[os-steve]

Fixes #14984

Two prose statements of the governed REPO set lived outside the register (GOVERNED_REPOS in scripts/pm/check-governed-merges.mjs) and went stale the minute the register grew to five repos with the landing of #14867 (5d4d55ae). Remedy shape ruled by the skills seat (triage comment 5529676559): remove the mirrors and point at the register — no repo name and no count outside GOVERNED_REPOS, so there is nothing left to drift and nothing to parse. Two one-line edits, both on governed surfaces (.claude/** x2) ⇒ draft PR, in-seat review, human merge; skip-changeset (nothing published from any package).

The two lines

.claude/skills/pm-dispatch/SKILL.md:723 — before (119 bytes):

`CLAUDE.md`;agent 指令文件按此跨仓同判 —— objectui、cloud、objectos 一并在内（2026-08-18 裁定）

after (118 bytes; the （2026-08-18 裁定） date pointer kept verbatim per the X0 quote policy):

`CLAUDE.md`;agent 指令文件按此跨仓同判 —— 仓集读 `GOVERNED_REPOS`,此处不列（2026-08-18 裁定）

.claude/hooks/guard-governed-enqueue.sh:54 — before:

#      the same call answers for all four governed repos.

after (75 columns, inside the header's 80-column comment block):

#      the same call answers for every repo configured in `GOVERNED_REPOS`.

No repo name and no count on either line; no issue number inside either line (the id lint covers .claude/skills/pm-dispatch/**). check-governed-merges.mjs itself, including its 2026-08-18 quotation, is untouched.

Deviation from the dispatch word, measured

The dispatch word asked the SKILL.md line to name both GOVERNED_REPOS and scripts/pm/check-governed-merges.mjs, at net 0 lines. check:pm-skill-ratchet also holds a 120-byte per-line budget the dispatch reading did not price: my first cut (0931bc0b, both names on the line) went red with L723 (180B). The line has 79 fixed bytes (the `CLAUDE.md`;agent 指令文件按此跨仓同判 —— prefix plus the date pointer), the path alone is 38 bytes and the constant 16, so both cannot fit; wrapping would spend a line the dispatch word forbids. The line now names GOVERNED_REPOS alone: the script is already named in this file at lines 264, 791 and 964, and git grep -n GOVERNED_REPOS resolves the constant in one hop. If the seat prefers the path on the line at the cost of one wrapped line (970 → 971, ceiling 1005), that is a one-line change on this PR.

Third-mirror sweep

git grep -n -i "objectui、cloud、objectos\|four governed repos\|four repos\|五个\|四个.*仓" -- '.claude' 'AGENTS.md' 'CLAUDE.md' 'skills' at BASE fddfc8db: 9 hits, of which only the two target lines state the governed repo set or its size. The other seven are 五个 in unrelated prose — .claude/agents/os-dev.md:82 (lock waiters) and :384 (a wave of devs), .claude/skills/checklist-author/SKILL.md:50 (five sweep angles), .claude/skills/pm-dispatch/SKILL.md:795 (five health metrics), references/lanes/engine.md:30 (driver columns), .claude/skills/spec-property-retirement/SKILL.md:292 and :316 (generated artifacts). A repo-wide git grep for the two literal strings hits only the two sites, so no test pins them. Neighbouring SKILL.md:724 ((objectos 指令面 PR 照样 draft/人工合并)) names one repo as a worked example of the path-fork, not the set or its size — left as is.

Verification — head 16bd3e51 (the pushed commit), each exit code captured before any pipe

Gate list derived by node scripts/pm/dispatch-gates.mjs --repo objectstack-ai/objectstack (no paths passed; change set read from the merge base: the two files, 14 families) — one family beyond the dispatch word's list, check:bash32-floor, matched via .claude/hooks/**, run and green. check-governed-queue-guard is CI-measured only (reads the workflow payload).

Gate	Exit	Verdict line
node scripts/check-closing-keyword-parity.mjs	0	check-closing-keyword-parity: OK (3 parsers agree on all 9 keywords and both measured separators; sweep found 5 file(s) carrying the grammar across 8178 tracked file(s), all registered).
node scripts/check-comment-mask-corpus.mjs	0	✓ comment-mask corpus sweep [scripts/js-comment-mask.mjs]: 5833 files, 0 disagree, 0 unparseable, 48.9s (comparator self-test: 12 cases pass).
pnpm --filter @objectstack/lint run check:doc-formula-expressions	0	✓ check:doc-formula-expressions: 22 record-scoped formula example(s) across 426 files / 1367 TS blocks judged clean by @objectstack/formula. (two earlier runs exited 3 = PREREQUISITE NOT MET, @objectstack/formula then @objectstack/lint unbuilt — NOT MEASURED in the gate's own words; both built under os-verify-lock.sh, VERDICT command-exit 0, then measured)
pnpm check:agent-test-spelling	0	✓ check-agent-test-spelling: 0 violations — 432 file(s) · 5850 bare -- token(s) · 1390 launcher-rooted run(s) · 9 separator(s) JUDGED · 5 vitest-backed script name(s) derived from 81 manifest(s)
pnpm check:doc-authoring	0	✓ doc authoring guard: 393 files clean — no bare metadata literals. (+ three further ✓ lines)
pnpm check:nul-bytes	0	check-nul-bytes: OK (scanned 8171 text file(s) -- 8171 tracked, 0 untracked-not-ignored; skipped 7 binary; no raw ASCII control bytes).
pnpm check:pm-governed-merges	0	✓ check-governed-merges --self-test: 245 assertions (…)
pnpm check:pm-governed-prose	0	✓ check-governed-prose: 2 instruction surface(s) name all 5 registered governed surfaces (docs/adr/** · .claude/** · skills/** · AGENTS.md · CLAUDE.md) and claim no others.
pnpm check:pm-skill-id-lint	0	✓ check-skill-id-lint: 23 file(s) clean (pattern /#[0-9]{3,}/g).
pnpm check:pm-skill-ratchet	0	✓ check-skill-line-ratchet: .claude/skills/pm-dispatch/SKILL.md is 970 lines (ceiling 1005; headroom 35). — net 0 lines; L723 118 bytes (was 119)
pnpm check:refd-timer-probe	0	OK check-refd-timer-probe: 5828 source file(s) swept; the process-global timer probe is read in packages/qa/refd-timer-testkit/src/index.ts and nowhere else.
pnpm check:skill-frame-sync	0	✓ check-skill-frame-sync: 2 copies of the decision frame are structurally isomorphic across 2 files
pnpm check:watch-hint-literal	0	✓ check-watch-hint-literal: 48 declaration(s) across 4 rostered name(s) -- ROOT_DIR_WATCH_HINTS 32, ROOT_FILE_WATCH_HINTS 9, ROOT_WATCH_HINTS 2, DECLARED_WATCH_HINTS 5 -- every one an array of quoted literals inside its own statement, every rostered name non-empty, and no unrostered spelling of the idiom in the tree.
pnpm check:bash32-floor (added by the re-derivation)	0	✓ check-bash32-floor: 26 tracked shell file(s) under scripts/**, .claude/hooks/**, .githooks/** name no bash 4+ construct outside a comment, a guarded ${VAR:-} read, or a non-command position.
bash -n .claude/hooks/guard-governed-enqueue.sh	0	(syntax check, no output)
.claude/hooks/guard-governed-enqueue.selftest.sh	0	49 passed, 0 failed

Both edits were proved on disk before the commit by anchored counts (old text 0, new text 1 in each file), not by the editor's exit code.

---

Generated by Claude Code