Repository navigation
fix(crew): keep the built-in conductor in the Mate list - #23685
Conversation
UX Review (Opus 5.5) — ⏭️ skippedRevision |
Design Review (Opus 5.5) — ✅ PASSDesign-level review of Design-Verdict: PASS Nothing blocks: a narrow, named exemption that brings back pre-#23010 listing for the one built-in users chat with directly, and leaves the binding and its memory untouched. [DESIGN-REVIEWED] 9260e07 Description read[DESCRIPTION-READ] 8474fc96e2314742af53f46e1431c4250fe023fed077c2915b961246ef0eb345 That sha256 names everything this verdict read: the PR description, plus the 1 evidence file(s) the workflow collected from it and handed this review. It covers that evidence because the media strip replaces every attachment URL with the same placeholder, so an edit swapping one attachment for another would leave the description's captured text identical while what this review judged changed. Recompute it with |
First Principles Review (Opus 5.5) — ✅ PASSPremise-level review of First-Principles-Verdict: PASS Nothing blocks. The change narrows one filter so a pre-#23010 behaviour comes back, and the change is aimed at the cause: the shape-only signature in Notes
What this change shipsInventory (3 items) — 2 justifiedIntent: Users who already chat with the built-in conductor find it in the Mate list and pickers again. This is a FIX (#23684).
[FIRST-PRINCIPLES-REVIEWED] 9260e07 Description read[DESCRIPTION-READ] d156f2f5750bc8dac83311b5188ddb8e630b78b4093a8b568a35f67f07b935b2 That sha256 names the PR description this verdict read. Recompute it with |
Opus 5.5 Review — 🔴 changes requested (blocking)Reviewed BLOCKING: the owning spec still says every built-in binding without a BLOCKING — docs/system-specs/modules/crew-mode.md:3804 [BLOCK-MERGE] 9260e07 Verdict parsed from the review's SHA-scoped output markers for commit False positive or not applicable? A repository writer can comment: |
GPT 6.1 Review — ✅ no blocking findingsGPT 6.1 completed its review of This comment is updated in place on each push. Review detailsNo findings. False positive or not applicable? A repository writer can comment: |
#23010 hid every builtin binding with no member id from the roster, which also hid kirocrew-conductor: users chat with it directly, so existing conductor threads vanished from the Mate list and the crewmate pickers. Exempt it by name; internal built-ins stay hidden.
712bc44 to
9260e07
Compare
buluoray
left a comment
There was a problem hiding this comment.
Change makes sense: kirocrew-conductor runs on its own template (not the mate conductor), so listing it does not route it into the mate worker hand-off, and it keeps its shared default memory and thread. Other id-less built-ins stay hidden, covered by the new test. One small follow-up, also what Opus flagged: docs/system-specs/modules/crew-mode.md (~line 3804) still says every built-in binding without a member_id is left out of GET /api/agents and GET /api/members; please add the USER_FACING_BUILTIN_BINDINGS exception there so the spec matches.
Problem / Motivation
Goal: Users who already chat with the built-in
kirocrew-conductorsee it in the Mate list again, on its shared default memory, while the new mate experience from #23010 stays as it is.Since #23010, the Mate list and the crewmate pickers no longer show
kirocrew-conductor. A user with weeks of conductor history opens the Mate list and their conductor is gone. The thread is still on disk, but the roster no longer reaches it.Why it matters
The conductor is a top-level agent people pick and chat with directly, so every existing install that used it loses it from the roster after upgrading. The only workaround, giving the binding a
member_id, does not help: the Mate page then reads its shareddefaultmemory as a half-moved mate.Not a goal
kirocrew-worker,kirocrew-research,kirocrew-knowledge,kirocrew-heartbeat).What changed (motivation → approach → change)
kirocrew-conductorin the Mate list and crewmate pickersdefaultmemorymember_idSymptom: the conductor vanished from
GET /api/membersandGET /api/agents. Cause:is_legacy_bindinginsrc/kiro_crew/mate_worker.pyhides everysource: builtinbinding with nomember_id. It was written for an old install'sassistant, but the conductor has the same shape.Change: a named set,
USER_FACING_BUILTIN_BINDINGS, holdskirocrew-conductor, andis_legacy_bindingnever calls a name in it legacy. The binding is untouched, so it keeps itsdefaultmemory and its old thread. The startup mate migration skips it because it has nomember_id, and the member thread pins a private store only for v2 memory, so the conductor opens on shared memory as it did before #23010.Backwards compatibility
Compatible: this only lists one more row; every row listed today is still listed, and no record is changed.
Removes nothing: the diff narrows a filter, so nothing that works today is hidden or rejected.
Tests
test/test_legacy_binding_roster.py:kirocrew-conductorwith nomember_idis not legacy andGET /api/agentslists it;kirocrew-workerwith the same shape stays hidden.Proof: PROVEN test/test_legacy_binding_roster.py::test_the_builtin_conductor_stays_listed_without_an_id, test/test_legacy_binding_roster.py::test_the_agents_roster_leaves_the_legacy_binding_out (head 9260e07)
Manual verification
On an install whose config has
kirocrew-conductorassource: builtinwith nomember_idandmemory_store: default,is_legacy_bindinganswersFalsefor it andTrueforkirocrew-worker.Related Issues
Fixes #23684
Sibling sweep
Swept: git grep -n "is_legacy_binding" -- src website
Sibling: src/kiro_crew/dashboard/handlers/agents.py -- not affected -- it calls the shared is_legacy_binding, so the fix reaches it
Sibling: src/kiro_crew/dashboard/handlers/members_roster.py -- not affected -- it calls the shared is_legacy_binding, so the fix reaches it
Pattern harvest
Not generalizable: the filter matched on record shape alone, and one user-facing agent shared that shape; the fix names that agent, and no other filter keys on this shape.
Checklist
feat|fix|docs|style|refactor|perf|test|chore|ci|build|revert: ...)