Repository navigation
docs: record the quiet-end token decision superseding the turn-end entry - #32257
Conversation
First Principles Review (Opus 5.5) — ⏭️ skippedRevision |
UX Review (Opus 5.5) — ⏭️ skippedRevision |
Design Review (Opus 5.5) — ✅ PASSDesign-level review of Design-Verdict: PASS This adds one decision entry and its index row and changes no code, so nothing needs a design check. [DESIGN-REVIEWED] 00dde56 Description read[DESCRIPTION-READ] a44b85682c785bb8012286d62c6f29eb82632b6e39a8128196f0bfaf1d2fcaa1 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 — ✅ human override acceptedReviewed Human judgment by @bolichen97 overrides the Opus 5.5 finding for [OPUS-OVERRIDE] 00dde56 Verdict recorded from an authorized human decision 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: |
8dce55c to
bb64218
Compare
|
self-added: no
[operator: bolichen97] |
|
self-added: no
[operator: bolichen97] |
bb64218 to
7853422
Compare
|
[operator: bolichen97] Rebased onto main a0d9a5a at 7853422 to clear the conflict in The GPT and Opus BLOCKs on bb64218 are the |
|
self-added: no
[operator: bolichen97] |
b9c8551 to
63f7e01
Compare
|
Rebased: main moved in docs/decisions/README.md (a merge conflict: main appended an index row where this entry's row sat) and other docs/decisions/ entries since this head went green (265b3ac -> ce89396); CI re-running. The entry file is byte-identical; its index row now sits directly under the 2026-10-03 row it supersedes, off the table tail other decision PRs append to, and |
|
self-added: no
[operator: bolichen97] |
|
Re-sync budget reset: this PR is inside its 48-hour watch window, and its only reds on 63f7e01 (Backend Tests 3.12 shards 5 and 6, Windows shards 2 and 4, and the Coverage Gate that follows them) are main's own: |
Add the maintainer's decision for #19276 as a new docs/decisions entry: after nothing_to_do the model writes only a fixed end token, which Kiro Crew drops where the model's text first arrives. It supersedes, in part, the 2026-10-03 turn-end entry, which stays byte-identical, and is indexed beside it.
63f7e01 to
00dde56
Compare
|
Rebased: main moved in docs/decisions/README.md and two other docs/decisions/ entries since this head went green (ce89396 -> 2f7e9f2); CI re-running. The re-sync picks up #35625, which fixes the two Backend Tests reds main carried at ce89396 ( |
|
self-added: no
[operator: bolichen97] |
|
/ai-review override fable 00dde56 owner-decision: agent: this entry deliberately supersedes, in part, the 2026-10-03 turn-end decision, as docs/decisions/README.md rule 3 requires for a change of mind; the 2026-10-03 entry stays byte-identical. Decision: #32257 (comment) [operator: bolichen97] |
Human judgment recorded@bolichen97 recorded a owner-decision release (class
This decision applies only to this commit. A new push requires a new judgment. |
KiroCrewCaptain
left a comment
There was a problem hiding this comment.
Review of head 00dde56. The diff is docs only: one new decision entry and one index row in docs/decisions/README.md. The 2026-10-03 entry has no hunk, so it stays byte-identical. The Supersedes: line quotes the replaced clause word for word and uses the existing partial-supersede form. Each Evidence link resolves to a record that matches the entry: the 190 and 323 counts in #27655, and the decision comment on #19276 from a listed maintainer. The entry also ships in its own PR, apart from the code change, as rule 3 asks. PR Readiness is green on this head.
Done well: the supersession is narrow. It replaces one quoted clause and says that the rest of the earlier contract, including the recovery ladder, still holds.
Optional: the override that clears the rule-3 flag on this head carries the agent: prefix. A maintainer may want to restate it in person. An acknowledgement from @CrysisDeu, who decided the earlier entry, would also make the record clearer.
Optional: until #32302 lands, the ledger describes end-token behaviour that main does not have yet. #32302 now has a merge conflict, so it is best to land the two close together.
Problem / Motivation
Goal: Record the maintainer's decision that after
nothing_to_dothe model writes only a fixed end token that Kiro Crew drops, as a new entry that supersedes the 2026-10-03 turn-end entry.The 2026-10-03 entry says a quiet turn ends on "a
nothing_to_docall with nothing after it". That leaves the backend's history ending on a tool result, which the backend fills with its own placeholder; the model then copies the placeholder after later quiet ends. #19276 asks for a quiet end that leaves the history on a real assistant message, and the maintainer decided how on the issue. Changing a recorded decision takes a new entry in its own pull request (docs/decisions/README.md, rule 3).Why it matters
Without the superseding entry, the code change for #19276 reverses a recorded decision, and the
recorded-decisions-are-not-regressionsrule rightly blocks it. With it, the ledger says what the product does, so later reviews and triage hold the new turn-end shape instead of the old one.Not a goal
What changed (motivation → approach → change)
One new entry,
docs/decisions/2026-11-07-a-quiet-turn-ends-with-an-end-token-kiro-crew-drops.md, with a partialSupersedes:line naming2026-10-03-every-agent-turn-ends-with-text-or-nothing-to-do.md(in part: "anothing_to_docall with nothing after it"), and its index row indocs/decisions/README.md. The entry's decision: afternothing_to_dothe model writes only a fixed end token, which Kiro Crew drops where the model's text first arrives, so no chat, channel, cron or sub-agent result shows it; every other rule of the superseded entry stands, including the recovery ladder for a bare stop after an ordinary tool. Its evidence is the maintainer's decision comment on #19276.The shipped 2026-10-03 entry is not edited (entries are immutable;
scripts/check_decisions_history.pypasses).Backwards compatibility
Compatible: a documentation entry only; nothing that runs reads it.
Removes nothing: the superseded entry stays in the ledger byte-identical, and this pull request only adds a file and one index row.
Tests
Proof: not provable -- docs or config only
DECISIONS_BASE_REF=origin/main python3 scripts/check_decisions_history.pyandscripts/docs-lint.shpass.Manual verification
N/A — a documentation entry; the two gates above are its checks.
Related Issues
Part of #19276
Maintainer decision: #19276 (comment)
This pull request carries the superseding entry, so the
recorded-decisions-are-not-regressionslane blocks it by design (docs/decisions/README.md, rule 3); a maintainer override is what lets it land.Checklist
feat|fix|docs|style|refactor|perf|test|chore|ci|build|revert: ...)