Repository navigation
Commit d34533c
pm-dispatch(sign-off): a seat going off shift hands every unlanded claimed card back to pm:queue in one stroke; a needs_decision card goes to the decision box; stay-behind narrows to ACCEPTed PRs waiting on the queue or a human (#27371)
Fixes #27359
Clause-②: no
Metadata-protocol: none
## 维护者速读
- **改了什么**:席位下班(收班)时,手上每张还没落地的认领卡一笔交回 `pm:queue`(`Release:` +
`Resume-from:` + 清 assignee);dev 报了 `needs_decision` 的卡改进决策箱;留守只剩已
ACCEPT、正在等合并队列或等人批的 PR。`check-handover.mjs` 同步收窄:声明留守的卡必须有已 ACCEPT 且在等的
PR,dev 报 `needs_decision` 却还挂 `pm:dispatched` 的卡点名拒收。
- **为什么改**:下班的席位不再把卡锁在自己账号下。之前留守是默认安置,结果两张卡停在 `pm:dispatched`
没人跑,其中一张的待裁问题根本没到你手里,别的席位只能强制接管。
- **风险与回滚**:只改两页席位协议文字和一个本地只读检查脚本(不进 CI、不写
GitHub)。风险是检查比以前严:以前能过的「随手留守」现在被拒,要么交回、要么先把 PR 送进队列或送审。回滚 = revert 本 PR。
- **先看**:`scripts/pm/check-handover.mjs` 的 `handoverVerdict` 与
`acceptedPullWait`(收窄逻辑);`seat-post-protocol.md`〈活性、写侧刷新与收班〉新第 35–36
行与〈交接收尾清单〉第 2 步;`seat-lifecycle.md` 第 40、43 行。
- **两分钟验证**:`node scripts/pm/check-handover.mjs --self-test`,应看到 `✓
check-handover self-test: 66 cases pass`。
- **需要你做的**:无,席位落地(`.claude/**` 为 Tier S,席位达档复核后落地)。
## What changes
The maintainer's rule, verbatim: 「下班时卡片应该交回队列,这个应该立卡更新 skills」. The
card's items 1–4 are the spec.
**Protocol text** (both files stay at their line ceilings, every line at
most 120 bytes):
| file · line | before | bytes | after | bytes |
|---|---|---|---|---|
| seat-post-protocol.md L35 | 安置默认离任留守;留不了才热移交:在飞卡落收单注记 + 点名看护者。 | 94 |
收班默认交回(〈交接收尾清单〉第 2 步);留守只剩已 ACCEPT 待队列或待人批的 PR。 | 109 |
| seat-post-protocol.md L36 (new) | — | — | 留守写面 = 简报点名的这些
PR,残余定时器同界;留不了才热移交:收单注记 + 点名看护者。 | 119 |
| seat-post-protocol.md L38 → L39 | 收班序列:贴正文与标题刷新在先,简报评论是最终一笔;继任的释放判据 =
简报在。 | 110 | 收班序列不看维护者措辞:`check-handover.mjs` 退出 0 → 刷正文与标题 → 简报最终一笔。 |
114 |
| seat-post-protocol.md L40 (deleted) | 留守写面恰为简报点名的项:已 ACCEPT 待队列
PR、落地中的评审链载体;残余定时器同界。 | 117 | — (content moved to L35/L36; 评审链载体 leaves
stay-behind by item 3) | — |
| seat-post-protocol.md L47 (checklist step 2) | 在手工件逐件处置:本席认领的卡落地,或
`Release:` 与清 assignee 同笔释放,转成真实状态。 | 118 | 认领卡落地,或同笔
`Release:`+`Resume-from:`+清 assignee 交回 `pm:queue`;`needs_decision`
进决策箱。 | 118 |
| seat-lifecycle.md L40 | 收班 = 状态 flush + 在飞显式安置,不是在飞归零;派发口径失效,简报 ⛔
不列作继承裁决。 | 118 | 收班 = 状态 flush + 在飞默认交回队列、不等落地;派发口径失效,简报 ⛔ 不列作继承裁决。 |
120 |
| seat-lifecycle.md L43 | 离任留守只写简报点名的项,⛔ 不读队列、不新派发、不碰未点名卡。 | 94 |
离任留守只写点名的已 ACCEPT 待队列或待人批 PR,⛔ 不读队列、不新派发、不碰未点名卡。 | 120 |
Bytes are the full source line, list marker included (the gate's
measure); the table shows the text after the marker. Line counts:
seat-post-protocol.md 88 → 88 (ceiling 88), seat-lifecycle.md 96 → 96
(ceiling 96). The new L36 is paid by deleting old L40, whose content
moved into L35/L36. One clause leaves old L38: 「继任的释放判据 = 简报在」. Its home
is seat-lifecycle.md 〈开轮互斥〉: 「收班简报是前任不再写的显式声明,是释放标记不是锁:简报即最新事件 ⇒ 立即坐席。」
- Item 1 (hand-back is the 收班 default): L35 + checklist step 2. The
stroke is `Release:`, `Resume-from:`, the assignee cleared, and the card
back to `pm:queue`.
- Item 2 (a `needs_decision` card goes to the decision box, ⛔ not
`pm:queue`): checklist step 2, and the new check below.
- Item 3 (stay-behind narrows to ACCEPTed PRs waiting on the queue or a
human): L35/L36, seat-lifecycle L43, and the check.
- Item 4 (the brief and the `check-handover.mjs` run are the last acts,
whatever words end the shift): L39.
**`scripts/pm/check-handover.mjs`** (a local refusal tool; no workflow
runs it):
- **(a′) `[needs-decision-unreleased]`.** A held card is refused by this
name when its newest `os-dev-report` is newer than the claim and
declares `"status": "needs_decision"`, and the card is not
`needs-user-decision`. The remedy names the decision box, ⛔ not
`pm:queue`. It is refused even when the card is declared as a
stay-behind.
- **Stay-behind narrowed.** A card listed in `--stay-behind` and
declared on the vacant body is reported as `[stay-behind]` only when an
open pull that delivers it meets two conditions. First, the pull's
newest `## 席位核对` record is an ACCEPT. Second, the pull waits: it is in
the merge queue (newest queue event on its timeline is the add),
auto-merge is armed, or it awaits a human approval
(`needs-user-decision` on the pull). Otherwise the card stays
`[claim-unreleased]`, and the refusal names the failed condition for
each pull.
- **No reader is restated.** `latestDevReport` /
`devReportPayloadCandidates` / `deliveryEvidence` come from
`check-half-states.mjs`, `SEAT_RECORD_MARKER` / `seatRecordAccepts` from
`seat-record.mjs`, and `enqueuerFromTimeline` from
`check-governed-queue-guard.mjs`. The only new reading is the report's
`status` field.
- **Reads are bought only for a declared tail.** That is one paged
listing of open pulls, then each delivering pull's thread, and its
timeline only when the label and `auto_merge` did not answer. Every
request is a GET.
## Tests
`node scripts/pm/check-handover.mjs --self-test`: **46 → 66 cases**, all
pass (handshake intact). Changed rows: ARM 1 now requires a waiting
ACCEPTed pull, and the decision-not-excused row uses a waiting tail so
it still isolates the decision refusal. New rows cover the narrowing
(not-waiting, unread), (a′) (measured shape, older-than-claim,
newest-decides, already-in-inbox, refused-by-name), `acceptedPullWait`
(queue, approval label, armed, needs-timeline, dequeued, no-accept,
REWORK-after-ACCEPT, ACCEPT-after-REWORK), the shortfall text, delivery
evidence, and two structural pins.
<details><summary>Ablation, four legs through
scripts/ablation-replace.mjs, each restored and hash-proven (HEAD
d457c56, blob 2557438)</summary>
- leg 1: `else if (twice && h.wait)` changed to `else if (twice)`. 2 of
66 failed: `SIGN-OFF NARROWING`, `absence never reads as waiting`.
- leg 2: the `needs_decision` status test changed to `return null`. 1 of
66 failed: `(a′) THE MEASURED SHAPE`.
- leg 3: the ACCEPT requirement changed to record existence. 1 of 66
failed: `a REWORK newer than the ACCEPT, is no ACCEPT`.
- leg 4: the queue reading changed to `if (true)`. 1 of 66 failed: `the
newest queue event is a removal, or there is none`.
- After each leg: blob after restore == blob at HEAD (2557438), and
`git diff HEAD` is empty. The unmutated control run: 66 cases pass.
</details>
<details><summary>Live reads (GET only, this container)</summary>
- `--probe`: transport prerequisite met.
- Departed `domain:server` seat (`session_01R1q2xdtYgU6FKrZ7LgjPCC`,
`os-sales`, since 2026-10-27): exit 0, clear. The measured instance has
moved on. #25805 is now `needs-user-decision`, and #23181's newest claim
names another session, so it is listed, not refused.
- This seat with `--stay-behind=#27359` and a vacant probe body: exit 2.
#27359 was refused with "declared a stay-behind on both, but no open
pull request delivers it". Coverage: `(c) 1 declared stay-behind
card(s): 14 open pull(s) listed`.
</details>
## Acceptance notes
- `落地中的评审链载体` leaves the stay-behind write surface. Card item 3 narrows
stay-behind to ACCEPTed PRs only.
- Checklist step 2 loses 「转成真实状态」. The ruled targets replace it:
`pm:queue`, or the decision box for `needs_decision`. Step 4 still
carries `pm:blocked` + `Blocked-by:` / `Restart-when:` for ledger
obligations.
- The stay-behind reading asks whether the pull's newest seat record is
an ACCEPT. It does not ask whether that record was made for the current
head; H69 patrols that half on Tier H.
- (a′) reads held claims only. A card released to `pm:queue` while its
dev's question is open leaves that population, and H52 (open questions
unrouted) still reports it.
- Queue membership goes through `enqueuerFromTimeline`. An add event
whose actor has no login reads as not queued, which is the refusing
direction.
- The coverage line prints `(c)` between `(a)` and `(b)`, because the
stay-behind reads follow the held-card reads. Cosmetic only.
- The file's self-test keeps its pre-existing shape: a handshake and no
battery floor. Adding a floor is outside this card.
- This PR carries no assignee. That follows the seat's ruling for
seat-driven drafts, while `os-dev.md` on main still says to send one.
## Verification
Final head `d457c56d`. Derivation: `node scripts/pm/dispatch-gates.mjs
--commands --repo objectstack-ai/objectstack` gives 36 commands. All 36
exited 0 locally. `check:doc-formula-expressions` first exited 3
(PREREQUISITE NOT MET: `@objectstack/lint` was not built). It exited 0
after `turbo run build --filter=@objectstack/formula
--filter=@objectstack/lint` under the verify lock. `--ran`
reconciliation: `36 derived famil(ies) accounted for — 36 run, 0
NOT-MEASURED (a DERIVED zero — all 36 recorded an exit code and none of
them is 3)`.
<details><summary>Gate readings</summary>
- `pnpm check:pm-skill-ratchet`: seat-post-protocol.md is 88 lines
(ceiling 88; headroom 0). seat-lifecycle.md is 96 lines (ceiling 96;
headroom 0). Self-test: 164 cases pass.
- `pnpm check:pm-skill-id-lint`: 105 file(s) clean.
- `pnpm check:pm-post-stamped`: 735 cases pass across 26 batteries. Its
self-test reads `check-handover.mjs`'s `carriesFacetMarker`, which is
unchanged.
- `pnpm check:doc-authoring`, `pnpm check:nul-bytes`, `node
scripts/check-self-test-wired.mjs` (+ `--self-test`), `pnpm
check:pm-governed-merges`, `pnpm check:skill-frame-sync`, and the
remaining 26 derived commands: exit 0.
- `pnpm exec eslint --no-inline-config scripts/pm/check-handover.mjs`:
exit 0.
- `check:pm-governed-prose` and `check:pm-half-states` are not derived
for this diff, so they did not run locally; CI runs the lint farm.
</details>
---
_Generated by [Claude
Code](https://claude.ai/code/session_01QSeKYmhU5KEj1nukYhp1Tg)_
---------
Co-authored-by: Claude <noreply@anthropic.com>1 parent 6e0e7d7 commit d34533c
3 files changed
Lines changed: 236 additions & 33 deletions
File tree
- .claude/skills/pm-dispatch/references
- scripts/pm
| Original file line number | Diff line number | Diff line change | |
|---|---|---|---|
| |||
37 | 37 | | |
38 | 38 | | |
39 | 39 | | |
40 | | - | |
| 40 | + | |
41 | 41 | | |
42 | 42 | | |
43 | | - | |
| 43 | + | |
44 | 44 | | |
45 | 45 | | |
46 | 46 | | |
| |||
| Original file line number | Diff line number | Diff line change | |
|---|---|---|---|
| |||
32 | 32 | | |
33 | 33 | | |
34 | 34 | | |
35 | | - | |
| 35 | + | |
| 36 | + | |
36 | 37 | | |
37 | 38 | | |
38 | | - | |
| 39 | + | |
39 | 40 | | |
40 | | - | |
41 | 41 | | |
42 | 42 | | |
43 | 43 | | |
44 | 44 | | |
45 | 45 | | |
46 | 46 | | |
47 | | - | |
| 47 | + | |
48 | 48 | | |
49 | 49 | | |
50 | 50 | | |
| |||
0 commit comments