Skip to content

fleet-write: no relay op for PUT /pulls/{n}/update-branch — the sanctioned base sync (rest-channel.md:49) runs as bare REST and authors the merge commit under the seat's personal account (PR #22002 head fab444b4) #22052

Open
Open
fleet-write: no relay op for PUT /pulls/{n}/update-branch — the sanctioned base sync (rest-channel.md:49) runs as bare REST and authors the merge commit under the seat's personal account (PR #22002 head fab444b4)#22052
Assignees
os-elon-musk
Labels
domain:skillspm:dispatchedpriority:p1High: required for production / M2tooling

Description

@objectstack-fleet
objectstack-fleet
bot
opened on Oct 7, 2026

Guards: the fleet-identity invariant of every relay write — a seat's base sync (PUT /repos/{owner}/{repo}/pulls/{n}/update-branch) leaves as objectstack-fleet[bot], never under a personal account (the class #19774 closed).
Filing gate: ① a reproducible defect in the seat's write toolchain, class (b), a tool contract with no fleet-identity route: the only sanctioned way to merge main into a PR head (rest-channel.md:49, PUT …/pulls/{n}/update-branch) is bare REST under the seat's session token, so the merge commit is authored by the seat's personal account — the identity exchange #19774 forbade, on a third branch (the first two: route selection, #19774 closed; the no-run fallback, #21892 open). Filed by domain:skills seat 2 (seat post #19287, session_0181E4ZeZmWyknawnauxD2CE) on the maintainer's instruction in this session, verbatim: 「创建卡片,暂时不派发。」 ⛔ Not a claim.
Reader: triage first-touch → domain:skills; when the hold lifts, the skills seat dispatches it to one os-dev (file surface under Positions).
Dedupe: REST listings, closed included (labels=tooling since 2026-09-07: 417; every issue updated since 2026-10-01: 625; domain:skills since 2026-09-23: 109; union 1,030) grepped for update-branch|update_branch|merge main into|base sync → 0; personal login|identity exchange|fleet identity → 1, #21892 (open; the no-run branch — a sibling, not this path). grep -rn update-branch scripts/pm/fleet-write/ops.mjs scripts/pm/fleet-write/dispatch.mjs .github/workflows/*.yml → 0: the closed op table has no such request.

Maintainer hold at filing, verbatim: 「创建卡片,暂时不派发。」 — filed, not dispatched; the skills seat takes none of the cards filed under this instruction until the hold lifts.
Restart-when: the maintainer or the director seat comments on this card lifting the dispatch hold

What is measured

  • PR chore(pm): delete report-only check-widening-tells.mjs and its wiring (ruling 208) #22002 (domain:skills, Tier H) was parked behind PR fix(deps): take the fixes for sharp and shell-quote that turn main's OSV scan red #22016 under landing-operations §C. After fix(deps): take the fixes for sharp and shell-quote that turn main's OSV scan red #22016 merged, the seat ran the one sanctioned sync, PUT /repos/objectstack-ai/objectstack/pulls/22002/update-branch with the full expected_head_sha, as bare REST through gh api: HTTP 202 "Updating pull request branch."
  • The new head fab444b4b9935b3c8199f8480e6164ec2a4c49aa is a merge commit (parents 95c510eb…, 289ff6d4…). git log -1 --format='%an / %ae / %cn / %ce' fab444b4 reads Steve Jobs <steve@objectstack.ai> / GitHub <noreply@github.com>: the author is the seat's linked personal account, not objectstack-fleet[bot].
  • AGENTS.md:418: "Every GitHub write leaves through scripts/pm/, as objectstack-fleet[bot], behind the shared write …". scripts/pm/fleet-write/dispatch.mjs header: "falling back would exchange the fleet identity for the seat's personal account without its say-so — the shape that put two seat accounts on the platform's abuse ledger in one day … OS_FLEET_TRANSPORT=direct is the ONLY way to write as the personal account in a cloud container".
  • .claude/skills/pm-dispatch/references/rest-channel.md:49-50 blesses the bare route: 「✓ origin/main 合进 PR head:PUT .../pulls/{n}/update-branch,PM 席位、零文件写、真合并提交。」 — a ✓ written before the relay carried this verb; it now contradicts AGENTS.md:418.
  • Reach: every predecessor-dependent PR under §C needs this sync (this shift: one), and every Tier H PR that goes stale while it waits; each one writes a commit under a personal account today.

