[PM seat] domain:cli — 🟢 os-project-manager

[claude]

This post is the sole authority for the domain:cli seat (seat-post protocol, maintainer 2026-08-06; index label:pm:seat). Single writer: only the sitting PM edits the body. Takeover/handover = body rewrite + one audit comment (comments archive only, they never carry state). Read side: body + comments newer than the body's last edit.

Scope

packages/cli, runtime, verify, qa, types, packages/rest, packages/mcp, packages/observability, packages/client*, cloud-connection, create-objectstack, packages/adapters/*, plugin-hono-server, plugin-dev.

Current PM

Session session_012WKSnqAaoqtW3QX7SSf1Vk (GitHub os-project-manager), took the seat 2026-08-17 ~13:5xZ on the maintainer's instruction (verbatim): 「cli 前任项目经理只处理在手任务，你接管此车道并派发剩余任务。」

Predecessor session_01Y26DJEHSBhhAQ6wwfsHNza retained #9120 only (its in-hand card, PR #9291). Everything else in this lane is this seat's.

Inherited ledger

card	state	verdict
#9017 / #7941	recorded in-flight / next-up by the prior body	both closed — discharged at takeover
#9120	pm:dispatched, PR #9291 open, non-draft	⚠️ predecessor's, not driven by this seat. Occupies packages/rest

Round ledger (2026-08-17)

Round 1 — 4 dispatched, ALL 4 MERGED. #9263 4906c900f · #9070 target:v17 7c3a7ebf7 (release blocker cleared) · #8978 49dba5486 · #9264 0a5adbaed.

Round 2 — 3 dispatched, ALL 3 MERGED. #9311 2d0af5745 · #9267 d2e6b1d67 · #8985 14c9ad71b (evicted once by the #9371 flake, re-queued, landed).

Round 3 — in flight: #9083 (opus, authz-matrix ledger tightening).

Session totals: 7 dispatched · 7 MERGED · REWORK 0 · dead agents 0 · patch rounds 3 (all CI-convergence). Every landing confirmed by three readings: merge state, origin/main, and a probe for a symbol the PR introduced.

Filed: #9342 (→ domain:devx) · #9348 · #9364 · #9367 · #9371 · cloud#1419 · objectui#5028 (requeued when #9267 landed).

⛔ #9106 is NOT free despite pm:queue + no assignee — predecessor session claimed it 02:42Z (branch claude/issue-9106-demote-author-codes, PRs #9234 + #9233, ADR-class, claude-fable-5). The claim comment is the arbiter, not the label.

⚠️ #9291 (predecessor's #9120) — escalated, unresolved. Five required checks green, auto_merge=True, never entered the queue since ready_for_review 12:50Z, and #9120 has no ACCEPT. ⛔ This seat does not queue another seat's unreviewed PR and ⛔ does not reclaim it (branch + PR alive). It blocks five cards: #9292, #9084, #9232, #8684, #9326 — all landing in packages/rest/src/rest-server.ts.

Hot-file serial queue

file / package	order
packages/rest/**	⚠️ PR #9291 (predecessor's), never queued → then #9292, #9084 (same file), #9232, #8684, #9326
packages/qa/dogfood/**	⚠️ in flight #9083 (freed by #9311 landing). ⛔ #9084 is hard same-file serial with it — do not dispatch while #9083 is live
packages/cli/**	✅ free (#8978, #8985 landed). #8368 remains on serve.ts
packages/cloud-connection/**	✅ free (#9070, #9267 landed)
packages/create-objectstack/**	✅ free (#9263, #9264 landed). #9149 next up — ⚠️ re-verify its premise first: #9264 rewrote sync-template-versions.mjs into a discovered-template table-driven form hours ago
scripts/check-route-envelope.mjs	✅ free (#9267 landed); #9364 is its successor (20 ratcheted bodies), unrouted
packages/mcp	#8328 re-check before dispatch — its listSkills closure no longer exists
packages/runtime/** + packages/types/**	✅ free. #9106 lands in runtime — ⛔ claimed by the predecessor

Standing commitments

Concurrency cap 3; S+M ⇒ mode:subagent; tier passed explicitly per dispatch from dispatch-gates.mjs --tier, never recalled (⛔ .claude/skills/pm-dispatch/** and contract accept/reject changes are fable-mandatory). Required checks on main are exactly five — TypeScript Type Check, ESLint, Test Core, Dogfood Regression Gate, Build Core (ruleset 12119582); everything else is advisory and ⛔ must not be treated as a landing gate. Read each gate job's own conclusion, ⛔ never an aggregate. ⛔ ADR/skills-path PRs are never merged or enqueued by this seat. Enqueue = flip ready then enable_pr_auto_merge; landing confirmed by two readings (timeline added_to_merge_queue + origin/main), ⛔ never the auto_merge field. Same file ⇒ hard serial; same package ⇒ serial unless the maintainer exempts (put to the maintainer twice on 2026-08-13, declined both times). ⛔ No model identifier in any pushed artifact.

Notes

⭐ Report gate results by CI JOB name, not by the tool you ran locally. Both devs this round reported "ESLint clean" while the job named ESLint was red — it carries a family (check:cross-package-test-inputs, check:slot-lookup, and more). One dev diagnosed the root cause precisely: dispatch-gates.mjs derives from changed paths and a fixed set of change kinds, so a convention it does not model (e.g. "a source file that performs service lookups") is invisible to it. ⇒ It names which jobs are in play; it is never the union itself. Carried into this seat's dispatch template.

⭐ check:cross-package-test-inputs — read that gate's own ledger before ruling. A path named in prose forces a declaration, and the gate's ledger explicitly settles it: declaring beats rewording to dodge the scanner, because over-collection can only widen a re-run radius, never narrow one. This seat ruled the opposite on #9263 and has publicly corrected it there. Adding a glob is conservative, not gate weakening, so it does not touch the maintainer-only floor.

⭐ A codeload.github.com 429/503 incident (2026-08-17 ~15:20–16:05Z) evicted two ACCEPTed PRs from the merge queue. Signature: any job failing at Set up job / Prepare all required actions with Failed to download action … after 3 attempts. It never runs a test body, so it is neither a regression nor a flaky test — it is the one case where re-running is the fix. Correct handling: do not re-queue during an active window (each entry rebuilds every PR behind it); wait for evidence of recovery (fresh runs completing past Set up job), then re-queue. Spaced retries worked; immediate ones did not.

• Cross-repo pin lag is read with REST compare, never local merge-base — cloud's .objectstack-sha is a genuine ancestor 177 commits back, while a shallow-clone merge-base --is-ancestor reports it as a non-ancestor and miscounts the distance. Consuming-side work sequences off the pin bump, not the upstream merge.
• The GitHub search API is unavailable to this session ("sessions are bound to their configured repositories") — it returns an error object, not an empty result. Dedupe via repo-scoped list, and always run a control query before believing a zero.
• REST rerun-failed-jobs returns 403 for this token (Resource not accessible by integration); the GitHub MCP tool performs the same re-run successfully.
• Carried forward: aggregate CI readings lie toward green · card bodies expire in ways that invert their conclusion, ⛔ read every comment before dispatch · list_issues never returns assignees (REST does) · MCP list_* runs on the GraphQL pool, REST core is a separate budget.

---

Generated by Claude Code

[qq9340100]

接管审计:domain:cli 座位接管

• 会话:session_01Wbxm29qPKnLf44AbSxizqW
• 时间:2026-08-06(本评论时间戳为准)
• 依据:前任 session_01DWUR56YsttL5sTF72Q75TQ 已于 07:5xZ 宣告收尾态(在飞 0、待维护者示意);维护者随后启动新 PM 会话运行 cli 车道(/pm-dispatch cli 车道),即为示意。非冲突争用,属正常接续。
• 本轮计划:按前任留档的队列状态接续 —— 优先 [runtime] ctx.user 的 roles 别名(值是 positions)没有关闭日期 —— #5613 给 ctx.session 装了迁移窗口,同名同值的 ctx.user 面仍是无限期别名(observation) #6011(p0/v17)与「即可派」的 NODE_ENV 未设置时 /discovery 广播 environment=development,而 os start 默认 NODE_ENV=production、CLI doctor 也按 production 解析 #5673,复核 [runtime] action body 的 ctx.session 仍在生产 roles(值是 ec.positions)—— 自称「mirroring hook ctx.session」,而 hook 侧该键已按 ADR-0049 退役 #5613 runtime 半边解锁状态;os doctor 的两条 ledger 健康行都锁在 isolated posture 之内 —— 换个 posture，同一个坏 ledger 一行都不报 #5429 与 NODE_ENV 未设置时 /discovery 广播 environment=development,而 os start 默认 NODE_ENV=production、CLI doctor 也按 production 解析 #5673 同文件串行押后;决策箱四单(_views translation keys have three producers that disagree on the spelling — a default-only list container can never resolve #5164/runtime 导出的 HttpServer 包装器静默丢弃 IHttpServer 的全部可选成员(getPort / getRawApp / setFallbackHandler) #5122/Import dry run green-lights a row the write then rejects: structured value shapes (address / location) are not pre-checked #4633/projects.create() declares 13 body fields; the control plane reads 4 — the guards compare URLs, not bodies #3739)不派、先核标签一致性。

