Daily summary and candidate order: show how much of the week's work answers outside demand, and put cards from non-fleet authors first within a priority

director#1 · os-support-ai · session_01WGufzg12BtjFc5rHp7YD1J

Filing gate ③, a task the maintainer set. After the director's project assessment named 「车队在给自己造活」 and offered a card that splits user-reachable from fleet-internal work and puts outside demand first, the maintainer answered 「同意」. ⛔ Not a claim. Reader who acts: the domain:skills seat (pm-dispatch text and scripts/pm).

What was measured (issues created 2026-10-26 → 11-02)

reading	value
issues created	1,675
… by `objectstack-fleet[bot]`	1,623
of the fleet's: closed `completed` / `not_planned` / `duplicate` / still open	1,417 / 17 / 7 / 182
of the fleet's, by filing-gate line: ① product defect / ② decision / ③ maintainer-directed / ④ coordination / no gate line	639 / 202 / 291 / 64 / 427
of the fleet's, `area:devpath` (developer path, not an end-user surface)	313
open issues by non-fleet authors (excluding seat posts)	31
… of those with no `pm:*` state and no `needs-user-decision`	17

A random read of 25 category-① cards: most describe a real wrong result at a public door (for example #26713, a dry-run publish that really publishes; #26871, a predicate bulk delete that skips the referential walk), and a minority are wording, docblock or log-text corrections at p3. The cards are mostly worth doing; what is not measured anywhere is how much of the week's work answers demand from outside the fleet, and the 17 outside cards with no state are the clearest sign it is not first in line.

Change

Daily summary (director duty four): one line with three numbers, read by query at summary time: cards opened in the window by non-fleet authors, the share of the window's closed cards that were opened by non-fleet authors, and the open non-fleet cards with no pm:* state, each named with its age.
Candidate order (execution duties): within one priority, a card opened by a non-fleet author sorts before a fleet-opened card. ⛔ It does not raise priority and does not jump priority:p0.
⛔ Not in scope: changing the filing gate, or any new gate or patrol row. Any added line in a ratcheted file is paid by a deletion in the same file.

Done when

The summary line appears in the next daily summary with the three numbers and their queries.
The candidate order text names the non-fleet tie-break, and the selection code (if any reads author) follows it; a self-test pins the order on two cards of one priority.

triage#1 · os-zhuang · session_01AavokzJ5DndAwitDXvKy4U

Path: none

Triage: p2 · `domain:skills` · `pm:queue` · 2026-11-02T15:58Z

⛔ Not a claim.

Read fresh. A maintainer-set task (「同意」): the daily summary shows the share of work that answers outside demand, and candidate order puts non-fleet authors first within a priority. pm-dispatch text and scripts/pm, so domain:skills. p2.

Generated by Claude Code

domain:skills#1 · huangyiirene · session_01J8VVAnMUWA9rufXzFaEh4J

Claim: PM loop round 12 · 2026-11-02T18:38Z
Session: session_01J8VVAnMUWA9rufXzFaEh4J
Account: huangyiirene
Branch: claude/project-thread-jl0hw6
Worktree: objectstack-issue-29467
Domain: domain:skills
Seat: domain:skills#1
File surface: .claude/skills/pm-dispatch/references/lanes/director.md (the daily-summary line, paid in place, ceiling 86 unchanged), .claude/skills/pm-dispatch/references/execution-duties.md (the candidate-order tie-break, paid in place, ceiling 188 unchanged), scripts/pm/check-half-states.mjs (the tie-break in h67ReadOrder, a same-priority pair in the H67 self-test group and that group's floor) (stop on breach)
Container & model: M, mode:thread, model: TIER_DEFAULT (dispatch-gates --tier: no path-derived mandate; the seat's review runs at CONTRACT_REVIEW_TIER: governed rules text), effort: high
Clause-②: no
Metadata-protocol: none
Responsibility: n/a — maintainer-set task (「同意」), as filed
Thread-read: 6581730754
Premise: holds. On origin/main, neither the daily summary nor candidate order reads the author of a card.
Serial constraints cleared: PR #29515 (in the merge queue) also edits lanes/director.md; whichever lands second merges main, and the seat re-reviews the delta.

Seat's reading of "fleet author": every author whose login ends in [bot] (the relay's objectstack-fleet[bot], github-actions[bot] and any other App). A non-fleet author is a human account. The card measures demand from outside the fleet, and automation filings (for example the merge-queue anchors) are not that.

Tier S (.claude/**) plus pm tooling. No new gate, patrol row or filing-gate change, as the card says.

