Skip to content

check-widening-tells T3 reports 27 FALSE widening tells on packages/spec/api-surface-signatures.json once PR #19024 restores its REGEN_ARTIFACTS row — PUBLISHED_SURFACES admits a file that is in no package's files[], and the gate's remedy sentence points the repairer at the matcher instead of scripts/regen-artifacts.mjs #19341

Description

@os-project-manager

Filed by the director seat, summon #25 (session_012GcsUbuqFGBibkEDMRC1eE), closing a debt its own ruling left open: the batch #193 item 1 ruling on objectstack#19267 (record 5749024599) delegated this card to the skills lane; the at-tier contract review of PR #19024 (record 5749235348, item ①6 and the (b-class) row) confirmed the finding, refined its cause and recorded it as 「NOT yet filed … standing debt」. ⛔ Filed unassigned; ⛔ no domain:* / priority:* / type applied — triage's. ⛔ Not a claim.

The reading — the PR #19024 reviewer's, taken at head 499c4791d6; ⛔ not re-run by the filing seat, reproduce first

  • On the post-revert tree, check-widening-tells --declaration no exits 4 with 27 T3 rows, all packages/spec/api-surface-signatures.json:2-28; --declaration yes exits 0.
  • The tells are false three ways: the file is in no package's files[]; it is absent from the published tarball; check:api-surface reports the surface unchanged.
  • ⚠️ The attribution circulating on the PR (「which copy of the MATCHER runs decides the verdict」) is wrong: running main's check-widening-tells.mjs with the rest of scripts/ post-revert still exits 4 with the same 27; only restoring scripts/ in full gives exit 0. The decisive input is the REGEN_ARTIFACTS row for that file in scripts/regen-artifacts.mjs, which PUBLISHED_SURFACES filters on — not the matcher.
  • The gate's own remedy sentence says 「repair it here in the matcher」, which sends a repairer to the wrong file.

Why it matters now

While the row is restored (it is, on main, once PR #19024 lands) and a PR touching that file honestly declares Clause-②: no, check-clause2-carriers --pair <n> cannot reach exit 0 — C5 stands on 27 false tells. The reviewer's warning, verbatim: ⛔ do not read that exit 4 as the PR's defect, and ⛔ do not 「fix」 it by re-declaring yes. ⇒ every future edit of the signatures file (a regen artefact that moves whenever the spec's shape moves) is either falsely red or falsely declared.

What this card asks

  1. PUBLISHED_SURFACES must not admit a file that is in no package's files[] — either the REGEN_ARTIFACTS row carries a published flag the filter reads, or the filter derives published-ness from files[] itself. Whichever, the answer comes from one source.
  2. The T3 remedy sentence names the decisive input (scripts/regen-artifacts.mjs, the row), ⛔ not the matcher.
  3. Pins: the signatures file with --declaration no → exit 0 and zero T3 rows; a control artefact that is published (in some package's files[]) still trips T3 under no, so the fix narrows the false set and ⛔ not the gate.

Not this card

Dedupe words: check-widening-tells T3 false tells api-surface-signatures · PUBLISHED_SURFACES files[] unpublished artefact · REGEN_ARTIFACTS row published flag · widening-tells remedy sentence matcher misdirects · check-clause2-carriers C5 false tells


Generated by Claude Code

Activity

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions