Skip to content

Re-verify a cleared stop condition: the reviewer line and the brief-construction note - #11334

Merged
hotlong merged 1 commit into
mainfrom
claude/issue-11251-premise-reverify-rule
Aug 23, 2026
Merged

Re-verify a cleared stop condition: the reviewer line and the brief-construction note#11334
hotlong merged 1 commit into
mainfrom
claude/issue-11251-premise-reverify-rule

Conversation

@claude

@claudeclaudeBot commented Aug 23, 2026

Copy link
Copy Markdown
Contributor

Fixes#11251

A stop condition whose only enforcement is the stopped party's self-report is not
enforced. A dev that reasons past one produces an artifact indistinguishable from one
that honoured it: an on-scope diff, green gates, and prose asserting the check passed.
This PR lands the graded scope — the PM-side rule — as exactly two lines on the
references face. SKILL.md is not touched.

What landed

references/review-checklist.md — one new checklist item, placed directly after the
falsification item (dev 验证过 issue 的前提吗?) so the two read as a pair: falsification
is good work, and a cleared stop condition is re-verified by the PM on origin/main,
never accepted from the dev's prose alone. Its verdict criterion is the reported evidence
surface (what was searched, what was found) plus the PM's own re-check; the incident is
compressed to the house one-line form.

references/dispatch-runbook.md — one brief-construction note in ## 派发词构造细则,
directly after the three-partition clause it is a sibling of: a premise/stop-condition
check in a brief must ask the dev to report evidence, not a verdict — the PM
adjudicates the fork.

Out of scope by the grading, and deliberately not attempted here: the separately-reported
artifact as a hard gate, and forbidding a same-turn commit past a stop condition. Both
change the dispatch state machine and are recorded as inputs to the whole-text
restructuring round rather than decided piecemeal.

How the two lines were paid for

Both files sit at their line-ratchet ceilings with zero headroom. Payment is in-place
content compression — no ceiling raise, no re-wrap as payment, and no dated ruling or
verbatim maintainer quote deleted.

filebeforeaddedcompressedafter / ceiling
references/review-checklist.md82+4−482 / 82
references/dispatch-runbook.md243+2−2243 / 243

What was compressed, and what was removed from each:

  • review-checklist.md — the local-gate-union item drops its restated rationale, which
    the preceding pin-to-head item and the dev-side contract already carry (−2); the
    cross-the-boundary item drops its closing aphorism (−1); the scope-check item folds its
    changeset-route tail (−1). Every operative branch survives: the contract requirement,
    the head.sha comparison, both-directions rejection, the no-HEAD-quoted handling, and
    both repos' tests/docs-only routes.
  • dispatch-runbook.md — the three-state recovery item folds its two parenthetical
    incident retellings into the house one-line form, per that file's own 行文纪律 (−1);
    the subagent fast-forward paragraph drops the redundant alternative spelling of the
    command it already gives (−1).

Provenance preservation was checked mechanically rather than asserted: both files carry an
identical date inventory before and after, and the runbook's bracketed-quote inventory is
byte-identical (same set, same count). The one bracketed span that leaves
review-checklist.md is the scare-quoted phrase 「本地全绿」 inside the compressed
sentence — not a ruling, not a maintainer quote.

Verification

Gate families derived on the actual change set with
node scripts/pm/dispatch-gates.mjs --repo objectstack-ai/objectstack (2 paths, committed
2 / working tree 0 / untracked 0), re-derived after the final commit — same 8 families, no
new one. The union below was run after the final commit, at HEAD 562cb567, exit codes
captured before any pipe:

  • pnpm check:pm-skill-ratchet — the two verdict lines this PR is accountable for:
    • ✓ check-skill-line-ratchet: .claude/skills/pm-dispatch/references/review-checklist.md is 82 lines (ceiling 82; headroom 0).
    • ✓ check-skill-line-ratchet: .claude/skills/pm-dispatch/references/dispatch-runbook.md is 243 lines (ceiling 243; headroom 0).
  • pnpm check:pm-skill-id-lint✓ check-skill-id-lint: 19 file(s) clean (pattern /#[0-9]{3,}/g).
  • pnpm check:doc-authoring✓ doc authoring guard: 389 files clean — no bare metadata literals.
  • pnpm --filter @objectstack/lint run check:doc-formula-expressions✓ check:doc-formula-expressions: 22 record-scoped formula example(s) across 418 files / 1447 TS blocks judged clean by @objectstack/formula. (needed its dependency closure built first)
  • pnpm check:pm-governed-merges✓ check-governed-merges --self-test: 129 assertions
  • pnpm check:required-contexts and node scripts/check-required-contexts.mjs✓ check-required-contexts: 6 required context name(s) pinned across 2 workflow(s)
  • pnpm check:skill-frame-sync✓ check-skill-frame-sync: 4 copies of the decision frame are structurally isomorphic across 3 files

Control-byte self-scan over both edited files: no match (escaped spellings only).

No changeset — this is instruction-surface text under .claude/, publishing nothing;
skip-changeset per this repo's convention.

Governance

This PR edits .claude/**, a governed surface. It opens draft, it is for human merge
only
, and no auto-merge is or should be enabled on it.

Generated by Claude Code


Generated by Claude Code

…onstruction note
A stop condition whose only enforcement is the stopped party's self-report is
not enforced. A dev that reasons past one produces an artifact indistinguishable
from one that honoured it: on-scope diff, green gates, and prose asserting the
check passed. Two lines close the reviewer-side half.
review-checklist.md — a new item paired with the falsification item directly
above it, so the two read together: falsification is good work, AND a cleared
stop condition is re-verified by the PM on origin/main, never accepted from the
dev's prose alone. The verdict is the reported evidence surface (what was
searched, what was found) plus the PM's own re-check.
dispatch-runbook.md — a brief-construction note in the dispatch-wording
section: a premise/stop-condition check must ask the dev to report EVIDENCE,
not a verdict; the PM adjudicates the fork.
Both files sit at their line-ratchet ceilings with zero headroom, so both
additions are paid in place by content compression, with no ceiling raise:
review-checklist.md (82/82) — the local-gate-union item drops its restated
rationale (-2), the must-cross-the-boundary item drops its closing aphorism
(-1), and the scope-check item folds its changeset-route tail (-1).
dispatch-runbook.md (243/243) — the three-state recovery item folds its two
parenthetical incident retellings into the house one-line form (-1), and the
subagent fast-forward paragraph drops the redundant alternative spelling of
the command it already gives (-1).
No dated ruling and no verbatim maintainer quote was touched: both files carry
the same date inventory before and after, and the runbook's bracketed-quote
inventory is byte-identical.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01RMTpSRF5CjMmQBFfPtPCwJ
@github-actionsgithub-actionsBot added size/s documentation Improvements or additions to documentation labels Aug 23, 2026
@os-zhuangos-zhuang added the skip-changeset PR has no user-facing published change; bypasses the changeset gate label Aug 23, 2026 — with Claude
@os-zhuang
os-zhuang self-requested a review August 23, 2026 12:09
@hotlong
hotlong marked this pull request as ready for review August 23, 2026 12:26
@hotlong
hotlong added this pull request to the merge queueAug 23, 2026
Merged via the queue into main with commit b0fe0baAug 23, 2026
27 checks passed
@hotlong
hotlong deleted the claude/issue-11251-premise-reverify-rule branch August 23, 2026 12:42
Sign up for freeto join this conversation on GitHub. Already have an account? Sign in to comment

Labels

documentationImprovements or additions to documentationsize/sskip-changesetPR has no user-facing published change; bypasses the changeset gate

Projects

None yet

3 participants

@hotlong@os-zhuang@claude