Repository navigation
quest: plan moq's 2026-11-03 /quest-complete follow-ups - #453
Conversation
Three follow-ups from running quest's skills on moq-dev/moq: - The land fixer epic gains two quests: the watcher leaves a PR another session holds alone, and the fixer runs the repository's own checks, such as moq's JavaScript type check, before it requeues. - /quest-complete offers no grouped or blanket pick across PRs. - The decision-records quest also says where a decision goes when the PR deletes its quest: the PR's Decisions instead of the quest Plan. 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 59 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. CodeRabbit Verify
|
There was a problem hiding this comment.
✅ No new issues found.
Reviewed changes: planning docs for three follow-ups that running quest's skills on moq-dev/moq turned up. This is a docs-only change; I verified every source-code and skill claim each quest makes, and they all hold up:
holds.md—isHoldLabeldoes live atsrc/core/hold.ts:27and accepts anyquest:*label exceptquest:merge, soquest:activeand an older pin'squest:iterateboth hold, as the goal states. The two watcher files (src/cli/land-watch.ts,src/cli/land-watch-fix.ts) really never read PR labels for holds. TheendPickrecommendation matches the code:src/cli/land-watch-fix.ts:143ends a pick "as a writer's push or a person's removal does", and the proposed open question for the starter (end the pick vs. wait out the hold) has enough context to answer.checks.md—assets/land-fix.md:51says exactly "Before reportingfixedorrequeue, build and test the packages the PR and your fix touch", which supports the proposed cause for the moq-dev/moq#6949 miss. The fallback for a repository that names no check command is spelled out, so the fixer's prompt stays actionable.complete-per-pr.md—assets/skills/complete.md:20does say "one question per PR, never grouping PRs", and line 30 frames a blanket answer as valid, so the diagnosed tension between the two paragraphs is real. The guidance to also checkunblock.mdandafk.mdpoints at skills that do put PR picks to the user.decision-records.md— folding the where-to-record-a-decision work in here is apt:plan.md:26andunblock.md:28both send decisions to "the quest's Plan" while a quest-completing PR deletes that file, anditerate.md:19records in the PR description without naming a## Decisionssection. The quest name and README/m0-index entries were updated to match the widened goal.- READMEs — the m0 index entry and the land-fixer epic's
Requiredlist both follow the existing link-list style, and the ranking claim ("land work first, then Kix") matches m0's recorded 2026-11-01 decision.
quest check --base origin/main passes on this branch: 69 documents ok.
glm-5.3-flash | 𝕏
moq-dev/moq runs its worktree hooks through its flake's pinned quest (`nix run ...#quest -- worktree hook`), and `quest init` refuses it because it matches only the exact `quest worktree hook`. The maintainer decided init treats any command hook ending in `worktree hook` as its own. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
There was a problem hiding this comment.
✅ No new issues found.
Reviewed changes: incremental re-review of the new commit on this PR since the prior Pullfrog review (27f8241). All documentation-only.
- Added
quest/m0/wrapped-hooks.md— an [XS] quest forquest initaccepting any command hook whose arguments end inworktree hook(moq-dev/moq'snix run "${CLAUDE_PROJECT_DIR}#quest" -- worktree hookwrapper). It carries the maintainer's decision record (wrapped match ✅; configured hook command and no-change rejected), implementation guidance forisQuestHookandquest uninstall, and a test plan covering init, re-init, and uninstall. - Indexed the new quest in
quest/m0/README.mdimmediately after [Worktree hooks here], matching the doc's stated rank. - Appended a missing trailing newline in
quest/m0/land-fixer/holds.md— no content change.
I re-verified the new doc's claims against this checkout: isQuestHook (src/core/init.ts:240) does count wrapped hooks as someone else's by matching only handler.command === HOOK_COMMAND, and uninstall shares that exact check (src/cli/setup.ts imports the same match at line 30 and filters with it at line 242). quest check --base origin/main passes on the head commit: 70 documents ok.
(written by GLM-5.3-Flash)
glm-5.3-flash | 𝕏

Problem
Running quest's skills on moq-dev/moq on 2026-11-03 turned up four gaps. The maintainer approved planning them:
/quest-iteratehold was active. It also requeued without running moq's JavaScript check, and the next build failed on it./quest-completeoffered a grouped "merge all 20 landable PRs" option, even though the skill already says to ask one question per PR. The maintainer: "Ask per PR. Update quest to always ask per PR."nix run "${CLAUDE_PROJECT_DIR}#quest" -- worktree hook, so agents use thequestmoq pins (moq-dev/moq#7674).quest initrecognizes its hook only by the exactquest worktree hook, so it refuses to run in moq. The maintainer decided init treats any command hook ending inworktree hookas its own (✅ that; rejected: a configured hook command, no change).Changes
quest/m0/land-fixer/holds.md(new, [S]):quest land watchchecks for a hold before it starts a fixer job and again before it pushes or re-enqueues. Todaysrc/cli/land-watch.tsandland-watch-fix.tsnever read hold labels. The frozen batching gate is out of scope.quest/m0/land-fixer/checks.md(new, [S]):assets/land-fix.mdmakes the fixer run the repository's own check and test commands, such asjust checkandjust test. Today it says only "build and test".quest/m0/complete-per-pr.md(new, [XS]):complete.mdforbids offering any option that covers more than one PR. A blanket answer the maintainer volunteers still counts.quest/m0/decision-records.md(folded in): a decision goes in the quest Plan, or in the PR's## Decisionsif the PR deletes the quest.quest/m0/wrapped-hooks.md(new, [XS]):isQuestHook(src/core/init.ts) accepts any command hook ending inworktree hook.quest uninstallshares that check, so it removes wrapped hooks too.Choices made without asking
plan.mdandgrill.md./quest-completeand the land fixer every day.quest/m0/worktree-hooks.md, installs the hooks in this repository and is held by moq-dev/quest#423, which uses the plain command. No quest covers how init matches its hook.Search: the quest tree, git history for deleted quests, and open PRs. moq-dev/quest#452 touches
assets/AGENTS.mdandspawn.md, but for follow-up planning, so it does not overlap. No other overlap.Validation
quest check --base origin/main(built from this checkout): 70 documents ok.just checkpasses.Review
quest review --base origin/main, Codexgpt-6.1-sol, medium effort. A new session started because the base's instructions changed. Round 1 reviewed head80b80520021c6850fd23b675d75128dc986e01f7: no findings.(Written by Claude Opus 5.5)
🤖 Generated with Claude Code