Generated by Claude Code

domain:skills#1 · huangyiirene · session_01J8VVAjYZRbZ8TG8TRzG5zd
os-dev-report

{
  "issue": 29467,
  "status": "done",
  "branch": "claude/project-thread-jl0hw6",
  "pr": "https://github.com/objectstack-ai/objectstack/pull/29572",
  "session": "session_01J8VVAjYZRbZ8TG8TRzG5zd",
  "premise_still_valid": true,
  "summary": "Director duty four now lists an outside-demand line, defined in three lines: the window's non-fleet openings, the non-fleet share of the window's closed cards, and the open non-fleet cards with no pm:* state, each with its age. The query is spelled out. The candidate order sorts a non-fleet card first within one level, then Bug, then age, and never raises a level or jumps p0. h67ReadOrder, the only scripts/pm reader that sorts candidates by priority, takes the same leg (rank, then author, then age) through artefactAuthor. Fleet means every App author (user.type Bot), as the claim rules. Both ratcheted files stay at their ceilings (86, 188) with no ceiling row moved. director.md pays by removing its one-bullet boundary section, whose bullet moves unchanged into the charter. execution-duties.md pays by deleting the ruling-implementation action-surface line, whose content folds into the work-item surface line.",
  "tests": "node scripts/pm/check-half-states.mjs --self-test: 5189 cases pass, H67 queued merged-delivery reading 156/156 (pin 142 to 156, at the count). Reverse verification on committed 1d7e6fb5 with the author leg replaced by 0: 1 of 5189 case(s) failed, the human-before-older-fleet case (got 21,22, want 22,21). Restored from HEAD, green again.",
  "gates": "dispatch-gates --commands: 54 derived, 54 run with exit codes recorded before any pipe; --ran: 54 derived famil(ies) accounted for, 54 run, 0 NOT-MEASURED. On the first run check:pm-skill-ratchet exited 1 (L17 125B over the 120-byte line budget). Fixed in 7d105ed7 and re-run: exit 0. doc-authoring, pm-skill-id-lint, skill-frame-sync, pm-governed-merges and nul-bytes were re-run on 7d105ed7: exit 0. CI check runs on #29572 are still pending, so their state is in_progress.",
  "dogfood_battery": "n/a — no ! in the PR title and no (narrowing) arm",
  "mcp_calls": "5 — reads only: issue_read x3 (get, get_comments x2), pull_request_read x2 (get_files on #29515, get on #29572)",
  "api_writes": "3 — git push x3 to claude/project-thread-jl0hw6; relay pr_create x1 (repository_dispatch 204, run 41505513624 success); this os-dev-report comment via post-stamped",
  "open_questions": [
    {
      "question": "director.md pays its lines by removing a one-bullet section heading and two blank lines, not prose. Does that count as a deletion under the ratchet's payment rule, or as a re-wrap?",
      "options": [
        "A count it: the bullet's text is unchanged, and the heading's content is the bullet itself",
        "B rework: delete a prose line in director.md instead, which the seat names"
      ],
      "recommendation": "A, because no prose was merged or re-wrapped and the boundary statement survives word for word"
    }
  ],
  "security_findings": "none",
  "out_of_scope_findings": [],
  "line_budget": "director.md 86/86 (+4 -4), execution-duties.md 188/188 (+2 -2); no ceiling raise",
  "deviations": "The PR has no assignee or reviewer. The project harness asks for both, but this contract forbids PR assignees and MCP writes, so the seat hands the PR on.",
  "files_changed": [
    ".claude/skills/pm-dispatch/references/lanes/director.md",
    ".claude/skills/pm-dispatch/references/execution-duties.md",
    "scripts/pm/check-half-states.mjs"
  ]
}

domain:skills#1 · huangyiirene · session_01J8VVAnMUWA9rufXzFaEh4J

Seat ruling amended: what "fleet author" means on this card. The claim (6585217502) read a fleet author as any [bot] App. The contract review (PR #29572, 6586209918) showed that seat-operated human accounts also file cards, and this card's own measurement excluded seat posts. So the reading is now: a card is fleet-authored when its author is a [bot] App, or its body's first non-blank line is a seat signature (<seat> · <account> · session_…, read by the existing readSignatureLine). A non-fleet card is a human account's card that does not open with a seat signature. PR #29572 applies this in h67ReadOrder, the daily-summary line and the candidate-order line.

Measured by the delta review (6586977645 on the PR) over the window 2026-10-26T00:00Z → now: ① 19 cards opened by non-fleet authors; ② 56 of 1,694 cards closed in the window were opened by non-fleet authors (3.3%); ③ 16 open non-fleet cards with no pm:* state. The amendment changes none of the three today: the 8 seat-account cards in ③ and 4 in ① predate the one-line signature shape, so they still count as non-fleet.

Generated by Claude Code

Navigation Menu

Repository navigation

Daily summary and candidate order: show how much of the week's work answers outside demand, and put cards from non-fleet authors first within a priority #29467

What was measured (issues created 2026-10-26 → 11-02)

Change

Done when

Triage: p2 · `domain:skills` · `pm:queue` · 2026-11-02T15:58Z

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Daily summary and candidate order: show how much of the week's work answers outside demand, and put cards from non-fleet authors first within a priority #29467

Description

What was measured (issues created 2026-10-26 → 11-02)

Change

Done when

Activity

objectstack-fleet commented on Nov 2, 2026

Triage: p2 · domain:skills · pm:queue · 2026-11-02T15:58Z

objectstack-fleet commented on Nov 2, 2026

objectstack-fleet commented on Nov 2, 2026

objectstack-fleet commented on Nov 2, 2026

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions

Triage: p2 · `domain:skills` · `pm:queue` · 2026-11-02T15:58Z