Skip to content

docs(rfc): Preserve Autopilot incident evidence - #16146

Merged
iamwhatever merged 1 commit into
kirodotdev:mainfrom
junjiequ:docs/autopilot-evidence-final
Oct 2, 2026
Merged

iamwhatever merged 1 commit into
kirodotdev:mainfrom
junjiequ:docs/autopilot-evidence-final

Conversation

@junjiequ

@junjiequ junjiequ commented Oct 2, 2026

Copy link
Copy Markdown
Contributor

Problem / Motivation

Goal: Preserve the evidence from six Autopilot timeout incidents after the feature's retirement, without overstating how much cleanup has landed.

Autopilot's 30-minute stage timeout repeatedly told users that work had stopped while native ACP work continued. The user-visible mode has since been removed, but the incident record is not on main, the generic cancellation defect remains open, and the retirement RFC still describes the pre-removal state.

Why it matters

Two recovered incidents show concrete loss of control: one successful build result became unreachable from the dashboard, and another native turn began three source writes after the dashboard said the stage stopped. Preserving sanitized evidence keeps issue #14058 grounded in observed lifecycle failures without publishing raw transcripts or project content. Accurate RFC status also prevents maintainers from treating unfinished StageBoundary cleanup as complete.

Not a goal

What changed (motivation → approach → change)

Area Current record
rfc-autopilot-stage-budgets.md Preserved as a superseded proposal and historical incident record, not current product direction.
assets/autopilot-stage-timeout-evidence.md Records six deduplicated timeout incidents, two confirmed dashboard/native continuations, provenance hashes, and a deterministic reproduction without raw transcripts.
rfc-retire-chat-autopilot.md Records Rollout steps 1 and 2 as merged and keeps status partial because step 3 still has StageBoundary / _in_stage_execution references under src/.
docs/request-for-change/README.md Adds both audited rows while preserving newer RFC entries, including Crew Log Wake.
assets/README.md Indexes the Markdown evidence appendix.

The superseded RFC points the remaining generic four-hour turn-cancellation defect to issue #14058 and explicitly forbids restoring the retired mode. The retirement RFC links back to that historical record.

Backwards compatibility

Compatible: this changes documentation only and does not alter runtime behavior, persisted data, configuration, or public APIs.

Tests

  • python3.12 scripts/docs_lint.py --test
  • ./scripts/docs-lint.sh — 394 Markdown files scanned; all documentation checks passed.
  • git diff --check

Manual verification

Verified on main at 6916ba17f that #15362 and #15359 are merged, issue #14058 remains open, Autopilot's visible surfaces are absent, and StageBoundary / _in_stage_execution references remain in 16 files under src/. The appendix contains no raw transcript rows, local paths, prompts, tool arguments, or credentials.

Screenshots / video

N/A — this PR changes RFC and evidence text only; it has no user-visible UI delta.

Related Issues

no linked issue: This PR preserves evidence and references #14058; it does not implement or close that runtime defect.

Supersedes #15987 and #14043.

Checklist

  • At most two commits (one is the norm), with a Conventional Commits title (feat|fix|docs|style|refactor|perf|test|chore|ci|build|revert: ...)
  • Existing relevant tests pass; no runtime functionality is added.
  • Self-review completed; documentation follows project conventions.
  • Documentation updated.
  • No secrets, credentials, or internal references in the diff.

Retain the sanitized timeout and split-brain evidence after Autopilot retirement. Record the retirement as partial until StageBoundary cleanup lands, and link the remaining generic cancellation defect to issue kirodotdev#14058.
@junjiequ
junjiequ requested a review from a team as a code owner October 2, 2026 11:07
@junjiequ
junjiequ requested a review from bolichen97 October 2, 2026 11:07
@github-actions github-actions Bot added fork Pull request from a fork (external contributor) readiness: action required A blocking check or review needs attention labels Oct 2, 2026
@github-actions github-actions Bot added readiness: checking Automated validation is still running and removed readiness: action required A blocking check or review needs attention labels Oct 2, 2026
@github-actions

github-actions Bot commented Oct 2, 2026

Copy link
Copy Markdown
Contributor

GPT 5.6 Review (fork) — ✅ no blocking findings

Reviewed b9117f143706108592ecc2415a23e76e3edc7aec via the fork AI-review pipeline; updated in place on each push.

Review details

No findings.
[GPT-REVIEWED] b9117f1

@github-actions

github-actions Bot commented Oct 2, 2026

Copy link
Copy Markdown
Contributor

Design Review (Fable 5, fork) — ✅ PASS

Design-level review of b9117f143706108592ecc2415a23e76e3edc7aec via the fork AI-review pipeline — updated in place on each push. A BLOCK verdict blocks PR readiness; PASS/CONCERNS are advisory.

Design-Verdict: PASS

Verify the appendix's live claim — the four-hour turn ceiling still skipping native session/cancel at 6916ba17f — before the record ships.

I spot-checked the committed facts: StageBoundary / _in_stage_execution do occur in exactly 16 files under src/, the retirement RFC's unchanged Rollout section matches the accepted→partial flip (steps 1–2 merged, step 3 open), the assets/ directory exists and gains its first index, and the docs-only diff renders no UI surface, so the rendered-evidence rule does not apply. The historical sections are clearly dated and banner-marked superseded; the one assertion made against current main (ceiling cancellation without native session/cancel, the premise keeping #14058 open) is plausible from _bounded_turn's ceiling path but is the single fact a maintainer should confirm, since the document commits it as verified.

[DESIGN-REVIEWED] b9117f1

Description read

[DESCRIPTION-READ] a565beb8f018b4542dc335505c275a45ffea5545a764b281b386651935f1c570

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 .github/scripts/pr-description-capture.sh over the same inputs; if it differs, the description or that evidence changed after this verdict was formed and any finding drawn from either is unproven.

@github-actions

github-actions Bot commented Oct 2, 2026

Copy link
Copy Markdown
Contributor

Opus 5 Review (fork) — ✅ no blocking findings

Reviewed b9117f143706108592ecc2415a23e76e3edc7aec via the fork AI-review pipeline; updated in place on each push.

Review details

No findings.

[OPUS-REVIEWED] b9117f1

@github-actions github-actions Bot added readiness: passed Eligible automated validation passed for the current revision and removed readiness: checking Automated validation is still running labels Oct 2, 2026

@iamwhatever iamwhatever left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Reviewed: docs-only under docs/request-for-change/. The stage-budgets RFC lands as superseded with a matching superseded-by/supersedes pair, and the retire RFC moves accepted -> partial backed by merged #15362/#15359 and audited-at 6916ba1, so no new acceptance is claimed. Evidence appendix carries no prompts, paths, URLs or credentials. Nit: assets/README.md has two H1s; the second should be ##.

@iamwhatever
iamwhatever enabled auto-merge (squash) October 2, 2026 16:44
@iamwhatever
iamwhatever merged commit 26e26fc into kirodotdev:main Oct 2, 2026
99 checks passed
@github-actions github-actions Bot removed the readiness: passed Eligible automated validation passed for the current revision label Oct 2, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

fork Pull request from a fork (external contributor)

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants