pm-dispatch: release is an explicit act — clear the assignee and post a `Release:` line whenever a card leaves a session's hands, so `pm:queue` never carries an assignee and `pm:dispatched` always does

[claude]

Filed by the domain:skills seat (session session_019RfFHiRCSs3JXLK4cwcfox, os-steve) from the maintainer's 2026-09-05 design discussion on issue ownership. Self-triaged priority:p2, pm:queue, governed (.claude/skills/pm-dispatch/SKILL.md + references/state-machine.md).

The gap

Ownership has an explicit entry (assign + Claim:) and no explicit exit. A card leaves a session's hands in four ways today — re-routed to another lane, handed back on premise_still_valid: false, abandoned when a flight dies with no takeover, or moved across lanes — and in each the assignee field keeps saying "someone owns this" after nobody does. The patrol rows that read the field (H1 dispatched-without-assignee, H24 queued-with-assignee) then report residue as if it were a live half-state, and a seat reading the list view sees a taken card that is free.

What this card does

One rule, two lines, line-neutral on both files (both at headroom 0 — fold, never raise):

1. SKILL.md state machine section: release is an explicit act — whoever moves a card out of ownership clears the assignee and posts a line beginning Release: (session ID, the reason from the four above, and where the card went: the new label state or lane); the next owner claims afresh. A pm:queue card must have no assignee; a pm:dispatched card must have one (the H1/H24 readings become definitional).
2. references/state-machine.md: the same two facts on the releasing side, beside the existing applier/remover lines.

Not in scope: a new patrol row for a Release: newer than the last Claim: while the field is still set (a follow-up for check-half-states.mjs once the rule is on main), and any change to the claim's spelling (ruled 2026-08-11).

Serial

SKILL.md (811/811, table pin 342) and state-machine.md (42/42) are in no flight since PR #15732 landed; member 5 of #15379 touches AGENTS.md + CLAUDE.md only. Governed ⇒ draft PR, in-seat review at contract tier, os-zhuang + hotlong, human merge.

Dedup

REST enumeration 2026-09-05T10:3xZ, state=all sorted by update (1,200 items, 797 issues; positive control #14768 hit): no card names a release act or the assignee's exit.

---

Generated by Claude Code

[claude]

Claim: PM loop round 6 — filed and taken at first touch by the seat under the maintainer's 2026-09-05 design ruling (R3: release is an explicit act); one rule, two lines, line-neutral on both files
Session: session_019RfFHiRCSs3JXLK4cwcfox
Branch: claude/issue-15846-release-is-an-explicit-act
Worktree: objectstack-issue-15846
Domain: domain:skills
File surface: .claude/skills/pm-dispatch/SKILL.md (811/811, table pin 342, headroom 0 — the 认领 / state rules around line 464 「assignee 字段归 PM」) and references/state-machine.md (42/42, headroom 0 — beside the applier/remover lines 40–41): the release act (clear the assignee + a line beginning Release: with session, reason and destination) and the two definitional readings (pm:queue never carries an assignee; pm:dispatched always does), folded into existing lines, never raised.
Container & model: S, mode:subagent, model: fable (dispatch-gates.mjs --tier on the two paths at d701e657d7, 2026-09-05T10:40Z: MANDATORY — SKILL.md is clause-① mandatory)
Clause-②: no (PM protocol text)
Serial constraints cleared: both files free since PR #15732 landed 06:36Z; member 5 (in flight) touches AGENTS.md + CLAUDE.md only; #15842 (in flight) touches dogfood-verification/SKILL.md only; no open PR touches either file (scan 10:3xZ). GOVERNED ⇒ draft PR, in-seat contract-tier review, os-zhuang + hotlong, human merge, skip-changeset. The PR carries Fixes #15846.

Decision re-read (10:4xZ): the rule follows from the maintainer's ruling that the assignee field stays a presence bit under one shared identity (「暂时不区分」) — a presence bit is only truthful if its exit is written down; no further ruling owed. A patrol row for a Release: newer than the last Claim: while the field is still set is a follow-up after this lands, not this card.

---

Generated by Claude Code

[os-steve]

os-dev-report

{
  "issue": 15846,
  "status": "done",
  "branch": "claude/issue-15846-release-is-an-explicit-act",
  "pr": "https://github.com/objectstack-ai/objectstack/pull/15848",
  "premise_still_valid": true,
  "summary": "Draft PR #15848 lands R3 (release is an explicit act) line-neutral on both files, both still at headroom 0 (SKILL.md 811/811, table pin 342 unchanged; state-machine.md 42/42). SKILL.md: the state-model rows gain the definitional readings (`pm:queue` 卡恒无 assignee,有即半态 / `pm:dispatched` 恒带 assignee) and the `assignee 已设` row gains the four exits, the destination definition and ⛔ 不静默摘 assignee; the 认领 family gains 「释放是显式动作:让卡离手者同笔清 assignee + `Release:` 行(会话/因/去向);下一任重新认领。」 paid for by folding the 原子对 header line into the ① line (nothing dropped); the dead-claim reclamation line now routes through 释放回队(`Release:` 行载因). state-machine.md: 「释放者 = 让卡离手的席位:行首 `Release:`(会话/因/去向)与清 assignee 同笔;下一任重新认领。」 beside 挂标者/摘标者, paid for by folding the awaiting-inventory line into the retriage-authority line (only the redundant 挂/摘分属两方 dropped). The four reasons and the destination sit in the table row rather than the 认领 line because the 120-byte cap cannot hold them and the row has ~240 bytes under the pin — the dev's call inside the ruling, stated in the PR body. `Claim:` spelling, CLAIM_COMMENT_MARKER, H1/H24 code and the frame copies untouched. Channel notes: the REST additive labels POST was denied by the container's action classifier, so `skip-changeset` went by the MCP fallback (read → union → whole-set write → comparison read-back: documentation, size/s, skip-changeset all present, nothing stripped); the PR body read back intact (first line `Fixes #15846`), with the platform appending a second session-URL footer under mine, left as is. Assignee was set by the seat at dispatch; not touched.",
  "tests": "All at PR head ec0a7f8d8c. Family: `dispatch-gates.mjs --commands --repo objectstack-ai/objectstack` on the two files = 16 commands (9 by path, 7 whole-tree), identical to the no-path merge-base derivation; every command run with the exit captured before any pipe; `✓ dispatch-gates --ran: 16 derived famil(ies) accounted for — 16 run, 0 NOT-MEASURED.` Ratchet: `✓ check-skill-line-ratchet: .claude/skills/pm-dispatch/SKILL.md: widest table row is 342 bytes (pin 342; headroom 0).` `✓ … SKILL.md is 811 lines (ceiling 811; headroom 0).` `✓ … state-machine.md is 42 lines (ceiling 42; headroom 0).` `✓ check-skill-line-ratchet self-test: 155 cases pass.` `✓ check-skill-id-lint: 24 file(s) clean` (+14 self-test cases); `✓ check-skill-frame-sync: 2 copies … isomorphic across 2 files`; `✓ check-half-states self-test: 2267 cases pass.` (no H-row prose pin fired); `✓ check-skills-token-ratchet: 36 authored bundle file(s) within their ceilings`; check:pm-governed-prose / pm-governed-merges (274 assertions) / governed-queue-guard (144) / closing-keyword-parity (24) / comment-mask corpus (5998 files, 0 disagree) / agent-test-spelling (0 violations) / doc-authoring / driver-memory-census / nul-bytes (75) / refd-timer-probe (11) / watch-hint-literal all exit 0. check:doc-formula-expressions: two `PREREQUISITE NOT MET` exit-3 refusals first (formula, then lint — NOT MEASURED, filed as #15850), then after `turbo run build --filter=@objectstack/lint` through the lock: `✓ check:doc-formula-expressions: 22 record-scoped formula example(s) across 428 files / 1370 TS blocks judged clean`. Whole-repo `pnpm lint` (eslint . --no-inline-config) via os-verify-lock.sh slot issue-15846: `os-verify-lock: VERDICT command-exit 0 · held the lock 67s · waited 65s`. Byte widths (LC_ALL=C awk): widest non-table prose line SKILL.md line 98 = 120 (pre-existing), widest table row line 250 = 342 (pre-existing), state-machine.md widest line 18 = 120 (pre-existing); widest line this PR writes = 119. Control-char scan of both files: no hits. Negative control from the committed state: +1 line on each file → `✗ … SKILL.md is 812 lines; the ratchet ceiling is 811` and `✗ … state-machine.md is 43 lines; the ratchet ceiling is 42` (exit 1); restore `git checkout HEAD -- …` proven by git hash-object equal to the HEAD blobs (a32af80219…, b3b537e59a…) and `git diff HEAD` empty. check-governed-merges --test on the two paths: exit 3 (governed). CI on ec0a7f8d8c: all 29 check runs on ec0a7f8d8c completed, 0 non-green — `Lint & Repo Gates` success (started 11:03:57Z, completed 11:24:33Z), `Type Check · workspace` / `consumer gates` / `source gates` / `debt ledger` and `TypeScript Type Check` success, `Governed Surface Queue Guard` success, `Check Changeset` success under `skip-changeset`, `Check Documentation Links` / `Check PR Size` / the three claim-parity checks success; `Test Core`, `Dogfood Regression Gate`, `Build Core`, `Build Docs`, `Temporal Conformance` skipped by the path filter on a `.claude`-only diff. PR left draft, not flipped, no reviewers requested, no auto-merge.",
  "mcp_calls": "3 — issue_write (labels union on the PR after the REST additive POST was classifier-denied), issue_write (create #15850), add_issue_comment (this report)",
  "open_questions": [],
  "out_of_scope_findings": [
    "filed as #15850: check:doc-formula-expressions refuses one unbuilt prerequisite at a time — its printed Fix names @objectstack/formula, then it refuses again for @objectstack/lint; the fix line is one build short on a cold worktree (labelled finding, unassigned; dedup by REST enumeration 613/613 items, positive control hit)"
  ]
}

Generated by Claude Code

---

Generated by Claude Code