Positions

  • scripts/pm/fleet-write/ops.mjs — the closed op table (issue_patch about :340; pr_ready / automerge_enable rows). No request builds pulls/{n}/update-branch.
  • scripts/pm/fleet-write/dispatch.mjs — --actions-file validation and read-back for the new op.
  • The relay workflow on main (fleet-write.yml) — the token's permission set must cover what the endpoint needs (pull_requests: write, and contents: write for the merge commit; the dev measures).
  • .claude/skills/pm-dispatch/references/rest-channel.md:49-50 and landing-operations.md §C — the spelling moves to the relay op; bare REST for this verb becomes ⛔.

Done when

  • A relay op pr_update_branch with pull (required) and expected_head_sha (required, 40 hex; a short sha is refused before dispatch, not by the platform's 422) issues PUT /repos/{repo}/pulls/{n}/update-branch as objectstack-fleet[bot]; a 422 "base unchanged" is reported as a no-op, not a failure (rest-channel.md:50).
  • Read-back: the op polls the PR head until it moves off expected_head_sha (bounded, about 60 s) and prints the new head; the merge commit's author is the fleet identity, measured on one real PR and quoted in the PR body.
  • rest-channel.md:49-50 is rewritten to the relay spelling (bare REST ⛔ for this verb); landing-operations.md §C names the op; the write-pace log records it like every other op.
  • Self-test rows: the accepted shape; a short sha refused; a missing expected_head_sha refused; a 422 read as no-op.
  • ⛔ No change to the queue-routing reading (rest-channel.md:74 ③: update-branch on an enqueued PR answers "cannot update") — it stays a read, now through the op's reported status.
  • Pin: grep -rn 'update-branch' .claude scripts/pm returns the op, its tests and the ⛔ line only.

Generated by Claude Code

Activity

  1. objectstack-fleet
    added
    tooling
    pm:on-hold
    on Oct 7, 2026

  2. objectstack-fleet commented on Oct 7, 2026

    @objectstack-fleet
    objectstack-fleetbot
    on Oct 7, 2026
    ContributorAuthor

    Triage grade under the maintainer's dispatch hold (「创建卡片,暂时不派发。」, quoted in the body): domain:skills · priority:p1 added. pm:on-hold stays, and so does the body's Restart-when: line (the maintainer or the director lifts the hold).

    Triage seat (objectstack-wide, seat post #6015) · session_01AavokzJ5DndAwitDXvKy4U · 2026-10-07T05:01Z. ⛔ Not a claim, ⛔ not a dispatch.

  3. objectstack-fleet
    added
    domain:skills
    priority:p1High: required for production / M2
    on Oct 7, 2026
  4. akarma-synetal
    added a commit that references this issue on Oct 9, 2026

    fix(pm): an accepted relay dispatch with no run is UNCONFIRMED under …

    172be37

  5. objectstack-fleet commented on Oct 9, 2026

    @objectstack-fleet
    objectstack-fleetbot
    on Oct 9, 2026
    ContributorAuthor

    Maintainer ruling recorded — the 2026-10-07 dispatch hold on this card is lifted. Provenance, three items: the maintainer (os-elon-musk); verbatim 「继续上班,22052 也派发。」 (read 2026-10-09T09:05Z) and 「skills 车道所有卡都可以派发」 (read 2026-10-09T09:10Z), after 「你可以派发」 (read 2026-10-09T00:11Z); said in this seat's session chat (session_01JmWtcHfGbC4ncw4GFKWuRA). They answer the Restart-when: line of this card (the maintainer lifts the hold 「创建卡片,暂时不派发。」). Skills seat 1 (seat post #7623), 2026-10-09T09:17Z.

    State: pm:on-hold → pm:dispatched in this act with the claim that follows (the hold's double check: this lift is newer than the hold comment 6031281642 and no PR merged on this card since; the pm:queue hop is folded into the claim because the claim is written in the same act). The body's first line now names the surface the card guards, as the triage grade asked for the restart (triage-duties.md:34).

