[PM seat] domain:skills — 🟢 os-warren · session_01KTZmMfzVzjNvyaLyQ8mHvg · R9 (PR #20899 awaiting approval; PR #20902 in review) · landed 12

[huangyiirene]

This post is the single authoritative registration of the skills seat (seat 1, domain:skills#1). Four-section template; the job description is versioned at .claude/skills/pm-dispatch/references/lanes/skills.md — ⛔ not copied here. Body r116, written 2026-09-30T17:12Z by session_01KTZmMfzVzjNvyaLyQ8mHvg (os-warren) at the 17:12Z patrol. r115 (R9 dispatch), r114, r113, r112, r111, r110, r109, r108, r107, r106, r105, r104, r103 and r102 (the previous shift's hand-over, session_01MjvgiFAmjHqsxy1XLiVYfH, os-steve) are in the edit history, ⛔ not restated. r102's 「Maintainer items open at stand-down」 (1)–(5) and 「Seat tasks owed」 (a)–(h), r81's and r69's lists and r64's fact lines are carried BY REFERENCE.

Current PM

🟢 os-warren · session_01KTZmMfzVzjNvyaLyQ8mHvg · seated 2026-09-30T04:10Z on the maintainer's /pm-dispatch skills (round-open marker 5903866929; R2 marker 5905508713).

• Write identity: every content write goes through the scripts/pm/ tools over the fleet-write relay, stored as objectstack-fleet[bot]; the session ID is the attribution.
• Served model claude-opus-5-5, below CONTRACT_REVIEW_TIER ⇒ every owed ## Contract review is rendered by an isolated at-tier subagent, ⛔ never in-seat.
• batch 3, mode:subagent. Patrol Routine trig_014UfhpWroArwBsHBSK2YKPv (self-bound, hourly at :11).

The review-tier deviation is closed. It ran from 2026-09-30 ~08:40Z to 13:47Z. Two review subagents had died on a host-reported weekly-limit error; a one-word probe at 13:47Z was served claude-fable-5-1, and all four owed records were rendered at tier.

• Seat lesson: probe the tier before treating a quota error as a wall that lasts days.
• The one-shot timer trig_01RxTU4seRwCTS5c7asHdo6G is deleted.

Maintainer's standing word this shift: when a dev's skip-changeset / PR-assignee write is refused by its session classifier, the seat applies them itself, once it has checked that the diff publishes nothing, and re-runs the stale Check Changeset. Dispatches from R4 on give the dev no label write at all.

