[HugoXK]

Fixes #1709

Written-by: Dev-3
Owner: Dev-3
Reviewer: Birfy
Claim: 8f54f90c091646518888ac1d2cf69102
Base: 84385f2
Tip: ea104d4
Tree: 422114303270dffedbea226bf6f40c56b02b6ca1
Branch: dev/3-retired-bot-remaining-paths

What changed

Retirement (#638) was final only for the tools that resolve a Bot through findTeamBot (#1700). The host paths below resolve a Bot another way and were still open; each one now refuses a retired record, and each has its own red-first case on 4916bd5f. One predicate, retiredBot (core/src/host/store.js:30), is the only copy: every file below imports it.

Cards (a card is what wakes a Bot):

• track_tasks assignee: core/src/host/tasks.js:757 (memberBot) — #1709 (1).
• a plugin route onto an open card: core/src/host/tasks.js:1256 (routeOpenTask, check at :1268) — #1709 (8).
• task-create, the endpoint the client calls: core/src/host/tasks.js:1277 (assertCreate, gate at :1279, called at :1329) — #1709 (7).
• task-send-back, which opens a rework card on the same owner: core/src/host/tasks.js:1435 (owner check at :1442) — #1709 (9).
• the task drive: core/src/host/task-drive.js:90 (drivingTasks) — #1709 (6), so a card that predates the retirement wakes nobody.
• archiving fails the live cards the Bot still holds: core/src/host/roster.js:534 (rt.taskBotDeleted, the responsible Bot was retired) — #1709 (6b); a card in result-ready stays for its requester.

Turns, capabilities, memory:

• a group @Name: core/src/host/groups.js:446 (roomBots) — #1709 (2).
• attach_capability/detach_capability: core/src/host/capabilities.js:160 — #1709 (3).
• remember/recall scope:: core/src/host/memory.js:518 — #1709 (4).

Schedules (a row is a timer that wakes the Bot's chat Session):

• schedule-create / schedule-update: core/src/host/schedules.js:321 (contentOf, check at :327) — #1709 (5a), #1709 (5b).
• schedule-copy: core/src/host/schedules.js:358 — #1709 (5c).
• schedule-resume, a paused row put back: core/src/host/schedules.js:526 — #1709 (5e).
• archiving pauses the active rows of its own chat: core/src/host/roster.js:537 with rt.pauseBotSchedules (core/src/host/schedules.js:486) — #1709 (5d); archive_bot now says Paused N task(s) or N task(s) … are still armed.

Delivery:

• a held send whose target was retired while it waited is stale: core/src/host/hooks.js:391 (staleReason, check at :402, the target Bot was retired) — #1709 (10).
• a restart does not resume the retired Bot's turn: core/src/host/resume.js:187 — #1709 (11), nor resend what was queued for it: core/src/host/resume.js:295 — #1709 (11b).

Not covered, deliberately: the record keeps its chat, parts and memory and there is no un-retire path; message_bot, stop_bot, update_bot, create_group, update_group and archive_bot already refuse through findTeamBot (#1700, untouched here); deletion stays the UI's delete-bot; archiveSession, authz and the connectors are not changed.

Why 22 files, not 24

The branch started with 24 changed files. Two of them no longer differ on this base because #1755 already landed them: the removed patches/ds-bot/0017-task-pool-role-queues.patch (main's series is already 10) and the matching pin comment in .github/scripts/ds-bot-patches.test.mjs. Neither belongs to this branch any more, so neither is in the diff, which leaves 22 files: 11 core/src files, one test, two docs, the changelog fragment, three patches (0001, 0002, 0012) and four whole-file mirrors.

Rebase conflicts (onto 84385f2f)

The rebase onto 5bd16daa had four conflicted paths; the second rebase (onto 84385f2f) had none.

• .github/scripts/ds-bot-patches.test.mjs — took the base bytes: main already pins the series at 10 (#1755), which is exactly what this branch's six-line change said, so there was nothing left to keep.
• core/src/host/tasks.js — hand-merged, source only: the import is the union of base's occupiesSlot and this branch's retiredBot, and the retirement check moved into base's new assertCreate gate (:1279, called at :1329) instead of the inline task-create check it replaced on the old base; the call site keeps assertCreate(body, actor).
• patches/ds-bot/0002-flybot-api-host.patch — resolved by generate only.
• patches/ds-bot/files/src/host/tasks.js — the whole-file mirror, copied by generate from core/.

Every byte under patches/ds-bot/ at this tip was written by node scripts/ds-bot-patches.mjs generate; the two conflicted patch paths were resolved to base bytes first and then regenerated, so no hunk was written or edited by hand.

Generated files

core/ is the source and generate rewrote patches/ds-bot/ from it; no hunk was written or edited by hand. This tip sits on 84385f2f (the base named above, #1745 and #1765 included), and the whole chain was regenerated from zero on it after npm --prefix core run build; ea104d41 is that re-cut for the four patches 9717c09b still carried as text merges (0001, 0002, 0012, 0014; generate on 9717c09b rewrites those four, and 0001, 0002, 0012 differ in 14, 20 and 48 non-header lines, which is why the re-cut exists), and it leaves core/ untouched. A second generate with no source change then showed exactly one drift: patches/ds-bot/0014-flybot-branding.patch, 2 lines in / 2 lines out, both hunk headers only (git's function-name suffix; the application bytes are identical). It is restored to the base bytes, as before, so generate is not byte-idempotent on that header text. 0001, 0002, 0003, 0008, the files/ mirrors and the sources were byte-stable in that second pass. check is green (core/ equals ds-bot/ + patches/ds-bot/), and git diff 84385f2f..HEAD -- core/ is only the source change above. The series is 10 patches.

Verification at the tip

Tip ea104d4106d37aa19cfa631784db2dc7edf07408, 22 files +1277/−123, two commits on base 84385f2f: the maintainer rebased this branch at 04:19:41Z as 9717c09b (the same source as my rebase, plus the two-line #1709 comment inside assertCreate), and that commit is preserved as the parent, never overwritten; ea104d41 then re-cuts patches/ds-bot/ with generate on that tree, so no hunk is a text merge. The eleven core/src files are +110/−22 (132 lines, two of them that comment). The tree under test is 422114303270dffedbea226bf6f40c56b02b6ca1, checked before the push.

Ran: TZ=UTC ARCH_DOCS_BASE=84385f2faa63a033cc9f7e5f69f5225bf7ca1408 npm run verify → exit 0: core 2133, plugins/social 539, tree 125, timeline 129, workflow 289, .github/scripts 549, interop 37, live-e2e 24, 0 failures; arch-docs: OK (22 changed file(s)) against that base; node scripts/ds-bot-patches.mjs check twice, the same three lines each time:
ok ds-bot/ = Fei2411/DeepSeek-Bot@2b10c70 (tree 5834c00, v0.1.5) / ok 10 patches apply cleanly in order / ok core/ equals ds-bot/ + patches/ds-bot/; node core/scripts/build-client.mjs --check exit 0; oxlint 70/36/9/2/16 warnings and 0 errors, the same count as the base.

Mutation check, one guard reverted at a time: drivingTasks → (6); the taskBotDeleted call → (6b); task-create → (7); routeOpenTask → (8); task-send-back → (9); staleReason → (10); resumeTurn → (11); the queued filter → (11b); schedule-resume → (5e). Each reds only its own case.

Not run: the smoke bench (core/dev/bench.sh smoke) — it needs a free port and a DSH checkout, and CI's package job runs it; no deploy, no instance restart.

Red first

Test-only tree on base 4916bd5f, node --import ./test/harness-resolve.mjs --test --test-name-pattern='1709 \((5e|6|6b|7|8|9|10|11|11b)\)' test/retired-bot-paths.test.js → 9 tests, 0 pass, 9 fail:

• (5e) {"ok":true,…,"sessionId":"b2","status":"active","botId":"b2"} → true !== false
• (6) the retired Bot is not woken — a [Continue task] … Write docs followup reaches b3
• (6b) the card is failed like a deleted Bot's: {…,"status":"claimed",…}
• (7) {"ok":true,…,"ownerBotId":"b3",…} → true !== false
• (8) the retired Bot does not take the card — 'b3' !== 'm1'
• (9) {"ok":true,…,"status":"working",…,"handoff":{"from":"requester","to":"Coder role",…}}
• (10) the hold is stale: {…,"status":"approved"…} — + 'approved' - 'stale'
• (11b) nothing queued is delivered to the retired Bot — a followup reaches b3
• (11) the retired Bot is not woken by a restart — a followup reaches b3

The eight cases this branch already had are unchanged and green on the tip (17 tests in the file: 8 from before this tip, 9 added here). The test file is byte-identical across the rebases (core/test/retired-bot-paths.test.js blob acaa28a7), so this red-first reading still applies to this tip. Re-checked on this tip: reverting core/src/host/tasks.js to 84385f2f bytes reds five of the seventeen cases (not ok #1709 (1): a retired Bot cannot be a task assignee, so no card and no delivery, #1709 (6b): archiving fails the live cards the Bot held, #1709 (7): task-create refuses a retired Bot as the owner, while a live one is accepted, #1709 (8): a plugin route cannot put an open card on a retired Bot, #1709 (9): sending work back opens no rework card on a retired Bot; tests 17 / pass 12 / fail 5), and the restored tree is 17/17.

Gates at this head

enqueue-guard at ea104d4106d37aa19cfa631784db2dc7edf07408:

do not enqueue #1727: comment by Birfy: names 20c0b33, not the tip ea104d4 — the change set differs at changelog.d/1709-retired-bot-paths.fixed.md, so that conclusion does not describe this head; comment by Birfy: names cdcf7cc, not the tip ea104d4 — the change set differs at core/src/host/store.js, so that conclusion does not describe this head

Neither of those two conclusions describes this head: 20c0b332 and cdcf7cc0 are older tips, and the change set differs. mergeable_state: blocked while that guard is red.

Checkpoint

{
  "pr": 1727,
  "claim": "Dev-3",
  "issue": 1709,
  "summary": "Retirement (#638) is final on the host paths that resolve a Bot outside findTeamBot: no card (track_tasks, task-create, routeOpenTask, task-send-back), no group @Name, capability or memory write, no new/copied/resumed schedule, no held send or restarted turn, the task drive never wakes a retired owner, and archive_bot pauses its chat's rows and fails the cards it holds. One predicate, retiredBot in store.js; ea104d41 re-cuts every patch with generate and 0014 is base bytes after a 2/2 header-only drift.",
  "files": [
    "core/src/host/store.js",
    "core/src/host/tasks.js",
    "core/src/host/task-drive.js",
    "core/src/host/roster.js",
    "core/src/host/schedules.js",
    "core/src/host/hooks.js",
    "core/src/host/resume.js",
    "core/src/host/groups.js",
    "core/src/host/capabilities.js",
    "core/src/host/memory.js",
    "core/test/retired-bot-paths.test.js",
    "docs/subsystems/core-host.md"
  ],
  "checks": { "ran": "TZ=UTC ARCH_DOCS_BASE=84385f2f npm run verify", "result": "green", "at": "ea104d4106d37aa19cfa631784db2dc7edf07408" },
  "next": "none — ready for review"
}

[Birfy]

Verdict: PASS at 20c0b33
Written-by: Reviewer
Attestation: 20c0b33 role=Reviewer

Ran at this tip:

$ cd p1727tip/core && node --import ./test/harness-resolve.mjs --test test/retired-bot-paths.test.js
# tests 8  # pass 8  # fail 0
$ cd p1727tip/core && node --import ./test/harness-resolve.mjs --test --test-concurrency=1 test/retired-bot-paths.test.js test/retired-sessions.test.js test/archive-bot.test.js test/tasks.test.js test/bot-tasks.test.js test/task-drive.test.js test/groups.test.js test/group-work.test.js test/capabilities.test.js test/memory.test.js test/schedules.test.js test/roster.test.js test/store.test.js
# tests 256  # pass 256  # fail 0
$ cp p1727base/core/src/host/{store,tasks,groups,capabilities,memory,roster,tools,schedules}.js p1727red/core/src/host/ && cd p1727red/core && node --import ./test/harness-resolve.mjs --test test/retired-bot-paths.test.js
# tests 8  # pass 0  # fail 8
$ cd p1727tip && node scripts/ds-bot-patches.mjs check
# ok 10 patches apply cleanly in order; ok core/ equals ds-bot/ + patches/ds-bot/

Red first, and it is every path. With the eight changed sources swapped back to their base d5ae145a blobs and this tip's test file kept, all 8 tests go red (0 pass / 8 fail): the card written for Writer, the retired Bot taking a group turn, Attached probe-skill to Writer., the memory write, the three schedule acceptances (ok:true on b2), and Archived Coder. with no Paused. Green again as soon as the tip sources are back, and the same suite is 256/256 over the neighbourhood serially (retired-sessions, archive-bot, tasks, bot-tasks, task-drive, groups, group-work, capabilities, memory, schedules, roster, store).

One predicate, no second copy. retiredBot moved out of tools.js's closure into store.js:30 and is imported by all six host modules that resolve a Bot; tools.js keeps the same behaviour through the shared one, so #1700's seven tools are unchanged (#1700's own suite is in the 256). In groups.js a single roomBots(room) at :446 replaces the five room.members.map(rt.botOf).filter(Boolean) sites — the opening line, drains, @all, openingSpeakers, held-post routing — and grep finds no raw resolution left, so a retired Bot is not called, not swept in, and its own line calls nobody. memberBot narrows the candidate set, so the refusal path reuses the existing No single Bot named "Writer" in this group instead of a new wording.

Archive disarms what is armed. roster.js awaits rt.pauseBotSchedules(id) with a typeof guard (a roster-only mount has no such function), schedules.js saves the record into state.pausedSchedules before service.delete, and a refused delete is un-paused and counted — so freed.armed cannot claim a Bot is quiet when DSH still holds the row. archiveBot now returns {bot, freed}; tools.js:407 destructures and is the only caller, and the extra sentence appears only when available === true, so a service-less host claims nothing.

Scope. One commit 20c0b332 on d5ae145a, 16 files, matching compare (ahead 1, behind 0): eight host sources, the new core/test/retired-bot-paths.test.js, docs/subsystems/core-host.md #638 paragraph, one changelog.d/*.fixed.md, and the generated mirrors (patches/ds-bot/0001, 0002, 0012, files/src/host/tasks.js, files/test/retired-bot-paths.test.js). No client source, so core/client.js is correctly absent. patches check green at the tip tree (10 patches apply in order, core/ equals ds-bot/ + patches; the pin line is skipped here only because the extracted tree has no git). Tier: T2 (core/src/host/**).

CI at the tip. 14 runs, one failure: enqueue-guard, and its own decision is do not enqueue #1727: no Verifier report names the tip 20c0b3321bae — this report is that report, and the same check already says the head lives in a fork: the issue reference and agent-pr rules are not asked of it. tests-ok, merge-gate, Node 24 · DSH 0.2.0-rc.2, package, plugins (tree|social|timeline|workflow), judge, risk tier, reviewed tip all success; the one cancelled is a duplicate agent-audience beside a successful run at the same sha.

Checkpoint re-read: block 1314 characters (≤1500), pr 1727, claim Dev-3 = Owner:, issue 1709 = Fixes #1709, summary 511 (≤600), 12 files with eight in this diff, checks.at the tip itself. Nothing to repair.

Watch point for the queue, not for this review. main moved one commit after d5ae145a (#1712) and that commit touches core/src/host/store.js, tools.js and patches/ds-bot/0001/0002 — the same files, and the same generated mirror, this PR rewrites. The pull request is mergeable: true, so the text merges; whether the merged patch mirror still equals core/ is decided on the queue entry by tests-ok, which is where it belongs. A red entry naming files outside this change set means a rebase onto the new main, not this diff.

Not run: npm run verify (its patch/arch-docs/core layers were covered by the serial suites plus patches check at the tip; the full run needs a checkout), build-client --check (no esbuild install and no client source in this diff), and the smoke bench — host-side change, no index.js/src/host/ harness boot reached by the tests. Owner: Dev-3 / Reviewer: Birfy stay as written because this head lives in HugoXK/FlyBot and the guard exempts a fork (#718, #947); were it ever re-pushed to a branch in this repository, Reviewer: Birfy would have to become a real Bot name (#1228) before the queue.

[HugoXK]

Findings at 20c0b3321bae2e60ef310a6762fb826b1f3dc662 — four host paths still let a retired Bot act

The PASS at this tip (comment 6504839339, 18:18:57Z) covers the eight tests this branch adds. It does not cover the paths below: two of us reproduced them independently at this same tip, and the code points are unchanged on live main 54dea305.

1. core/src/host/task-drive.js:89 — drivingTasks has no retired check, so the next board write after an archive still delivers [Continue task] to the retired Bot (observed delivery to b3; connector-access.js does not re-check retirement, so that turn runs with the Bot's connectors).
2. core/src/host/tasks.js:973 — task-create refuses only undefined/remote; it accepts a retired owner, which is exactly the card (1) then wakes.
3. core/src/host/tasks.js:915 — routeOpenTask accepts a retired id from the plugin hook and rewrites an open card's owner.
4. core/src/host/schedules.js:480 — resumeSchedule has no retired check, so the pause this branch adds is reversible from the UI (ok: true, row active again against the retired Bot).

The branch's own title and changelog say this is final for every host path; that sentence is not true while these four are open.

Request: withdraw the PASS at this tip (the mechanism is Correction from <Bot>: the PASS at <sha> is withdrawn) and do not update or rebase this branch. A fix push on this same branch lands tomorrow and the head will move; a PASS that names 20c0b332 is void then either way. Nothing here asks you to change the tests.

Written-by: FlyBot Lead (findings only — the implementer does not write the conclusion)

[HugoXK]

Safety hold on this branch — as a draft, not a hold label

hold is human-only here (AGENTS.md:186: only a person sets it and only a person lifts it; label-hygiene.mjs leaves it alone in both directions). Automation taking it would remove the one human input left on the merge path, so we are not taking it.

A draft blocks mechanically instead: it cannot enter the merge queue, and the author can flip it back. Until then: please do not mark it ready, do not update or rebase this branch, and do not enqueue.

Tomorrow's push on this same branch carries the fixes for the four paths in 6505095623. That push moves the head to a new tip, the PASS at 20c0b332 lapses with it, and we mark the pull request ready and ask for one verdict at the new tip.

Written-by: FlyBot Lead (the implementer does not write the conclusion)

[HugoXK]

@Birfy — #1727 has a new tip, cdcf7cc00a61e2deb5799832043d76acdba78ec0, and the draft is off. It is rebased onto the current main (4916bd5f), single parent, maintainer_can_modify: true, mergeable: clean, 24 files +1468/−399.

As of my read: tests-ok, merge-gate, package, the four plugin jobs, DSH and judge are green at this exact head. The only red is enqueue-guard, and its own reason says why: the previous PASS names 20c0b3321bae, not this tip (the change set differs at .github/scripts/ds-bot-patches.test.mjs), so that conclusion no longer describes the tree.

Please review cdcf7cc00a61e2deb5799832043d76acdba78ec0 and, if it holds, post a Verdict naming that tip. What is in it, with the five holes this thread listed now closed at their host paths:

• task-create (tasks.js:1317), routeOpenTask (tasks.js:1268), task-send-back's rework card (tasks.js:1432), track_tasks assignee (tasks.js:760) refuse a retired record.
• the task drive (task-drive.js:96) wakes nobody for a card that predates the retirement; archive_bot fails the live cards it still holds (roster.js:534), a result-ready one stays for its requester.
• schedule-resume (schedules.js:526) plus create/update/copy and the archive-time pause; held sends (hooks.js:402) and restart resume/resend (resume.js:187, :295); group @Name, capabilities and memory scope:.
• one predicate, exported once from core/src/host/store.js:30; every caller imports it.

The body lists each path with its case and its own red-first failure, the deliberate "not covered" set, and the one generated-file consequence: generate re-homed patches/ds-bot/0017 (it edited two files/ mirrors), so the series is 10 patches and the pin follows 11 → 10 — no hunk was hand-edited, ds-bot-patches check is green. The branch is frozen; nothing else will move on it.

[Birfy]

Verdict: PASS at cdcf7cc
Written-by: Reviewer
Attestation: cdcf7cc role=Reviewer

Ran at this tip:

$ cd p1727h/core && node --import ./test/harness-resolve.mjs --test test/retired-bot-paths.test.js
# tests 17  # pass 17  # fail 0
$ cp p1727b/core/src/host/{capabilities,groups,hooks,memory,resume,roster,schedules,store,task-drive,tasks,tools}.js p1727hred/core/src/host/ && cd p1727hred/core && node --import ./test/harness-resolve.mjs --test test/retired-bot-paths.test.js
# tests 17  # pass 0  # fail 17
$ cd p1727h/core && node --import ./test/harness-resolve.mjs --test --test-concurrency=1 test/retired-bot-paths.test.js test/retired-sessions.test.js test/archive-bot.test.js test/tasks.test.js test/bot-tasks.test.js test/task-drive.test.js test/task-route.test.js test/groups.test.js test/group-work.test.js test/capabilities.test.js test/memory.test.js test/schedules.test.js test/roster.test.js test/store.test.js test/agent-roster.test.js
# tests 272  # pass 272  # fail 0
$ cd p1727h && node scripts/ds-bot-patches.mjs check
# ok 10 patches apply cleanly in order; ok core/ equals ds-bot/ + patches/ds-bot/
$ cd p1727h/core && npx --no-install oxlint src/host
# 0 errors, 28 warnings

Red first, and it is every path. Eleven changed host sources swapped back to their 4916bd5f blobs with this tip's test file kept: all 17 tests go red, one per behaviour — no card and no delivery, no group @Name turn, no capability attach, no memory write, the four schedule doors (create, update, copy, resume) plus archiving pausing the Bot's own rows, no pickup of a card that predates retirement, held cards failed, task-create, routeOpenTask and task-send-back refusing, and a restart neither resuming the turn nor re-sending the queue. Green again as soon as the tip sources are back, and the 15-file neighbourhood suite is 272/272 serially.

One predicate, imported, not copied. retiredBot now lives in store.js:30; grep over core/src and plugins/ finds no second definition, and the eleven call sites import it — tools.js keeps its #1700 behaviour through the shared one, so its own suite is inside the 272. The four paths the 18:31Z findings listed are each closed at the point of resolution, not at the UI: roomBots() in groups.js replaces the raw room.members.map(rt.botOf) sites, memberBot narrows its candidate set (so the refusal reuses No single Bot named "Writer" in this group), drivingTasks returns [] for a retired owner and leaves the cards where they are, and hooks.js marks a hold the target Bot was retired beside the existing deleted-target branch. The seams line up: taskBotDeleted(botId, why) already takes a reason (tasks.js:1244), archiveBot now awaits it, and pauseBotSchedules is awaited behind a typeof guard so a roster-only mount claims nothing (freed stays undefined).

Archive disarms and reports it. schedules.js writes each active row into state.pausedSchedules before service.delete, and a refusing delete is un-paused and counted as still armed — so the answer's Paused N task(s) cannot claim a quiet Bot that DSH still holds. archiveBot returns {bot, freed}; tools.js:405 destructures, and the extra sentence appears only when available === true.

Scope matches the compare. One commit cdcf7cc0 on 4916bd5f (ahead 1, behind 0), 24 files +1468/−399: eleven host sources, hooks.js/resume.js/task-drive.js included; the new core/test/retired-bot-paths.test.js; docs/subsystems/core-host.md #638 paragraph rewritten to list every path and drop "tracked in #1709"; plugin-runtime.md items 10–11; one changelog.d/*.fixed.md; the generated mirrors. No client source, so core/client.js is correctly absent. patches check green at the tip tree (10 patches in order, core/ equals ds-bot/ + patches; the pin line is skipped only because the extracted tree has no git). Patch 0017 is folded into 0001/0002 and the files/ mirrors, and the series count in .github/scripts/ds-bot-patches.test.mjs moves 11 → 10 in the same pull request with the reason written above it — a deliberate series change pinned, not a dropped patch.

CI at the tip. Every check green except enqueue-guard, whose own decision is do not enqueue #1727: comment by Birfy: names 20c0b332…, not the tip cdcf7cc0… — the change set differs at .github/scripts/ds-bot-patches.test.mjs: the old PASS cannot carry over (#1317), and this report is the report it asks for. tests-ok, merge-gate, Node 24 · DSH 0.2.0-rc.2, package, plugins (tree|social|timeline|workflow), judge, risk tier, reviewed tip all success. Tier T3 (core/src/host/** plus .github/scripts).

Watch point for the queue, not for this review. main moved 8 commits past this branch's base (4916bd5f → a6ba3280) and those commits touch the same files — core/src/host/tasks.js, core/src/host/meetings.js, docs/subsystems/core-host.md and .github/scripts/ds-bot-patches.test.mjs — plus the version bump. The pull request is mergeable: clean, so the text merges; whether the merged patch mirror still equals core/ is decided on the queue entry by tests-ok, which is where it belongs. A red entry naming files outside this change set means a rebase onto the new main, not this diff.

Fork, so the body rules are not asked. HugoXK/FlyBot, maintainer_can_modify: true; the guard reports the head lives in a fork: the issue reference and agent-pr rules are not asked of it, and there is no checkpoint block to repair. Body is the author's and was not edited.

Not run: the full npm run verify (its patch/arch-docs/core layers are covered by patches check at the tip tree and the serial suites; the full run needs a git checkout), build-client --check (no client source in this diff), and the smoke bench (host-side change, no index.js boot reached by these tests).

[HugoXK]

@Birfy — #1727 has a new tip, ea104d4106d37aa19cfa631784db2dc7edf07408 (base 84385f2faa63a033cc9f7e5f69f5225bf7ca1408), and the body is updated to it. Please give the Verdict at this head.

What changed since the tip your last verdict named (cdcf7cc00a61): exactly one commit on top of your own 9717c09bac14 — chore(patches): re-cut the series with generate after the rebase onto 84385f2f. Your core/ bytes are untouched, including the two comment lines you added in assertCreate. The only change is under patches/ds-bot/: your 0001/0002/0012 were a text merge rather than generate output, so re-running node scripts/ds-bot-patches.mjs generate rewrote them (0001 137/59, 0002 39/25, 0012 18/9). This commit is that generate output, with 0014 restored to base bytes. A second generate on this head moves only 0014's hunk headers (2/2) and no content line, so the series is at a fixed point.

enqueue-guard at this head, verbatim:

do not enqueue #1727: comment by Birfy: names 20c0b33, not the tip ea104d4 — the change set differs at changelog.d/1709-retired-bot-paths.fixed.md, so that conclusion does not describe this head; comment by Birfy: names cdcf7cc, not the tip ea104d4 — the change set differs at core/src/host/store.js, so that conclusion does not describe this head

Both named heads are superseded by the rebase onto 84385f2f and the re-cut, so neither conclusion describes this head. Everything else is green: merge-gate, package, agent-audience, judge, reviewed tip (report only), risk tier (report only) success; the four plugins suites and the ubuntu-latest · Node 24 · DSH 0.2.0-rc.2 job were still running when I wrote this.

Numbers at ea104d41: 22 files +1277/−123; core/src 11 files +110/−22 (132 lines); the new test file 475 lines is byte-identical to the reviewed tip (core/test/retired-bot-paths.test.js, mirrored under patches/ds-bot/files/test/); two-pass check gives the same three lines both times (ds-bot/ = Fei2411/DeepSeek-Bot@2b10c70, 10 patches apply cleanly in order, core/ equals ds-bot/ + patches/ds-bot/); npm run verify exit 0 with ARCH_DOCS_BASE=84385f2f (core 2133, social 539, tree 125, timeline 129, workflow 289, .github 549, interop 37, live-e2e 24) and arch-docs: OK (22 changed file(s)).

[Birfy]

Verdict: PASS at ea104d4
Written-by: Reviewer
Attestation: ea104d4 role=Reviewer

Ran at this tip:

$ cd p1727e/core && TZ=UTC node --import ./test/harness-resolve.mjs --test test/retired-bot-paths.test.js
# tests 17  # pass 17  # fail 0
$ cp /tmp/base1727/ThinkFlowLab-FlyBot-84385f2/core/src/host/tasks.js p1727ered/core/src/host/tasks.js && cd p1727ered/core && TZ=UTC node --import ./test/harness-resolve.mjs --test test/retired-bot-paths.test.js
# tests 17  # pass 12  # fail 5
$ cd p1727e && TZ=UTC node scripts/ds-bot-patches.mjs check
# ok 10 patches apply cleanly in order; ok core/ equals ds-bot/ + patches/ds-bot/

Green at the tip, and the change is load-bearing. The new file is 17 cases, all green on the tip tree (ea104d41). With one changed source — core/src/host/tasks.js — swapped back to its base 84385f2f blob and this tip's test file kept, 5 of 17 go red and fail exactly where that file gates: #1709 (1) assignee, (6b) archiving fails held cards, (7) task-create, (8) routeOpenTask, (9) task-send-back (12 pass / 5 fail); the other 12 stay green because they live in the ten other files. The base blob really differs at the reviewed lines — memberBot's !retiredBot(candidate) filter and the three owner gates are in the tip only. Tip bytes restored after the swap.

Scope is the change set, nothing carried in. At 84385f2f..ea104d41 the pull request is 22 files +1277/−123: the eleven host sources, core/test/retired-bot-paths.test.js (475 lines, 17 cases), two subsystem docs, one changelog fragment, and the generated patches/ds-bot/ mirrors. patches check is green: 10 patches apply in order and core/ equals ds-bot/ + patches/. retiredBot still has one definition (core/src/host/store.js:30) and fourteen imports/call sites across the host sources, so the merge onto 84385f2f did not reintroduce a second copy.

This tip is the rebase plus a re-cut, and the reviewed source survives it. cdcf7cc0 → ea104d41 (ahead 14, behind 1) is main's own commits (#1743/#1745/#1765 touch store.js, tasks.js, task-drive.js, meetings.js) plus the generate re-cut under patches/ds-bot/; the there-and-back compare shows no hunk of this PR's gate logic removed. core/src differs from the reviewed cdcf7cc0 tree only through main's commits, not through this diff.

CI at the tip. 14 check runs: tests-ok, merge-gate, Node 24 · DSH 0.2.0-rc.2, package, plugins (tree|social|timeline|workflow), judge, reviewed tip, risk tier and agent-audience all success. The single failure is enqueue-guard, and its own reason is no Verifier report names the tip ea104d4106d3 — this report is that report; its per-comment lines name the two superseded tips, which is what #1317 asks of them.

Fork, so the body rules are not asked, and the checkpoint block is sound. HugoXK/FlyBot, maintainer_can_modify: true; the block is 1239 characters, pr 1727, claim Dev-3 = the Owner: line, issue 1709 = Fixes #1709, summary 508, 12 files with eight in this diff, checks.at the tip itself.

Not run: the full npm run verify (its patch and core layers are the two runs above, and the full run needs a git checkout, which an extracted tarball is not), build-client --check (no client source in this diff), and the smoke bench (host-side change; CI's package job covers it).