Repository navigation
docs(crews): sync the Crews pages and tool contract with what ships - #306
Conversation
The Crews definition, the agent tool contract, and nearby docs had drifted from the shipped surface: the retired /crew command, the struck agent seat alias, stale propose_crew and request_crew_member contract strings, the approver field #234 removed, and a seat-finish paragraph that described every run including cancelled ones. Every sentence now matches j5/main: History reads in date order with the command's retirement, the contract strings come from tools.ts, seats list their custom-seat runtime overrides, the snapshot keeps each member's reason, and seat finish notices post on change for completed and failed runs. The artifacts user page says one thing about handoff versions, and FORK.md case 3 names the Crew stop replay test. Closes #229 Closes #230 Closes #232 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
|
Important
This repository does not receive automatic reviews because it has fewer than 10 stars. ⚙️ Run configurationConfiguration used: Repository: Jacksondr5/j5code/.coderabbit.yaml Review profile: CHILL Plan: Advanced Run ID: Comment |
Conflicts in FORK.md, crews.md, agent-tools.md, the persona contract, and the artifacts page: merged word by word, keeping this branch's sync of the shipped Crew surface in the sync's handoff artifact wording, and kept both History entries. Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
|
[Review panel: Opus 5.5 + coordinator (Opus 5.5)] Missed file: Suggested fix: change it to "(seat, agent, why it joined, the version it joined at)" and add a fleet-page.md History line. Note that #315 also rewrites AC28 and keeps "who approved it and why", so whichever PR lands second needs the corrected wording. |
… typo Fleet-page AC28 still listed "who approved it and why" in the retired roster snapshot, which has no approver field; it now matches Crews AC6 and AC20, with a History line. Also fixes "Handoffartifacts" on the artifacts page. Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
|
@Jacksondr5 An update on #306:
|
|
[Review panel: Opus 5.5 + coordinator (Opus 5.5)] Re: the fleet-page AC28 finding. Verified at e72d27a. AC28 now reads "(seat, agent, why it joined, the version it joined at)", which matches |
Conflicts in the Crews docs resolved to the shipped wording from #313 and #315, keeping this branch's corrections: - crews.md: the Lifecycle sentence now says a Crew comes back only with its Captain, and the 2026-09-24 History line drops "Crews never unarchive as a unit" (Jackson's suggestion). The roster snapshot records each member's reason and no approver. History is back in date order. - agent-tools.md: the launch report keeps not_started, seat_not_created, and resolve-once, and the seat finish notice adds the unavailable status and over-read-limit case. - fleet-page.md: AC28 takes #315's wording, with both History lines. - personas.md: #315's paragraph, with Retired crews described as one section across every Squadron. Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Conflicts: agent-tools.md takes main's quoted propose_crew and request_crew_member descriptions and tables, which match the shipped strings after #347; crews.md takes main's Definition (Full access default, seats ask their Captain) with this branch's no-approver roster wording, and History keeps one date-ordered list with #301's and #307's 2026-09-24 lines. Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Problem
The Crews pages and the agent tool contract had drifted from what ships (#229), the artifacts user page contradicted itself about handoff versions (#230), and FORK.md case 3 didn't name the Crew-stop replay test that lives in the same upstream test file (#232). None of this is a policy change, except the approver narrowing I ruled on in #234.
What I changed
docs/j5/product/features/crews.md:/crewretirement (Compose Crews in ordinary chat and let members coordinate directly #211) and a 2026-09-24 line for today's changes.docs/j5/product/a2a/agent-tools.md:propose_crewandrequest_crew_membercontract strings are regenerated fromJ5_PROPOSE_CREW_DESCRIPTIONandJ5_REQUEST_CREW_MEMBER_DESCRIPTION.model_selectionandruntime_modeand drop the struckagentalias;reasonis up to 500 characters.CrewSeatFinishNotifier: completed and failed runs only, posts on change, folds into a queued digest, andmissingmeans checked and not there.docs/j5/product/glossary.md: Persona replaces Role in the row and the casing example.docs/j5/product/agent-personas/index.md: the Captain is told on every finish whether the handoff was written, missing, or never declared.apps/server/src/j5/a2a/README.md: the "no cascade" line now namesCrewCaptainArchiveCascade.tsandcrewSeatArchiveGuard.ts.docs/j5/worklog/2026-09-16-crew-command-decoupling.md: one pointer that the Inbox failure alert shipped in feat(agents): Crews, a roster the Captain proposes and the person approves #148; the history is untouched.docs/user/artifacts.md(Handoff files keep every version: fix the artifacts user page #230): a rewritten handoff adds a version at the top of the same file.docs/user/personas.md: retired crews sit in Fleet's Retired section, across Squadrons.FORK.mdcase 3 (Record the Crew-stop replay test under FORK.md case 3 #232): namesreplaying a Crew stop does not dispatch a second real interrupt.apps/web/src/j5/a2a/ThreadA2ARenderer.tsx(no/crewturn) and theFleetCrewcomment inpackages/contracts/src/j5.ts.Why this shape
Definitions are rewritten, not appended (
docs/j5/process/docs.md), so every sentence was changed in place and History got one line per amendment. The contract strings are copied fromtools.tsrather than paraphrased, becauseEnvelopeFormatter.test.tsalready reads this file for thesend_messagecontract and paraphrases drift.Invariants
Surfaces
packages/contracts)Out of scope
docs/j5/product/features/roles.mdto match the persona vocabulary; the glossary still links it.Upgrade and data
None.
Verification
EnvelopeFormatter.test.ts, which readsagent-tools.md, passes.Review focus
agent-tools.mddescribes fix(crews): Crew notices report only what the platform measured #292's retry behavior ("a handoff that cannot be read sends nothing"), so this should merge after fix(crews): Crew notices report only what the platform measured #292.tools.ts.Closes #229
Closes #230
Closes #232
Claude Opus 5.5 via Claude Code
🤖 Generated with Claude Code