Inherited + running ledger

• Landed this shift: 12 PRs, 14 cards closed completed.
  • PR docs(adr): ADR-0005 appendix (c) — the persisted document is the parsed body once every round-trip key is declared #20777 (docs(adr): ADR-0005 appendix (c): the persisted document is the parsed body once every round-trip key is declared (#20051 stage iii, ruling 甲) #20457) · PR docs(skills): objectstack-automation routes an outbound credential to a connector credentialRef and drops the Slack webhook-url example #20778 (skills(objectstack-automation): the http example routes an outbound credential to credentialRef and stops teaching a secret-bearing webhook url (the skills half of #20590) #20657) · PR docs(skills): objectstack-query — engine aggregate row admits search / searchFields #20776 ([finding] skills/objectstack-query/SKILL.md tells an agent that engine aggregate refuses search / searchFields — false since PR #20487 #20500).
  • PR docs(governed): the retirement route is retiredKey() on any shape, with the retiredAfter stamp and the step-18 registry shape #20797 ([finding] the retirement route for a .strict() schema: spec-property-retirement §2 and AGENTS.md prescribe delete + guidance map, while retired-key.ts and this week's retirements use retiredKey() #20465 + [finding] spec-property-retirement SKILL.md tells a step-18 author to append to conversionIds and extend the rationale string: both change shape when PR #20572 lands #20575) · PR docs(skills): objectstack-automation calls the api flow secret required and routes explicit-only starts to autolaunched #20796 (skills/objectstack-automation still calls the api flow secret optional and says type: 'api' can be invoked explicitly only — both false since PR #20551 (split from #20553) #20569).
  • PR fix(fleet-write): read every relayed body back — a stored body that is not the bytes sent exits 4, never 0 #20806 (finding(fleet-write): an issue_patch body of about 41 KB comes back with two multi-byte characters replaced by U+FFFD, about 16 KB apart, identically on a re-send — and dispatch.mjs reports success without comparing the read-back #20571) · PR fix(pm): check-expected-skips derives schedule/dispatch-only jobs as expected skips under their raw name #20835 (check-expected-skips reads scaffold-e2e.yml's by-design 'Registry canary' skip as outside the roster (exit 4) on every PR that touches create-objectstack #20756).
  • PR docs(pm-dispatch): the draft-conversion reading splits by queue state — no queued PR left the queue on a draft conversion #20845 ([finding] pm-dispatch platform reading: in this repo a draft conversion alone did NOT remove a queued PR from the merge queue (PR #20695), contrary to platform-readings.md:42 #20764) · PR docs(skills): retirement kit routes a RETIRED_KEYS_BY_MAJOR entry through entries/retired-keys/ and gen:migration-registry, and an enum-value narrowing through enumWithRetiredValues() #20828 ([finding] retirement kit §3 and build-schemas.ts gate (b) tell an author to paste the RETIRED_KEYS_BY_MAJOR entry into migrations/registry.ts, which is generated from entries/ since #7297 #20803) · PR docs(skills): objectstack-query teaches the served route for a related record's column, not the refused nested form #20811 ([finding] skills/objectstack-query teaches the nested relation filter { relation: { field: value } } as a working form; no data-path driver serves it, and PR #20781 makes the engine refuse it #20782) · PR docs(agents,pm-dispatch): a draft conversion drops only an unqueued PR's auto-merge; a queued governed pull-back goes to the maintainer #20854 ([finding] AGENTS.md says a draft conversion removes a PR from the merge queue — every measured queued draft in this repo stayed queued and merged (Prime Directive #14's pull-back rests on it) #20847, p1).
  • PR fix(pm): the queue guard's refusal remedies send a queued PR to the maintainer's manual removal, never claiming a DRAFT conversion dequeues it #20894 ([finding] the Governed Surface Queue Guard's refusal text tells a seat to dequeue a governed PR by converting it to DRAFT — measured not to dequeue in this repo (same family as #20847) #20855, p1), MERGED 17:05Z as da93a8b3e. Non-governed: CI plus the seat's read, ACCEPT 5915517440. The maintainer's one-card exception to the tooling slot is on record there.
• Closing keyword: Fixes closed no card on the first eleven queue merges; the seat closed each with close-cards. On the twelfth (PR fix(pm): the queue guard's refusal remedies send a queued PR to the maintainer's manual removal, never claiming a DRAFT conversion dequeues it #20894) it did close [finding] the Governed Surface Queue Guard's refusal text tells a seat to dequeue a governed PR by converting it to DRAFT — measured not to dequeue in this repo (same family as #20847) #20855. close-cards skips a closed card, so the seat posted the closing note and stripped pm:dispatched + the assignee by hand.
• --no-skip-pr-referenced: used once, on [finding] skills/objectstack-query teaches the nested relation filter { relation: { field: value } } as a working form; no data-path driver serves it, and PR #20781 makes the engine refuse it #20782. Open PR feat(objectql): serve the nested-relation filter in where — lowered at the engine seam, the related object read as the caller, a loud cap, drivers untouched (#20802) #20872 names it only in prose.
• R9, both Tier H:
  • docs(adr): ADR-0005 appendix (c): the storage switch is in force since PR #20868, and a saved view stores the parsed value of every key its body carried, not the whole parse output (#20051 ruling B) #20879 → PR docs(adr): ADR-0005 appendix (c) — the storage switch is in force for view, and a saved view stores the parsed value of every key its body carried (ruling B) #20899, draft, head 950180d8. Contract review PASS 5915913589 (isolated claude-fable-5-1 subagent); ACCEPT 5915930227; 速读 5915939488; needs-user-decision; review requested from os-zhuang and hotlong. CI green. Awaiting an authorized APPROVED review.
  • [finding] skills/objectstack-query now teaches that the engine refuses { relation: { field: value } } (PR #20811), but since PR #20872 (ca5408c62) the engine serves it; the published skill states a refusal that no longer happens #20888 → PR docs(skills): objectstack-query states the served nested-relation filter in where, its limits, and the two-step route past them #20902, draft, head b73023a3. Contract review PASS 5916062797 (isolated claude-fable-5-1 subagent), with three non-blocking precision items the seat lands as they are (named in the ACCEPT). Lint & Repo Gates is still running at this refresh; the ACCEPT, 速读 and review request follow when CI is green.
• Awaiting a human 1 (PR docs(adr): ADR-0005 appendix (c) — the storage switch is in force for view, and a saved view stores the parsed value of every key its body carried (ruling B) #20899) · queued 0.
• Filed by this seat this shift (all triaged or bare, none claimed):
  • [finding] two content/docs sentences enumerate the engine aggregate option set as "only where / groupBy / aggregations" — false by context, having, search, searchFields (the docs half of #20500) #20792: the content/docs half of [finding] skills/objectstack-query/SKILL.md tells an agent that engine aggregate refuses search / searchFields — false since PR #20487 #20500, domain:devx.
  • [finding] retirement kit §3 and build-schemas.ts gate (b) tell an author to paste the RETIRED_KEYS_BY_MAJOR entry into migrations/registry.ts, which is generated from entries/ since #7297 #20803: closed above.
  • [finding] AGENTS.md says a draft conversion removes a PR from the merge queue — every measured queued draft in this repo stayed queued and merged (Prime Directive #14's pull-back rests on it) #20847: closed above.
  • [finding] the Governed Surface Queue Guard's refusal text tells a seat to dequeue a governed PR by converting it to DRAFT — measured not to dequeue in this repo (same family as #20847) #20855: closed above.
  • [finding] docs: query-syntax.mdx "Filtering Across Relationships" says SqlDriver compiles a nested relation object and emits a dotted key to Knex — both are refused at the engine before any driver since 4b4ee88f #20876: query-syntax.mdx's stale mechanism paragraph, for the docs lane.
• Watch items:
  • PR feat(objectql): serve the nested-relation filter in where — lowered at the engine seam, the related object read as the caller, a loud cap, drivers untouched (#20802) #20872 MERGED as ca5408c62; its skill half is [finding] skills/objectstack-query now teaches that the engine refuses { relation: { field: value } } (PR #20811), but since PR #20872 (ca5408c62) the engine serves it; the published skill states a refusal that no longer happens #20888 (in flight above). The analytics half is #20802 analytics half (domain:services): the cube read and the analytics read scope answer { relation: { field: value } } as the engine seam now serves it — as the caller, capped, one answer on every face #20887 (domain:services); the skill says nothing about analytics until it lands.
  • spec: gate (b) of build-schemas.ts tells an author to paste RETIRED_KEYS_BY_MAJOR lines into migrations/registry.ts, which is generated from entries/ since #7297; two guidance-slot docblocks describe the old retirement route (spec half of #20803) #20809 (domain:spec) quotes PR docs(skills): retirement kit routes a RETIRED_KEYS_BY_MAJOR entry through entries/retired-keys/ and gen:migration-registry, and an enum-value narrowing through enumWithRetiredValues() #20828's kit wording when it lands.
• Queue at this refresh (objectstack domain:skills):
  • pm:queue: empty (unlock scan at 17:12Z: [decision] 半状态巡逻的送法与形状：每天都在改的仪器靠「手工复制三个文件」分发，副本必然掉队——换成引用并把「写坏的」那半搬到写侧 #18471 still open, objectui#10887 unreadable). Standby between rounds: the hourly patrol, the unlock scan, and following queued PRs to MERGED.
  • pm:blocked: docs(react-pages,skills,showcase): teach the data-invalidation nonce in a react page's live-data effect, and opt the two showcase react pages in (ruling A on objectui#10887 member 3) #20395 (Blocked-by: objectui#10887, unreadable here) and [finding] objectui's half-state findings reach no reader: the board has no anchor by ruling, and its copy of check-half-states still refuses in agent containers with the conclusion #13544 disproved #18290 (behind [decision] 半状态巡逻的送法与形状：每天都在改的仪器靠「手工复制三个文件」分发，副本必然掉队——换成引用并把「写坏的」那半搬到写侧 #18471, open).
  • Program [program] CLAUDE.md is a pointer to AGENTS.md in every repo — no second instruction set to maintain (maintainer 2026-09-09) #17161. Decision box (this lane): 0. Seat 2 [PM seat] domain:skills · seat 2 — ⏳ vacant · last incumbent session_017ETYWqMQD4qMtZzAGovWNi (os-steve) · no own output since 2026-09-21 · liveness patrol 2026-09-25 #19287 ⏳ vacant.

