skills(pm-dispatch): scope-narrowing follow-up — revert the decision-card assign half per the 2026-08-19 rulings

[os-warren]

Refs #9976 (already closed by the earlier PR — intentionally no closing keyword here)

Scope-narrowing follow-up. PR #9982 merged carrying edit 2 (the decision-card same-stroke assign/unassign of os-zhuang, with the inbox defined as the Assigned queue). After delivery and before merge, the maintainer narrowed the scope; this PR reverts exactly that half and keeps everything else #9982 landed.

Maintainer rulings (2026-08-19, PM chat, verbatim):

我感觉决策卡推给我太麻烦了,我需要和ai讨论才能判断,这个还是维持之前的样子。我会定期和ai讨论。

只有需要我审核的pr 推给我。

What this reverts (vs current main)

• State-model needs-user-decision row back to its pre-skills(pm-dispatch): notification handover — review-request and assign os-zhuang, retire the decision digest #9982 wording (「维护者的收件箱」) — no Assigned-queue semantics, no same-stroke assign, no unassign-on-transcription, no reclaim carve-out.
• Escalation flow step 2: drop the same-stroke assign os-zhuang.
• Escalation flow step 3 (atomic four-piece record): drop the same-stroke unassign clause.

What this keeps and grounds

• The governed-PR review-request step (four-piece ACCEPT fork) stays — explicitly confirmed by the second ruling, which is now quoted in the step as the scope grounding (push channel is PRs needing review, and only that).
• The decision-digest retirement stays — 「我会定期和ai讨论」 means chat replaces the digest; the inbox is worked by periodic maintainer-AI discussion. The first ruling's full verbatim is recorded once, compactly, at the facets-block provenance line.

Line budget (682-ceiling shrink-only ratchet)

• main after skills(pm-dispatch): notification handover — review-request and assign os-zhuang, retire the decision digest #9982: 678 · this PR: 680 · ceiling: 682 · headroom 2. The +2 pays for the two verbatim ruling recordings; the reverts themselves are line-neutral (single-line table row and inline clauses).

Gates (at head 9368fde78, fresh merge base 90f4d5dc3, exit codes captured before any pipe)

Derivation: node scripts/pm/dispatch-gates.mjs (no args) — "change set derived from git — 1 path(s) vs merge base 90f4d5d", same 7 families, tier line "Model tier — MANDATORY: claude-fable-5" (this card ran on it).

• check:pm-skill-ratchet — "✓ check-skill-line-ratchet: .claude/skills/pm-dispatch/SKILL.md is 680 lines (ceiling 682; headroom 2)."
• check:pm-governed-prose — "✓ 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."
• check:pm-skill-id-lint — "✓ check-skill-id-lint: 15 file(s) clean"
• check:skill-frame-sync — "✓ check-skill-frame-sync: 4 copies of the decision frame are structurally isomorphic across 3 files"
• check:skill-compatibility — "✓ check-skill-compatibility-version: 11 SKILL.md file(s) reconciled against 77 workspace packages"
• check:doc-authoring — "✓ doc authoring guard: 379 files clean — no bare metadata literals."
• check:pm-governed-merges --self-test — "✓ 77 assertions"
• check:doc-formula-expressions — "✓ check:doc-formula-expressions: 22 record-scoped formula example(s) across 404 files / 1414 TS blocks judged clean"

Reverse verification at this head (direction predicted first): 683 lines turns the ratchet RED — observed "is 683 lines; the ratchet ceiling is 682" (exit 1); restored, green at 680 (exit 0), worktree clean.

Merge note: the branch merged origin/main after #9982 landed; the textual merge auto-resolved the three revert hunks to main's side (they were net-unchanged vs the old base on this branch), so they were re-applied in a follow-up commit — the final file is byte-identical (git hash-object match) to the reviewed amendment commit 288a1df47.

No changeset: instruction-file-only PR, nothing published ships (skip-changeset).

---

Generated by Claude Code