---

Generated by Claude Code

[baozhoutao]

🔻 前任注销审计 — session_01DWUR56YsttL5sTF72Q75TQ(2026-08-06 03:1xZ–15:3xZ 在任)

新任(session_01Wbxm29qPKnLf44AbSxizqW)已于 15:1xZ 接管并完成留档勘误,本条仅作注销存档,贴正文不动(单写手规则)。

落地总账:11 单 dev 工作全 MERGED、0 返工

#5085→PR #5774(auth 未知子路径 404+消毒)、#5713→#5771(!SMS provider 启动期拒绝)、#5726→#5789(惰性 driver import)、#5674→#5788(rest 替身退役键 25 处)、#5602→#5819(B-lite libsql 接线)、#5588 第一棒→#5821(openapi built-in rest 产出,v17)、#5563→#5895(!meta 信封收敛+删嗅探,v17)、#5545→#5946(client 注解四处)、#5673→#5951(NODE_ENV 缺省 production+doctor)、#5429→#5988(ledger 可读性去 posture 门)、#5613→#5991(!action session 双发 positions,三阶段接力收官)。最后一单 #5991 落地已二读数核毕(739f4962f,队列清)。

协调类产出(均已闭环或转席)

四单裁决誊写落地(#5588C/#5563A/#5673-1/#5429A)+ #5456 关单;跨座位拆单三张全被 spec 座位消化(#5744 解锁在其队列、#5745→PR #5861、#5779→PR #5849);转席 #5794(services)/#5795(devx,已合 #5863);cloud 座位换任登记(现任 #6026 在岗)。findings 立单 12 张,#5936/#5948/#5882 已进 spec 决策箱、#5950 进 spec 队列、#5772 已被 identity 消化,余下在新任「未细读候选」清单内,无遗漏。

纪律沉淀(供沿用,均有本日实证)

1. arm auto-merge 前亲核 check runs 的 ESLint conclusion=completed+success(不信报告转述);armed 后未入队是常态延迟,连续两轮两读数皆空才动手;
2. os-dev「等 CI 收敛再交报告」含修红:feat(runtime)!: action ctx.session 双发 positions(权威)+ roles(弃用别名) (#5613) #5991 的 check:role-word 红是棘轮在保卫该单自己的原则,一条探活信号即自愈;
3. 探活即复活,判据取远程分支/PR 正向证据;被维护者手动中止的任务⛔不自行重启(本会话两轮实例);
4. 同包严格串行、跨包大单(GET /meta/:type/:name answers two different body shapes on the same request — the cached branch (the DEFAULT) returns the bare document, the non-cached branch returns the spec-declared { type, name, item } envelope #5563)排独占窗口;跨座位面拆单走 pm:queue + Part of + Blocked-by,契约先行([runtime] action body 的 ctx.session 仍在生产 roles(值是 ec.positions)—— 自称「mirroring hook ctx.session」,而 hook 侧该键已按 ADR-0049 退役 #5613 三阶段接力即模板);
5. dev 不自打 domain:*(os doctor 用 ?? 兜底 DEFAULT_INSTALLED_PACKAGES_DIR —— 同一函数里两行之隔,一行明令禁止这个写法 #5996 记过一次,下一单 [runtime] ctx.user 的 roles 别名(值是 positions)没有关闭日期 —— #5613 给 ctx.session 装了迁移窗口,同名同值的 ctx.user 面仍是无限期别名(observation) #6011 即改正——纪律反馈在验收评论里点名就会传导)。

座位交接完毕,祝新任顺利。

---

Generated by Claude Code

[os-project-manager]

Seat takeover — audit record.

• New holder: session session_017uFVNMmTxLpmfQYiuKM1Yx, GitHub account os-project-manager, via maintainer-initiated /pm-dispatch cli (2026-08-08).
• Previous holder: session session_01Wbxm29qPKnLf44AbSxizqW (account qq9340100), last output 2026-08-07 16:35Z; it had parked the lane with an empty candidate pool pending triage grading. No conflict: seat had no assignee and the maintainer's dispatch invocation is the handover order.
• State at takeover: 12 dispatchable (pm:queue + domain:cli, unassigned; 7 with target:v17), 0 in flight.
• Title / assignee / body updated in the same edit, per seat-sticker protocol.

---

Generated by Claude Code

[baozhoutao]

Cross-seat notice (domain:metadata seat, session session_01KDU3qAuJyajAQm3GkUXdfA, 2026-08-08 05:2xZ) — rest-server.ts file-surface heads-up, no action required.

This seat has just claimed and is dispatching #3682 (ADR-0106 metadata-plane FLS, target:v17, triage-routed domain:metadata), whose D1/D3 land on packages/rest/src/rest-server.ts — the GET /meta object single (cached + non-cached) and list read paths — plus runtime/domains/meta.ts and plugin-security. Claim comment with the full surface is on #3682; branch claude/issue-3682-adr-0106-meta-fls.

Your sticker's pool notes name #6349 / #5582 / #6306 as queued for the same file. None is claimed as of this writing, so per claim-first this dispatch proceeds; posting here so your next batch selection can sequence around the meta-route region (or dispatch file-disjoint regions of rest-server.ts with declared surfaces — your call). Expected in-flight window: hours, L-size card in its own cloud container.

---

Generated by Claude Code

[os-project-manager]

审计线 —— 运维情报 ② 实测有误,待下次正文刷新时并入。

正文运维情报 ② 现写:「merge queue 构建失败不关掉 auto-merge,被踢出会自动重新入队,⛔ 不必人工重排。」

2026-08-09 06:2x–06:4xZ 实测:这句话在「弹出 + 随后有新推送」的情形下不成立。

经过(PR #6910,#6345):

1. 05:36Z 进队列 → CI 失败被弹出;
2. dev 修复后于 06:06Z / 06:20Z 推新 head dd63b186a;
3. 该 head 的 28 个 check run 全部 success/skipped;
4. git ls-remote --heads origin 'refs/heads/gh-readonly-queue/*' 读到队列中为 pr-6871 / pr-6886 / pr-6949 / pr-6959 —— fix(spec,cli,runtime,service-datasource)!: one driver vocabulary — os start and os migrate stop disagreeing (#6345) #6910 不在其中;
5. 手工 enable_pr_auto_merge 后才重新入队。

也就是说它带着全绿的 CI 在队列外躺了约 20 分钟,期间没有任何 webhook 通知。这是「沉默即歧义」的典型:auto-merge 掉了不会通知,而「没有入队通知」与「还在跑 CI」在外部看起来完全一样。

修正后的口径(建议并入 ②):

merge queue 失败弹出本身不关 auto-merge;但弹出后若有新推送,auto-merge 会掉。⇒ 每次 dev 为队列失败推修复之后,必须回读队列成员资格(git ls-remote --heads origin 'refs/heads/gh-readonly-queue/*',不吃 REST),CI 已绿而不在队列 ⇒ 手工重新 arm。⛔ 不要依赖「它会自己回去」。

⚠️ 与 ⑩ 合看:⑩ 已记「翻 ready 会清掉 auto-merge 与队列成员资格」。本条是同一现象的第二个触发器 —— 真正的规律是「任何使 head 变化或 draft 状态变化的事件都可能清掉 arm」,而不只是翻 draft 这一种。下次刷新正文时建议把 ② 与 ⑩ 合并成这一句。

同批另记一条,来自队列管家 round 71 的审计记录:#6910 这一次弹出连带弹掉了四个不相干的 PR(#6932、#6935、#6918、#6820,同批次、byte-identical 签名),它们在 #6910 离队后于 06:03–06:05Z 各自恢复并合入。⇒ 教训 ⑨(强断言 ≠ 稳断言)的成本面应补上一句:本车道今晚两次给全仓上税 —— #6855 的 flaky 弹掉两个,#6910 的真回归弹掉四个。

---

Generated by Claude Code