { value: 'no',  … };
:1228  return { value: null, … };

⇒ yes has two carriers — a durable label, or the body line. no has exactly one: the body line. There is no label that expresses no at all.

⭐ So the loss measured on #16949 was not bad luck — no is the only arm that CAN be lost, and it is the majority arm: 85 of the last 120 merged PRs declare no, and 63 of those 85 carry no changeset file at all (which is why the cheaper "put it in the changeset" route was refused — three quarters of them have no file to write it in).

Letter C is: give no a governance label of its own, at parity with needs:contract-review on the yes side.

What this card delivers

1. A label name, in the PM label vocabulary, with its description — the no counterpart of needs:contract-review.
2. Its mint, through whatever this repo's label vocabulary actually uses.
3. ⛔ Not the parser change. That is [finding] the Clause-②: declaration lives in the PR body, so any unrelated body edit destroys it silently — measured: three runs of one check on one head SHA #16949.

⚠️ Two premises this seat did NOT measure — settle them before writing anything

Both are copied verbatim from #16949's 置信缺口, because they are this half's, not the parser half's:

• ⛔ untested: whether the label can be minted by ensure-pm-labels.sh without touching a governed surface. If minting it requires editing a governed file, that changes how this lands (draft PR + human merge), ⛔ not whether it lands.
• ⛔ untested: whether the label vocabulary has a cap or a ratchet that this new entry would break (check:pm-label-desc-cap exists and is named in [finding] 32 gates declare only their own artifacts and no scan surface, so path derivation scores them silent for every card — per-gate producer-side remedy, routed by lane #15082's roster list).

Why it is here and not in domain:devx

Triage's own prior ruling on #16949 (5599637112) wrote this split in advance, verbatim: 「⛔ 但若答案是『给 no 也发一个标签』,那是治理标签词表的变更 …… 那一半属 skills …… 若认领方选了新增标签这条路,回报本席同笔改判 skills,⛔ 不要自行跨过去。」 ⇒ letter C was chosen; this is that re-lane, executed by triage rather than by a claimant crossing lanes.

Filing gate

Class ① — a reproducible defect with a named landing surface, with a maintainer ruling already attached. Dedupe: #16949 and its thread are the only prior record; the label vocabulary itself carries no Clause-② entry today (that absence is the finding).

Generated by Claude Code