diff --git a/assets/skills/merge.md b/assets/skills/merge.md index b8179e68..e4e38e08 100644 --- a/assets/skills/merge.md +++ b/assets/skills/merge.md @@ -40,3 +40,8 @@ Ask instead when it refuses a head `/quest-iterate` left unchanged, and never pa Report once it is queued, and do not wait for it to merge. A PR that left the queue unmerged shows why under `landing-stopped` in `quest status `; it needs the user's pick again, except where `/quest-run` says an earlier pick still holds. Never close a PR to unstick it, and never work around a refused merge; ask instead. + +Ask the user for the pick with your recommended action, one of `/quest-merge`, `/quest-iterate`, skip, or `/quest-delete`, plus `/quest-merge` when you recommend another, "Explain it first", and "Rethink it". +"Explain it first" walks them through the change and its open choices; ask again without it. +"Rethink it" starts a fresh background sub-agent, given the PR, its reviews and threads, and its quest's Goal, Plan, and evidence (read the deleted file on the base), not your recommendation, to start from scratch: brainstorm other approaches to the problem, decide whether this PR is the best path forward, and recommend merging it, iterating toward a named alternative, or dropping it. +Ask again without it, its recommendation beside yours, and where they differ. diff --git a/assets/skills/review.md b/assets/skills/review.md index 09dbcd28..5f38ccb7 100644 --- a/assets/skills/review.md +++ b/assets/skills/review.md @@ -48,7 +48,7 @@ Once the reset passes or the login is restored, the item has cleared: whoever fi Push, resolve each `duplicate` line `quest status ` then prints as the guide's Creation describes, then mark the PR ready for review: with `quest ready `, which also drops its `## Waiting` section, or with `gh pr ready ` for a draft you reviewed without a hold. Summarize the review, with its last round's score and rationale, and CI from `quest status `. -Name any decisions the review raised without answering them, then ask the user with your recommended action: +Name any decisions the review raised without answering them, then ask the user for the pick as `/quest-merge` describes, recommending one of: - `/quest-merge`: the PR is landable as `/quest-merge` defines it, or will be once fixed with no decision left for the user, such as failing checks or unaddressed automated review findings; `/quest-merge` runs `/quest-iterate` on it first. - `/quest-iterate`: the review raised significant issues or decisions, which it settles with the grill skill's interview. - skip: the PR should not be merged yet. diff --git a/assets/skills/run.md b/assets/skills/run.md index a553bd76..cf7e6404 100644 --- a/assets/skills/run.md +++ b/assets/skills/run.md @@ -48,9 +48,9 @@ That covers only tests, benchmarks, CI, internal docs or comments, and behavior- A trivial PR, such as a dependency bump, carries it too whatever its author, under the same opt-out, but merges only once Pullfrog has approved its head; in a repository without Pullfrog, it needs the user's pick. Judge the diff, not the title, before every `/quest-merge` run, and ask when unsure: the standing pick ends at the first head that does not qualify, or once a person takes the PR out of the queue. -Ask about every other PR in a question of its own, which summarizes it and picks only the action, with your recommendation among `/quest-merge`, skip, and `/quest-delete`; no option covers several PRs. -Ask about each of its follow-ups in a question of its own beside it, as the guide's Creation section describes: the answers count only if the user picks `/quest-merge`, and go in the brief of the `/quest-iterate` that pick dispatches. -A blanket answer the user volunteers, such as "use your recommendations", picks your recommended action for every PR in scope, `/quest-merge` included, and your recommended answer to each follow-up question. +Ask about every other PR in a question of its own that summarizes it and picks only the action, with `/quest-merge`'s pick options; none covers several PRs. +Ask about each of its follow-ups in a question of its own beside it, as the guide's Creation section describes; the answers count only on a `/quest-merge` pick, and go in the brief of the `/quest-iterate` it dispatches. +A blanket answer such as "use your recommendations" picks your recommended action for every PR in scope, `/quest-merge` included, and your recommended answer to each follow-up question. Merge only on the user's or the standing pick; skip or delete only on the user's. A skip stands until the user lifts it. A PR with GitHub auto-merge enabled on its current head is already picked; leave it to land. Review comments are input, not approval, even a merge verdict on the maintainer's account. diff --git a/quest/m0/pick-options.md b/quest/m0/pick-options.md deleted file mode 100644 index f153f538..00000000 --- a/quest/m0/pick-options.md +++ /dev/null @@ -1,20 +0,0 @@ -# [S] Merge picks offer "Explain it first" and "Is this the best approach?" - -## Goal - -The merge-pick question `/quest-run` and `/quest-complete` ask about each PR offers "Explain it first" and "Is this the best approach?" beside the recommended action, in place of fixed skip and `/quest-delete` options. Skip or delete is an option only when it is the recommendation; otherwise the user types it under Other. "Explain it first" walks the user through the PR's change and open choices, then asks again. "Is this the best approach?" has a fresh sub-agent work out the best approach to the PR's problem and compare it with the PR, then asks again with its finding. - -Non-goals: other decision prompts, such as grill rounds, plan questions, and follow-up questions; the harness's question tool, which takes 2-4 options plus Other. - -## Plan - -Decided 2026-11-17 (maintainer, `/quest-plan`), from the maintainer's answers to agents' blocking questions since 2026-10-15 (`quest messages --json` on the desktop, fedora, and kixtop hosts: 16,237 answers; nixos's run failed on an old Codex rollout): - -- On the 4,527 merge-pick questions, the maintainer picked merge 84% of the time, iterate or a similar action 8%, skip 3.0%, and delete 0.5%. They typed an answer 3.9% of the time, and about half of those typed answers were questions such as "is this worth it?", "why is this needed?", or "is this the right approach?". 111 of the 161 skip or delete picks were the agent's own recommendation. -- Where a prompt offered a walk-through, the maintainer took it 14 times in 22, including "Walk me through them" 11 times in about 15, often over a recommended "Accept all". Verify- or investigate-first options were taken 9 times in 27. -- Scope? ✅ merge-pick prompts only, with skip or delete offered only as the recommendation (rejected: every decision prompt, where 2-3 real choices already fill the 4-option limit; dropping skip and delete as options even when recommended, since most of their picks were the recommendation). -- The second option? ✅ "Is this the best approach?", which matches the typed pushback above (rejected: "Are you sure?", which asks only to re-check the coordinator's call rather than the PR's design; changed by the maintainer the same day, after [#764](https://github.com/moq-dev/quest/pull/764) planned it). -- What answers it? ✅ a fresh sub-agent given the PR, its reviews, and threads, but not the coordinator's recommendation. Without reference to the PR's code, it states the problem the PR solves, works out the best long-term approach, and compares it with the PR, recommending to merge, iterate toward that approach, or drop the PR, as `/quest-review`'s checkpoint does. It runs in the background like any dispatched work, and the PR is asked again once it reports (rejected: the coordinator answering inline, which stays anchored on its own call; a new `/quest-review` round, which reviews the code rather than the approach). -- The change belongs in `assets/skills/run.md`'s Pull requests part, whose pick question `/quest-complete` reuses. This quest authorizes that edit, which `AGENTS.md` otherwise protects. - -Recommendation for the implementer: after "Explain it first", ask again without that option; after "Is this the best approach?", ask again with the sub-agent's approach and verdict beside the original recommendation, saying where they differ. Reuse the checkpoint's wording rather than restating it.