Hot-file serial queue

• skills/objectstack-query/** → [finding] skills/objectstack-query now teaches that the engine refuses { relation: { field: value } } (PR #20811), but since PR #20872 (ca5408c62) the engine serves it; the published skill states a refusal that no longer happens #20888.
• docs/adr/0005-metadata-customization-overlay.md → docs(adr): ADR-0005 appendix (c): the storage switch is in force since PR #20868, and a saved view stores the parsed value of every key its body carried, not the whole parse output (#20051 ruling B) #20879.
• Every other file of the lane: FREE.

Notes

• Job description and write paths. The job description is references/lanes/skills.md. Content writes go only through scripts/pm/post-stamped.mjs / label-write.mjs / issue-create.mjs / close-cards.mjs and fleet-write/dispatch.mjs --actions-file; a body refresh needs --ack-through=ID.
• Landing. Tier H = an authorized APPROVED review, then this seat clears needs-user-decision and lands, unless the approver already did. Non-governed = CI + the seat's read (plus an at-tier record only if a review surface is touched) ⇒ relay pr_ready + automerge_enable. Landing = MERGED through the queue.
• Platform readings this shift:
  1. One relay HTTP 500 with zero written; the informed re-send stored identical.
  2. Fixes #N did not fire on eleven queue merges out of eleven.
  3. issue-create UNCONFIRMED on a list lag: the relay job log showed HTTP 201 ([finding] retirement kit §3 and build-schemas.ts gate (b) tell an author to paste the RETIRED_KEYS_BY_MAJOR entry into migrations/registry.ts, which is generated from entries/ since #7297 #20803).
  4. A dev's label write is refused by its classifier on most PRs this shift; Check Changeset reads the label live, so rerun_failed_jobs clears the stale red.
  5. label-write answers "NOTHING WAS WRITTEN" when a --remove label is already absent. Read the card and re-send with the live labels.
• Checkout: /home/user/objectstack; check-harness-current.mjs CURRENT; the charter is unchanged since the R1 marker.
• Subscriptions: PR docs(adr): ADR-0005 appendix (c) — the storage switch is in force for view, and a saved view stores the parsed value of every key its body carried (ruling B) #20899 and PR docs(skills): objectstack-query states the served nested-relation filter in where, its limits, and the two-step route past them #20902 (PR fix(pm): the queue guard's refusal remedies send a queued PR to the maintainer's manual removal, never claiming a DRAFT conversion dequeues it #20894 unsubscribed at its merge).

[huangyiirene]

Audit: founding stock transfer executed (devx seat, session session_01GZKbx4xyF7U5WXj6ch49BM, 2026-08-11 ~09:2xZ)

Maintainer instruction (verbatim): 「所有现有 skills 相关卡片转到 新的项目经理」.

Enumeration: swept all open domain:devx cards (26) + a title search for skills/SKILL cards outside the label. 8 transferred — #7548, #7518, #7492, #7493, #7553, #7583, #7512 (borderline, noted on-card), #7498 (in-flight, 改标不改辖). Each carries a provenance comment quoting the instruction; label writes read back — label:domain:skills returns exactly the 8, pm-states intact (7 pm:queue, 1 pm:dispatched).

Deliberately NOT transferred (recorded so the boundary is auditable): QA-checklist authoring cards (#7547, #7573, #7569, #7530 — docs/qa/**, not a skills root), scripts/pm tooling findings (#7412), CI-gate/docs findings (#7484, #7353, #7347, #7329, #7009), and the rest of the devx stock — they stay domain:devx. domain:skills was auto-minted by the first label write; color/description cosmetics + the SKILL domain-table row land with #7548's PR.

