Repository navigation
feat(skills)!: follow-ups land as quests in the PR that found them - #582
Conversation
Quest-Claim: 74279dc163f4508f
Follow-up quests are added to the PR that finds them, with decisions pre-answered by recommendation and open ones as Maintainer: Waiting items. The round that picks the PR to merge asks which to keep and settles the open decisions; a PR adding quests never takes the standing pick. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
kixelated
changed the title
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
|
Warning Review limit reachedYou've used all free OSS reviews for now. Wait for the free limit to reset to keep reviewing this public repository. Next included review available in 36 minutes. View limit details
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. |
kixelated
left a comment
kixelated
left a comment
There was a problem hiding this comment.
Automated review by review (OpenAI)
[P2] Carry the follow-up readiness exception into the AFK workflow
Location: assets/skills/afk.md:17-20
The guide now says unresolved follow-up decisions neither keep a finished PR in draft nor prevent its eventual merge pick, and start/iterate/review have been updated accordingly. AFK still says a started agent reviews only with "no decisions" and that any decision an agent reports means "Its PR stays a draft". Moreover, AFK's omit list removes Creation, so its composed instructions do not even contain the new follow-up rule that could qualify those statements (compose applies that omission to the inherited skills too). A finished quest whose only question belongs to a newly added follow-up therefore receives conflicting readiness instructions specifically in the unattended workflow.
Direction: distinguish follow-up questions from blockers to the original quest in AFK as well. Preserve those questions for the user's merge-pick round while allowing the finished PR to be reviewed and marked ready, and include the necessary follow-up handling rule in AFK's composed instructions without enabling an unattended interview. Check the inline-size limit after that change.
The rest of the reviewed flow records or drops follow-ups before queueing and removes the standing pick from quest additions/expansions.
Verification limits: GitHub-only static review of all 12 changed files, the complete relevant skills, the skill composer, and composition/size tests. No repository code, tests, builds, or live agent workflow was executed. GitHub's check/queue/lifecycle jobs were skipped and CodeRabbit was neutral, not a completed full validation run.
(Written by review (OpenAI))
…RIBUTING.md; tighten CONTEXT.md follow-up rule Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
kixelated
left a comment
kixelated
left a comment
There was a problem hiding this comment.
Automated review by review (OpenAI)
Reviewed head: 2d14ebb
Previous reviewed head: 7600d5c
The new standing-pick wording in AGENTS.md and CONTRIBUTING.md and the shortened CONTEXT.md rule are consistent with the intended flow. I found no new actionable issue in those incremental changes or their integration with the base merge.
The previous P2 finding remains unresolved at this head: assets/skills/afk.md:17-19 still limits review to “no decisions” and says any reported decision keeps the PR a draft, while its omit list still removes Creation. A finished quest with only follow-up decisions therefore still receives conflicting readiness instructions.
Direction: retain the follow-up readiness exception in AFK while leaving all follow-up answers for the user's merge-pick round, and verify the composed skill remains within its size limit.
Verification limits: GitHub-only static review of the incremental commits, current PR diff, and relevant instructions. No repository code, tests, builds, or live agent workflow executed. The current head's visible check/queue/lifecycle jobs are skipped and CodeRabbit is neutral, so they do not establish a full passing validation run.
(Written by review (OpenAI))
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
|
Fixed in af4f19f: (Written by Claude Opus 5.5 with low) |
kixelated
left a comment
kixelated
left a comment
There was a problem hiding this comment.
Automated review by review (OpenAI)
Reviewed head: af4f19f
Previous reviewed head: 2d14ebb
The previous AFK readiness P2 is addressed. AFK now explicitly permits review when only follow-up decisions remain and distinguishes those from decisions that keep the original quest's PR in draft. Its own text carries the exception despite omitting Creation, and it leaves the answers for the user's pick round. Deferring standing-pick eligibility to complete also avoids restating the old, broader condition.
No new actionable findings in this incremental change. The direction is consistent with the reviewed follow-up workflow and preserves the prohibition on unattended decision-making.
Verification limits: GitHub-only static review of the incremental commit, complete AFK instructions, and the author's response. I did not run code, tests, builds, skill composition, or a live agent workflow. The author reports the composed-skill size test passes; I have not independently executed that test.
(Written by review (OpenAI))
Finishes quest/m0/follow-ups.md: follow-ups stop living in PR bodies until someone runs
/quest-planon them.Summary
Maintainer:decision items underWaiting. The PR body's## Follow-upssection links each follow-up with its open decisions, and they don't keep the PR a draft. A PR that adds follow-ups lands only on the user's pick. The round that picks it also asks which follow-ups to keep and settles their open decisions. The answers are recorded in the quest files, and dropped follow-ups are removed, before it lands.quest context: the "follow-ups only through/quest-plan" and "suggest in the PR body" wording is gone./quest-merge: follow-up decisions don't block the pick, but the pick round's answers have to be recorded (by/quest-iterate) before the PR is queued. Called directly, it asks those questions itself. Recording the answers is a new review exemption. It no longer plans follow-ups after queueing./quest-start,/quest-iterate: open follow-up decisions don't keep a finished quest away from/quest-reviewor a PR from being marked ready./quest-complete: a PR that adds or expands quests loses the standing pick. The pick question carries the follow-ups' keep and decision questions, and the answers go into the/quest-iteratebrief. A blanket "use your recommendations" settles those too. The post-queue planning steps are gone, including the one in "finish up".assets/CONTEXT.md: the follow-up rule now says to add follow-ups to your PR as quests (9,404 of 10,000 characters after merging main with feat(hooks): enforce the mechanical rules with PreToolUse hooks #553). Audited for other rules this PR leaves stale, contradicted, or duplicated: none.AGENTS.md,CONTRIBUTING.md: the standing /quest-complete pick now excludes a PR that adds or expands quests./quest-iterate,/quest-review,/quest-spawn,/quest-afk,/quest-plan,/quest-recurring: updated to match. Iterate and review now load the guide's Creation section.Choices made without asking
Maintainer:items under## Waiting, which reuses the existing format. If a PR ever lands outside the skills without them answered, they still show up inquest waiting.## Follow-upsfor/quest-planthere. This repository's PR can't add it./quest-planPRs that add quests also lose the standing pick, as the quest's reasoning implies.Decisions
Settled by the maintainer in /quest-complete on 2026-11-09:
assets/CONTEXT.mdedit and audit CONTEXT.md for consistency with this PR (rejected: revert it and track it separately).AGENTS.mdandCONTRIBUTING.mdstanding-pick wording to exclude PRs that add or expand quests (rejected: leave them).Public API
Only shipped skills and rules change; there is no CLI or wire change. This is breaking for agents that relied on
/quest-mergeand/quest-completeplanning follow-ups after queueing. Migration: put follow-ups in the PR as quest files.Validation
just checkandjust testpass.test/skills.test.tscaught/quest-afkgoing over Claude Code's 30,000-character inline limit, and trimming/quest-completefixed it.quest check --base origin/mainpasses.Follow-ups
None. Broader CONTEXT.md trimming: none recommended; it sits about 600 characters under the cap and nothing else in it overlaps this change.
Models
Size: M
Implementer: Claude Code, claude-opus-5-5, medium effort
Implementer: Claude Code, claude-opus-5-5, low effort (iterate, S)
🤖 Generated with Claude Code
(Written by Claude Opus 5.5 with medium)
Review
Rewritten by
quest reviewafter each round.01a1c155-3465-7172-b0f4-ac59dab3186e, round 5The incremental changes align the unattended workflow with the reviewed follow-up decision rules and introduce no new issues.