diff --git a/quest/m0/README.md b/quest/m0/README.md index 951092e1..8c692307 100644 --- a/quest/m0/README.md +++ b/quest/m0/README.md @@ -40,12 +40,16 @@ ownership-manifest design: ## Required - [Codex cloud off](/quest/m0/codex-cloud-off.md) - automatic Codex cloud reviews stop once no open PR relies on them +- [Live sandbox test](/quest/m0/review-sandbox.md) - the real reviewer CLIs run nothing a reviewed head plants - [Kix decision UI](/quest/m0/kix/README.md) - planning and implementation through prompts and summaries, run on the maintainer's own runners - [Agent test environment](/quest/m0/agent-test-env.md) - `just test` passes inside agent sessions +- [Stale skill stubs](/quest/m0/init-prune.md) - `quest init` and `quest uninstall` remove stubs of retired skills such as `quest-ready` - [Bun EFAULT report](/quest/m0/bun-efault.md) - condition: Bun fixes concurrent `rmSync` on Windows; report it, then drop the workaround +- [Codex project-config skip](/quest/m0/codex-project-config.md) - condition: Codex keeps a supported way to skip a project's `.codex/` layer; then reviews keep the user's config - [Gate bench](/quest/m0/gate-bench.md) - measure headroom admission on an idle host - [Gate admission on CI](/quest/m0/gate-ci-admission.md) - gated commands admit on loaded hosted runners like arm64 macOS - [Agent lifecycle integration](/quest/m0/worktree-agents.md) - reclaim completed and suspended Claude checkouts through its hooks - [Agent workflow helpers](/quest/m0/helpers/README.md) - CLI commands for CI waits, PR status, claims, quest moves, and serialized merges - [Quest log theme](/quest/m0/theme/README.md) - an MMO-style identity, copy, board, landing page, and CLI output for the demo and launch - [Launch material](/quest/m0/launch.md) - README, quickstart, demo, comparison, and launch-post drafts +- [Review harness check](/quest/m0/review-harness.md) - recurring each milestone: newest pinned review models, and the live sandbox test passes diff --git a/quest/m0/codex-project-config.md b/quest/m0/codex-project-config.md new file mode 100644 index 00000000..54b04ffe --- /dev/null +++ b/quest/m0/codex-project-config.md @@ -0,0 +1,23 @@ +# [XS] Codex keeps a way to skip project config + +## Goal + +Codex offers a supported way to skip a project's `.codex/` layer while +keeping the user's config. Once a release has it and the review harness check +passes with it, `quest review` uses it instead of `--ignore-user-config`, +which drops the user's model provider and profiles, and this quest is +deleted. + +## Plan + +Decided 2026-10-28 (user, /quest-plan of the follow-ups from [kixelated/quest#271](https://github.com/kixelated/quest/pull/271)). + +- [openai/codex#49333](https://github.com/openai/codex/issues/49333) reports + that `--ignore-user-config` also skips the project layer, unlike its help + text. #271 relies on exactly that to keep a reviewed head's MCP servers and + hooks from running. A fix that matches the help text would reopen that hole. +- Do not post upstream, on #49333 or a new issue; the maintainer declined. + Only watch for the change. +- This quest waits on Codex. Check #49333 and Codex release notes each time + ready work is triaged; [the live sandbox test](/quest/m0/review-sandbox.md) + catches a release that changes the behavior first. diff --git a/quest/m0/helpers/README.md b/quest/m0/helpers/README.md index 1730106b..25229a07 100644 --- a/quest/m0/helpers/README.md +++ b/quest/m0/helpers/README.md @@ -63,12 +63,18 @@ commands from Sep-Oct. Session counts in each child come from that window. [kixelated/quest#229](https://github.com/kixelated/quest/pull/229): the gate's rule and setup fixes rank after its adoption, which verifies forks live. +- Decided 2026-10-28 (user), from the follow-ups of + [kixelated/quest#271](https://github.com/kixelated/quest/pull/271): the + delegated follow-up rule ranks right after the review rules, as a prose + rule, and dropping `review-rounds` follows it. ## Required - [quest merge --land gaps](/quest/m0/helpers/merge-land.md) - squash on squash-only repositories, PR URLs outside a checkout, labels released after direct merges, and a warning when auto-merge outlives its head - [Visible merge-driver fallback](/quest/m0/helpers/merge-driver.md) - a merge without `quest` on PATH says so, and this repository's dev shell provides one - [Review rules](/quest/m0/helpers/review-rules.md) - a