---

Generated by Claude Code

[os-help]

Audit: seat takeover

• Incoming PM: os-help, session session_01ATJiqhFNL82YwV8udPkoqX, 2026-08-11.
• Staffing authority: the maintainer invoked /pm-dispatch skills in this session — the dedicated PM session the founding note (⏳ vacant, "awaiting the maintainer's staffing") was waiting for.
• Same-stroke triple: title → 🟢 os-help, assignee → os-help, body 当前 PM section → the triple above. Queue table refreshed to post-ruling states (all seven open cards now carry maintainer rulings recorded 08:10Z; decision box empty at takeover).
• First-round plan: consolidated batch of the seven ruled cards, one PR, mode:cloud, model: claude-fable-5, dispatched the moment PR pm-dispatch SKILL: require a four-lens block on every needs-user-decision card, Blocked-by-aware batch selection, round-brief decision flags #7552 (the only in-flight SKILL.md PR, awaiting the maintainer's manual merge) reaches terminal state. Round report goes to the maintainer in chat.

---

Generated by Claude Code

[os-help]

Audit: seat stood down → ⏳ vacant (founding PM, session session_01ATJiqhFNL82YwV8udPkoqX)

Maintainer instruction, verbatim (专题讨论, 2026-08-11): 「fable 5 没token了。剩余的任务开 issue 我让其他人处理」.

Exit checklist executed:

1. New dispatches stopped on receipt.
2. In-flight cleared to zero — PR pm-dispatch SKILL batch: eight ruled amendments — four-lens block, ADR-class skills, three-query board, queue criteria, five shift lessons, claim spelling, seat-post bounds, dispatch-contract rank (#7631) #7698 landed (all 8 member cards closed); its cloud dev session session_01KCFsFypEHm1Cnp26yAofxg archived.
3. Remaining work filed, not carried in context — pm-dispatch SKILL structural compression: 3050 → ~1400–1600 lines (PR₂ of batch #7631, maintainer-directed) #7754 (PR₂ compression, with the directive verbatim, measured composition, method, ⛔ no-rule-deletion constraint, gates, tier); [finding] A skills-root PR was queued by a sibling session ~1h after the ADR-class ruling — a ruling recorded only on cards cannot bind sessions that haven't read it #7701 updated with the second incident specimen; batch card [batch] pm-dispatch SKILL consolidated amendments + compression — eight ruled cards, two serialized PRs (skills seat, round 1) #7631 closed with its ledger.
4. Seat body rewritten to the six-section bounded template with the full handover ledger (prior revisions are the archive, per Seat-post protocol: registry bodies grow without bound — bound the live body to current state, move shift narration out #7583).
5. Title / assignee / body updated in one stroke: ⏳ vacant, assignee cleared.
6. Timers cleared — this session's patrol trigger trig_01SdxKWd2oJfk9gJcAdrdbBk deleted; no orphan PR subscriptions (the two subscribe_pr_activity attempts this session both failed, so none were ever armed).
7. Final report delivered to the maintainer in chat.

Takeover: /pm-dispatch skills, read this post first. Next action for an incoming PM is #7754.

---

Generated by Claude Code

[os-zhuang]

Audit: seat takeover

• Incoming PM: os-zhuang, session session_01XoWxhuq3WNNaZUTs6kWdk8, 2026-08-11 ~12:5xZ.
• Staffing authority: the maintainer invoked /pm-dispatch skills in this session — the takeover path the stand-down comment above prescribes.
• Same-stroke triple: title → 🟢 os-zhuang, assignee → os-zhuang, body 当前 PM section → the triple above.
• State read at takeover: queue = pm-dispatch SKILL structural compression: 3050 → ~1400–1600 lines (PR₂ of batch #7631, maintainer-directed) #7754 only (premise re-verified against origin/main: SKILL.md = 3050 lines, last touch = PR pm-dispatch SKILL batch: eight ruled amendments — four-lens block, ADR-class skills, three-query board, queue criteria, five shift lessons, claim spelling, seat-post bounds, dispatch-contract rank (#7631) #7698 merged today); finding [finding] A skills-root PR was queued by a sibling session ~1h after the ADR-class ruling — a ruling recorded only on cards cannot bind sessions that haven't read it #7701 held for the triage seat; decision inbox empty; zero open PRs touch either skills root; hot-file queue on SKILL.md empty.
• First action: claim + dispatch pm-dispatch SKILL structural compression: 3050 → ~1400–1600 lines (PR₂ of batch #7631, maintainer-directed) #7754 — mode:cloud, model: claude-fable-5 (mandatory clause). Its PR is ADR-class: this seat reviews but never merges/enqueues/arms auto-merge; round reports will carry the "awaiting a human merge" line.

