docs(pm-dispatch): release inventory is two queries, and the console bump linkage is NOT covered at the cut (#6906)

[os-project-manager]

Fixes #6906

Docs-only change to .claude/skills/pm-dispatch/SKILL.md (internal agent tooling — not a published skill). Two deliverables.

Deliverable 1 — the release inventory is stated as two queries

The board's consumer documentation previously described the maintainer's release inventory as one query (label:target:<major> is:open). That reading silently drops the entire frontend half, and a short read looks identical to "the board is clear". All three consumer sites are now updated coherently:

• the signal table row for label:target:<major> — "每个 backlog 恰好一个生产者", step 9 counts the sum of both boards;
• 「消费者三处」 — spells out repo:objectstack-ai/objectstack … and repo:objectstack-ai/objectui …, names the org-level search (org:objectstack-ai …) as the equivalent one-query form, and notes that the org form also sweeps cloud, where a hit is by construction a mislabel signal rather than a board item;
• step 9's release blockers health metric — 归零 now explicitly means both boards empty.

Written against the amended rule 1 (#7165, file-at-destination): a card whose landing site is objectui lives in objectui, and its target:<major> is produced by the objectui whole-repo seat in its own repo. The second query is therefore a structural consequence of the ownership model, not a migration leftover that some future cleanup retires.

Deliverable 2 — verified, not assumed: the cut moment is NOT covered

The dispatch asked for this half to be verified rather than assumed, and the verification came back negative. The #6162 window-close chore does not cover the release-cut moment, for two independent reasons (either sufficient alone):

1. its trigger is a window-settled predicate (.objectui-sha lags objectui main and objectui's merge queue is empty) — the second conjunct is false exactly when a cut is most inconvenient, and nothing on the release path re-evaluates it;
2. the card it files is labelled pm:queue and therefore carries no target:<major>, putting it outside both inventory queries by construction. Live specimen measured 2026-08-10: chore: bump the console pin to pick up objectui#4071 (approvals:inbox component ref) #7268, labels = pm:queue only.

The section now records that measured negative in place of prose that read as covered, and states what is guaranteed so the gap is not overstated: check:objectui-pin-fresh is required on the release PR and enforcing on the publish path (#3340 / #6170), so a stale pin cannot ship. What is missing is the process half — the card, or the explicit waiver, at the moment the light turns red.

The gap is filed as #7275 (unassigned, three candidate dispositions laid out, no decision taken — option B would create a second target:<major> producer and needs a maintainer ruling). Per #6906's own body, a real gap found by item 2 gets its own card rather than growing that issue.

Verification

• No stale "one query" wording remains in the file (the single surviving 「那一条查询」 refers to the objectui one of the two, which is correct).
• Docs-only: no package code, no content/docs/releases/ edit, no changeset (this file ships in no published package).

Note for the maintainer

This PR edits the operating skill of the PM seat that is reviewing it. Left as a draft, not queued — the review above is mine, but the merge decision is yours.

---

Generated by Claude Code

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

1 Skipped Deployment

Project	Deployment	Actions	Updated (UTC)
objectstack	Ignored		Aug 10, 2026 5:11am