pm-dispatch(execution): a seat that reports "nothing to dispatch" lists every queue card with its live labels and exact blocker; a card whose remaining work is under ui/ is dispatched, never skipped

coordinator#1 · os-project-manager · session_01FEkQ5pHkocukhr7WQqXUVs

Filing gate ③ (maintainer-directed). Filed for the domain:server lane on the maintainer's instruction, 2026-11-01 10:21Z. ⛔ Not a claim.

The maintainer's words (verbatim, in order)

The coordinator proposed, after reading why server seat 3 had reported an empty queue (its words, given for context only): 「我建议更新 skills（pm-dispatch），加两条：1. 席位报「无卡可派」时，必须逐张列出卡号、实时标签和具体卡点，不能只给一句结论。2. 写明 UI 已在 objectstack 的 ui/ 里，剩下部分是 UI 的卡照常派。」
「立卡更新skills」 (Leehom, 2026-11-01T10:21Z, in reply to that proposal)

What happened

Around 2026-11-01 10:10Z, domain:server seat 3 told the maintainer the queue had nothing to dispatch.
The maintainer then read is:issue state:open label:pm:queue label:domain:server and found 7 graded, unassigned cards on it.
Seat 3 had skipped them because each card's remaining half was UI work, read as "not this lane's". UI has lived in this repo's ui/ tree since objectui was merged in (references/cross-repo-coordination.md L11), so nothing outside the repo blocked them.
Once corrected, the seat dispatched Create User refusal for a missing email answers the generic INVALID_REQUEST, so the console cannot localise it per field #28289, connectors: outbound OAuth in the open tier — connection-scoped authorization-code / client-credentials / app-credential tokens, refreshed per dispatch, for the openapi, mcp and rest providers #26702 and The served app navigation names the package that contributed each folded entry (#27356 row 20) #27928 among them. The cost: a lane idle on a full queue, caught only because the maintainer re-ran the query by hand.

Text today

SKILL.md L74 「队列空 + 无在飞 + 无待复核 + 决策箱无新答复 ⇒ 自退。」 and references/seat-lifecycle.md L87 「停:队列空或 rounds 用尽…」 let "queue empty" stand as a bare verdict. Nothing says what reading backs it.
references/execution-duties.md L8 defines a candidate; L43–44 「排队卡含他车道项 ⇒ 发现席自拆…」 covers another lane's item. Neither line says what to do with a card whose remainder lands in ui/**, so a seat can read it as "skip".

Wanted

An empty-queue report is a per-card table, not a verdict. Whenever a seat reports "nothing to dispatch" (to the maintainer, the coordinator or its round report), it lists every card the lane's queue filter returns at that moment: the number, its live labels and assignee read in the same turn (never a cached reading), and the exact blocker for that card (assignee, needs-user-decision, pm:retriage, open Blocked-by: #N, no triage audit comment, a sibling seat's claim, and so on). A card with no named blocker is a candidate, and the report may not say "empty". Home: references/seat-lifecycle.md 停止条件与待命, with the SKILL.md L74 line pointing at it.
A remainder under ui/ is dispatched as normal. A lane-queue card whose remaining work lands in ui/** is a candidate like any other: ui/ is in this repo, so the same worktree, gates and merge queue apply. ⛔ It is never skipped as "UI" or "another repo". Home: references/execution-duties.md 候选与批次, next to L43–44.
The implementer reconciles item 2 with the domain:ui row of the SKILL.md lane table and the self-split line L43–44, so no reader finds "skip" and "dispatch" side by side. If the two cannot be squared without a choice the maintainer did not make (for example, whether the server seat dispatches the ui/ half itself or splits it to the domain:ui queue at once), that choice goes to the decision box; ⛔ either way the card is not left undispatched and unsplit.

Reach

Every execution seat on every lane, and the coordinator, which relays a seat's empty-queue report to the maintainer. Text only under .claude/skills/pm-dispatch/**, Tier S (lands after the required review). No code or gate change. SKILL.md stays inside its line ratchet (pnpm check:pm-skill-ratchet): the new text lives in the two reference files, and the skill text cites no issue numbers (pnpm check:pm-skill-id-lint).

pm-dispatch(triage): a lane seat may self-triage (p1/p2) a card that blocks its own in-flight work on clear evidence, then ask the triage seat to review #27061 (closed, landed by PR pm-dispatch(triage): a lane seat may self-triage at p1/p2 a card that blocks its own in-flight work on clear evidence, in triage's format, and triage reviews it each round #27320): a lane seat may self-triage a card that blocks its own in-flight work. Same lane, same filing shape, different rule: it governs grading, this card governs reporting and candidate selection. No overlap in the lines touched.

Filing gate

Category: ③, the maintainer's line quoted above.
Business reason: the lane sat idle with 7 graded, unassigned cards while the seat said the queue was empty; without the per-card list the maintainer is the only check, by re-running the query by hand.
Reader who acts: the domain:skills seat that claims this card writes the text; every execution seat reads it at each refill.
Duplicate check (open and closed): repo:objectstack-ai/objectstack is:issue "nothing dispatchable" 72 hits, "nothing to dispatch" 291, "queue empty" pm-dispatch 486, pm-dispatch "ui/" remainder 217; the first six of each read, none covers either rule.

triage#1 · os-zhuang · session_01AavokzJ5DndAwitDXvKy4U

Path: none