---

Generated by Claude Code

[huangyiirene]

跨席知会(devx 席 → skills 席):一个 skills-path PR 进入你的人工通道队列。

PR #7837 — [skills-path 人工合并] docs(skill): dogfood §2 — a drifted console dist now refuses to boot (#7752),分支 claude/issue-7752-skill-note,draft。

出处:#7752(console dist 漂移)的实现 PR #7783 原本把 .claude/skills/dogfood-verification/SKILL.md 的一小段和 CLI 代码混在一起。按维护者 2026-08-11「skills 更新与 ADR 同级,需人工审核」的裁决,devx 席未合并,已指令开发拆分:代码半留在 #7783(走 AI 正常流程,已武装),skills 半独立成 #7837 交你席位。

内容:#7783 让 os dev 在 console dist 的 objectui 戳记 ≠ .objectui-sha pin 时拒绝挂载 /_console(而非仅告警后照常服务)。SKILL.md §2 的配套改动是告诉跑手:此后 /_console/ 返回 404 意味着「用 pnpm objectui:build 重建」,不是「console 坏了」。同源的 RUNNER.md 那一行不在 skills 路径下,已留在 #7783 内。

时序建议(非指令,你席位自决):#7837 描述的是 #7783 落地后的行为。若 #7783 先合入而 #7837 待人工审核,期间跑手看到 404 会缺少这条解释 —— RUNNER.md 的 trap 行(在 #7783 内)已覆盖同一提示,所以缺口有限。

devx 席不再持有 #7837,不会对它做任何写操作。

---

Generated by Claude Code

[os-zhuang]

Audit: seat takeover

• Incoming PM: os-zhuang, session session_012Gg6rMAti8ZaueWb6BRsDn, 2026-08-12 ~01:5xZ.
• Staffing authority: the maintainer invoked /pm-dispatch skills in this session.
• Predecessor (same account, session session_01XoWxhuq3WNNaZUTs6kWdk8) left no stand-down comment — treated as an ungraceful session end; its ledger was inherited from the seat body and re-verified against live state before the body edit.
• Same-stroke triple: title unchanged (🟢 os-zhuang, same account), assignee unchanged, body 当前 PM section → the triple above.
• State re-verified at takeover: PR docs(pm-dispatch): cloud-card comms is event-driven by ruling — notes-21 verify-before-toggle, PR subscription as hard step #7794 open draft, CI fully green (Check Changeset re-ran to success 15:30Z) — awaiting a human merge. PR [skills-path 人工合并] docs(skill): dogfood §2 — a drifted console dist now refuses to boot (#7752) #7837 (devx transfer) reviewed this round against fix(cli): refuse to serve a drifted vendored console under os dev (#7752) #7783's MERGED diff, skip-changeset applied + failed job re-queued to success shape — awaiting a human merge. pm-dispatch SKILL: 5-hour usage-limit outages — queryable signals, cross-wall recovery timer (hourly cron, not send_later chains), wall-death playbook anchor #7765 / pm-dispatch SKILL Guardrails: dangling sentence fragment "someone else, never dispatch a needs-user-decision issue." — the bullet it belonged to is lost #7767 queued behind docs(pm-dispatch): cloud-card comms is event-driven by ruling — notes-21 verify-before-toggle, PR subscription as hard step #7794 (same-file serial). [finding] A skills-root PR was queued by a sibling session ~1h after the ADR-class ruling — a ruling recorded only on cards cannot bind sessions that haven't read it #7701 held for the triage seat. Decision inbox empty.
• Housekeeping executed: eight closed member cards' stale pm:queue stripped; read back clean (all 10 closed domain:skills cards carry the domain label only).
• Timers: predecessor heartbeat trig_01PA8om7ACs16icXhnXsaEXE already absent (delete → not-found); new cross-wall heartbeat trig_0139XfCW1b1oY5g6T6KeHkVg (hourly, self-bind) armed. docs(pm-dispatch): cloud-card comms is event-driven by ruling — notes-21 verify-before-toggle, PR subscription as hard step #7794 subscribed; [skills-path 人工合并] docs(skill): dogfood §2 — a drifted console dist now refuses to boot (#7752) #7837 subscription rejected by the platform — heartbeat covers it.

---

Generated by Claude Code