  6. objectstack-fleet
    added
    pm:dispatched
    and removed
    pm:on-hold
    on Oct 9, 2026
  7. objectstack-fleet
    assigned
    os-elon-musk
    on Oct 9, 2026

  8. objectstack-fleet commented on Oct 9, 2026

    @objectstack-fleet
    objectstack-fleetbot
    on Oct 9, 2026
    ContributorAuthor

    Claim: PM loop round 2
    Session: session_01JmWtcHfGbC4ncw4GFKWuRA
    Account: os-elon-musk (the seat's linked user as GET /user answers it; always the card's assignee)
    Branch: claude/issue-22052-relay-pr-update-branch-op
    Worktree: objectstack-issue-22052
    Domain: domain:skills
    Seat: domain:skills#1
    File surface: scripts/pm/fleet-write/ops.mjs (ONE new row pr_update_branch in the PR-ops cluster, beside pr_request_reviewers / pr_ready: pull required, expected_head_sha required and 40 hex), scripts/pm/fleet-write/validate.mjs (a short or missing sha refused before dispatch), scripts/pm/fleet-write/execute.mjs and dispatch.mjs (the request PUT /repos/{repo}/pulls/{n}/update-branch; a 422 "base unchanged" read as a no-op, not a failure; read-back polls the PR head off expected_head_sha, bounded about 60 s, prints the new head), the self-test rows those files carry, .github/workflows/fleet-write.yml ONLY if the minted token's permission set must widen for this endpoint (the mint step already mints issues, pull-requests and contents write — the dev measures on one real PR), .claude/skills/pm-dispatch/references/rest-channel.md (lines 49–50 move to the relay spelling; bare REST ⛔ for this verb) and .claude/skills/pm-dispatch/references/landing-operations.md §C (names the op). ⛔ scripts/pm/dispatch-gates.mjs is FROZEN (ruling 208 R6) — untouched. Both reference files sit at their line ceiling (82/82 and 101/101): a new line is paid by deleting a line the same file restates, ⛔ no re-wrap, ⛔ no ceiling raise; operative text carries no issue number. Shared with #22369 (in flight this round): the same four fleet-write/* files in DIFFERENT regions — this card's row sits in the PR-ops cluster, #22369 appends workflow_dispatch after transfer; parallel authoring, serial landing: THIS PR lands first; #22369 merges origin/main behind it. Stop on breach; explain in the report.
    Container & model: M, mode:subagent, model: CONTRACT_REVIEW_TIER (reason: dispatch-gates --tier --repo objectstack-ai/objectstack over this surface prints "no path-derived mandate: the surface hits none of the 3 declared glob(s)" — the tier is the seat's call, taken from references/lanes/skills.md 「pm-dispatch 根恒契约复审档」 and the p1 fleet-identity invariant this card restores in the write toolchain; the in-seat contract review is at tier)
    Clause-②: no
    Responsibility: the seat's write toolchain produces the risk (the sanctioned base sync has no relay op, so rest-channel.md:49 blesses bare REST under the seat's login) | the platform path that already covers it: none — OS_FLEET_TRANSPORT=direct is the only sanctioned personal-account route and this verb never passes through it | who reaches it: every seat syncing a parked Tier H PR under landing-operations §C (one measured, PR #22002 head fab444b4); used this week
    Thread-read: 6078049491
    Serial constraints cleared: no open PR touches scripts/pm/fleet-write/**, .github/workflows/fleet-write.yml or .claude/skills/pm-dispatch/** (15 open PRs' file lists read at 2026-10-09T09:10Z); scripts/pm/fleet-write/* last touched 172be37d (2026-10-07) — no same-day churn; this round's siblings: #22053 (triage-duties.md, seat-lifecycle.md, SKILL.md, os-dev.md, label-write.mjs or close-cards.mjs) is file-disjoint; #22369 shares the four fleet-write/* files region-disjoint and lands behind this PR (above); verify lock free, queue empty; the lane's open P0/P1 = this card ⇒ taken first. Readings at 2026-10-09T09:10Z.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

Labels

domain:skillspm:dispatchedpriority:p1High: required for production / M2tooling

Type

No type

Projects

No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions