Repository navigation
docs(work-ledger): say a patrol stops only once nothing is queued - #38779
Conversation
UX Review (Opus 5.5) — ⏭️ skippedRevision |
First Principles Review (Opus 5.5) — ✅ PASSPremise-level review of First-Principles-Verdict: PASS Nothing to check. What this change shipsInventory (2 items) — 2 justifiedIntent: make the written docs give the same patrol stop rule as the conductor's prompts and skill, so no reader thinks a patrol can stop while work is still queued. FIX.
[FIRST-PRINCIPLES-REVIEWED] 0eee668 Description read[DESCRIPTION-READ] 7e0ad8ce95baa942779db246746170505cf3193314aa1bcd19930c6965f2d13a That sha256 names the PR description this verdict read. Recompute it with |
Design Review (Opus 5.5) — ✅ PASSDesign-level review of Design-Verdict: PASS Doc-only change; it brings the two written copies of the patrol stop rule in line with the shipped prompts and skill, and removes nothing. [DESIGN-REVIEWED] 0eee668 Description read[DESCRIPTION-READ] 9707a8fe23347f12a8079fd8fd9382bac86b9c6a7417e8893bb0292b512d42da 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 |
Opus 5.5 Review — ✅ no blocking findingsReviewed 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: |
The goal-conductor skill and the conductor's patrol prompts end a patrol only when every ledger item is terminal and nothing is queued: a queued item has no ledger row yet, so an all-terminal ledger alone can still leave work undispatched. The work-ledger guide and the crew-mode spec's summary of the Bot conductor's patrol rule still gave the old two-part stop, so a reader of either was told the patrol may end with work queued. Both now carry the same clause. Closes #37192
|
Rebased: main moved in both files this PR edits (the Crew Mode spec and The one red on the previous head, [operator: bolichen97] |
68965f6 to
0eee668
Compare
Code change approver recommendation: A person should decideWhy: the decider estimates a 3% chance that this change introduces a defect, within the 12% limit for approving on its own, but it is held at tier 2: design-doc-gate: A. This PR changes wording only. It adds "and nothing is queued" to the conductor's patrol stop rule in the work-ledger guide and in the crew-mode spec. The new wording matches what main already ships: goal-conductor SKILL.md:722, the patrol_final_report prompt in spec_prompts.py:191, the PATROL_MESSAGE test at test_conductor_patrol.py:780 and the ledger golden fixture. Three other Markdown matches for "every item is terminal" are not the patrol stop rule: the cleanup sweep in session-work-ledger.md, the CLI --until-done flag in cli.md, and a historical RFC audit table. Nothing reads these lines as data, no test pins this prose, and every CI check passed. Tier: 2, held for a person: design-doc-gate: A. An automated recommendation, not a review: a maintainer still decides. It approves on its own only when the estimated chance of a defect is at most 12% and the change has at most 400 changed lines. Decider v3c (217fe607) · commit |
KiroCrewCaptain
left a comment
There was a problem hiding this comment.
Reviewed head 0eee668. The change is two documentation clauses: the work-ledger guide and the crew-mode spec now say a conductor's patrol stops only when every item is terminal and nothing is queued, or when the user says stop. I checked this against the conductor's final-report prompt and the goal-conductor skill's stop condition on this head, and all three now say the same thing. No code, prompt or test fixture changes, so nothing here can change runtime behaviour, and I found nothing that blocks merging.
Problem / Motivation
Goal: Every written copy of the conductor's patrol stop rule says what the conductor's own prompts say: a patrol ends only when every item is terminal and nothing is queued, or when the user says stop.
The goal-conductor skill (#36298) and the conductor's patrol prompts (#36312) end a patrol only once every ledger item is terminal and nothing is queued. Two written copies of that rule still give the old version, "every item is terminal, or the user says stop":
src/kiro_crew/docs/work-ledger.md, which agents read through thesearch_docstool;docs/system-specs/modules/crew-mode.md.Why it matters
A queued item has no ledger row yet, so a ledger whose rows are all terminal can still have work waiting. An agent or a person that reads the guide is told the patrol may stop at that point, which is the early stop the skill change was made to prevent.
Not a goal
main.SHIPPED_PATROL_MESSAGES: they are the texts already-armed loops still carry, and stay as shipped.What changed (motivation → approach → change)
The two docs now give the same stop rule as the skill and the prompts.
src/kiro_crew/docs/work-ledger.mddocs/system-specs/modules/crew-mode.mdWords only: no code, prompt or fixture changes.
Backwards compatibility
Compatible: documentation text only; nothing reads these lines as data.
Removes nothing: the paragraphs gain one clause each and keep everything they said.
Tests
No new test:
test/test_work_ledger_doc.pydeliberately does not pin prose, and the prompt text these docs describe is already pinned onmainbytest_conductor_patrol.pyand the spec-byte fixtures (#36312).Manual verification
Checked on Linux (Ubuntu 24.04), through the path agents use to read this guide,
search_docs(page="work-ledger")inkiro_crew.mcp_guide:search_docs(query="conductor stops patrolling")still returnswork-ledgeras its first hit. The branch merges intomainwith no conflict. In the merged tree every live copy of the stop rule carries the clause: goal-conductorSKILL.mdstop condition 1,patrol_final_report,patrol_ends,PATROL_MESSAGE, and the two docs here. The only copies without it are the earlier texts kept inSHIPPED_PATROL_MESSAGES.On the same host, the full project gate list (its backend and frontend related-test shards and the docs lint included), semgrep,
test/test_work_ledger_doc.pyandtest/test_search_docs.pyall pass.Real-machine verification
N/A -- the diff changes two Markdown files and runs nothing.
Related Issues
Closes #37192
Checklist
feat|fix|docs|style|refactor|perf|test|chore|ci|build|revert: ...); PR Hygiene does not count commitsContribution License Agreement
Contributed under the project's Apache-2.0 license.