head whose diff matches the reviewed diff keeps its review +- [Plan delegated follow-ups](/quest/m0/helpers/follow-up-plan.md) - a caller plans a background merge's follow-ups without asking +- [Drop review-rounds](/quest/m0/helpers/review-rounds.md) - `quest status` stops counting GitHub reviews now that `quest review` numbers rounds - [Headless agents](/quest/m0/helpers/headless-agents.md) - `/quest-spawn api` launches, resumes, and stops billed agents through one command auto mode allows - [Sibling PRs through a server-side merge](/quest/m0/helpers/queue-siblings.md) - children of one epic land back to back through a merge queue without conflicting on its link list - [kixelated/quest lands through quest land](/quest/m0/helpers/land-adopt.md) - this repository's `main` changes only through its own gate diff --git a/quest/m0/helpers/follow-up-plan.md b/quest/m0/helpers/follow-up-plan.md new file mode 100644 index 00000000..464f5a3c --- /dev/null +++ b/quest/m0/helpers/follow-up-plan.md @@ -0,0 +1,26 @@ +# [XS] Plan a delegated merge's follow-ups + +## Goal + +A caller that receives a merged PR's suggested follow-ups from a sub-agent, +such as a background `/quest-merge`, runs `/quest-plan` on them without +asking first, and its interview lets the user drop any. One rule in the +guide says this, replacing the copies in `complete.md` and `spawn.md`. + +## Plan + +Decided 2026-10-28 (user, /quest-plan of the follow-ups from [kixelated/quest#271](https://github.com/kixelated/quest/pull/271)). + +- [kixelated/quest#266](https://github.com/kixelated/quest/pull/266) made a + foreground `/quest-merge`, `/quest-complete`, and `/quest-spawn` plan + follow-ups without asking. A single `/quest-merge` run as a background + agent only lists them in its report, and no rule told the caller to plan + them: #271's follow-ups waited for the user to ask. +- The rule belongs in the guide, beside "A suggested follow-up becomes a quest + only through `/quest-plan`", so it covers any skill that delegates a merge. + `complete.md` and `spawn.md` keep only what is specific to them, such as + waiting until every sub-agent has finished. +- `/quest-afk` still holds follow-ups for the user, since it runs no + interview. +- Also update the guide's "`/quest-merge` decides which to plan", which + predates #266. diff --git a/quest/m0/helpers/review-rounds.md b/quest/m0/helpers/review-rounds.md new file mode 100644 index 00000000..d446eb08 --- /dev/null +++ b/quest/m0/helpers/review-rounds.md @@ -0,0 +1,19 @@ +# [XS] Drop review-rounds from quest status + +## Goal + +`quest status` no longer prints `review-rounds` or its JSON +`reviewRounds`. It counts GitHub reviews since a `Design change:` comment, +but the guide now counts design-checkpoint rounds by `quest review` session, +which prints each round's number itself. + +## Plan + +Decided 2026-10-28 (user, /quest-plan of the follow-ups from [kixelated/quest#271](https://github.com/kixelated/quest/pull/271)). + +- Remove the field from `src/cli/status.ts`, `--help`, the README's status + table and example, and the tests. The JSON field's removal is breaking, so + mark the commit `!`. +- Counting local rounds from the PR body's `## Review` record was rejected: + `quest status` does not parse PR body conventions (the helpers epic's + 2026-10-19 decision), and `quest review` already knows its round. diff --git a/quest/m0/init-prune.md b/quest/m0/init-prune.md new file mode 100644 index 00000000..9885d4a8 --- /dev/null +++ b/quest/m0/init-prune.md @@ -0,0 +1,22 @@ +# [S] quest init removes stale skill stubs + +## Goal + +`quest init` and `quest uninstall` remove Quest skill stubs whose skill +no longer exists, such as `quest-ready`, from `.claude/skills` and +`.agents/skills`. They recognize a stub by its Quest-written content, so a +user's own skill with a `quest-` name is never touched. + +## Plan + +Decided 2026-10-28 (user, /quest-plan of the follow-ups from [kixelated/quest#271](https://github.com/kixelated/quest/pull/271)). + +- #271 removed `/quest-ready`, but repositories that ran `quest init` + earlier still list it, and agents invoke it. +- `src/cli/setup.ts` recognizes a stub only by exact match with a current + skill's stub. Recognize a retired one by the stub's shape (frontmatter, then + the ``Run `quest skill ` and follow its output.`` body; the + description varies) for a name `quest skill` no longer has, rather than a + list of retired names. +- Re-running `quest init` after upgrading the pin is how a repository picks + this up; no new check command. diff --git a/quest/m0/kix/land-review.md b/quest/m0/kix/land-review.md index e8920783..3ed08c4d 100644 --- a/quest/m0/kix/land-review.md +++ b/quest/m0/kix/land-review.md @@ -31,6 +31,18 @@ Decided 2026-10-28 (user, /quest-plan trust audit): an AI comment from the author's account, and under protection that requires reviews without dismissing stale ones. +Decided 2026-10-28 (user, /quest-plan of the follow-ups from +[kixelated/quest#271](https://github.com/kixelated/quest/pull/271)): + +- The PR body's `## Review` record is self-reported: anyone who can edit the + body can write one. For a head a runner published, kix counts a local + review only when the runner published its `quest review` result as a + GitHub App check run on that exact head SHA. Only the App can write one, + and GitHub binds it to the commit. A maintainer's own `## Review` record + still counts for the maintainer's PRs. +- Test a body-only `## Review` record on a runner head, which does not count, + and a check run on an older head. + ## Related - [Web execution](/quest/m0/kix/execution.md) - merge jobs this rule gates diff --git a/quest/m0/review-harness.md b/quest/m0/review-harness.md new file mode 100644 index 00000000..240774ec --- /dev/null +++ b/quest/m0/review-harness.md @@ -0,0 +1,26 @@ +# [S] Review harness check + +## Goal + +Each milestone, `quest review`'s pinned models are each vendor's newest for +their tier, and the reviewers still run nothing from a reviewed head. + +## Plan + +Decided 2026-10-28 (user, /quest-plan of the follow-ups from [kixelated/quest#271](https://github.com/kixelated/quest/pull/271)). + +- Compare `MODELS` in `src/cli/review.ts` with each vendor's current models: + the default tier is the newest general model at medium effort, and the deep + tier the strongest one. Bump any that are stale, and check that both CLIs + accept the new ids. +- Run [the live sandbox test](/quest/m0/review-sandbox.md) with the CLIs + installed at the time, and plan a quest for any failure. +- This quest recurs. The PR that finishes this run creates + `/quest/m1/review-harness.md` with this Goal, listed in m1's `Required`, + and `Required: /quest/m0/README.md`, so the next run becomes ready when m0 + completes. Each run creates the next milestone's run the same way, carrying + forward the model ids it found and anything to check first. + +## Required + +- [Live sandbox test](/quest/m0/review-sandbox.md) - the test each run executes diff --git a/quest/m0/review-sandbox.md b/quest/m0/review-sandbox.md new file mode 100644 index 00000000..48d3cd28 --- /dev/null +++ b/quest/m0/review-sandbox.md @@ -0,0 +1,28 @@ +# [S] Live test that a reviewed head runs nothing + +## Goal + +An opt-in recipe runs `quest review` with the real Codex and Claude Code CLIs +on a head that plants configuration, and fails if any of it runs: a +`.codex/config.toml` MCP server, `.codex/hooks.json` hooks, an `.mcp.json` +server, and `.claude/settings.json` hooks. It also fails if the review writes +a trust entry for the reviewed directory into the reviewer's user config. + +## Plan + +Decided 2026-10-28 (user, /quest-plan of the follow-ups from [kixelated/quest#271](https://github.com/kixelated/quest/pull/271)). + +- #271's fake-CLI tests check only that the isolating flags are passed. Its + P1 (a trusted clone's worktree loading the head's `.codex/config.toml` MCP + servers outside the sandbox) lived in CLI behavior that fakes cannot see. +- Each planted item records that it ran, such as a marker file or a request to + a local port, and the test asserts nothing recorded. Run each CLI against a + disposable config home that holds a login and trusts the reviewed clone, so + the trust path is exercised. +- [openai/codex#49333](https://github.com/openai/codex/issues/49333) reports + that `--ignore-user-config` skipping the project layer is a bug, and that + the flag still writes a trust entry. Quest's protection relies on the + former, so this test is what catches a Codex release that changes it. +- It needs real logins and spends API usage, so CI does not run it. + [The review harness check](/quest/m0/review-harness.md) runs it each + milestone, and anyone can run it after upgrading either CLI.