From apple-agent-kit
Run a slim Codex-native closeout review for Apple Agent Kit branches and PRs before commit, PR, or merge.
How this skill is triggered — by the user, by Claude, or both
Slash command
/apple-agent-kit:codex-autoreviewThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
Use this skill when the user asks for autoreview, Codex review, PR closeout review, or a rigorous review before commit, push, PR, or merge.
Use this skill when the user asks for autoreview, Codex review, PR closeout review, or a rigorous review before commit, push, PR, or merge.
This is code review, not approval routing. Review output is advisory until verified against the real files.
Before reviewing, freeze:
Classify findings as:
in-scope blocker: introduced by this branch, concrete, and fixable without changing the PR purpose;follow-up: real but outside this branch's owner boundary;stop-and-escalate: requires a new public contract, private adapter decision, credentials, hardware dispatch, app launch, screenshots, physical-device automation, or release-process change.Fix only in-scope blockers. After any fix, rerun focused validators and repeat review only if the reviewed diff materially changed.
For public repo changes, confirm:
python3 scripts/validate-release.py passes;git diff --check passes;For private adapter changes, do not run hardware actions unless the user explicitly approved that specific dispatch. Planning and static review are allowed; install, app launch, UI smoke, screenshots, product build/test, physical-device automation, and Codex-session bridge execution are gated.
Lead with findings, ordered by severity, using file and line references. Then list validators run and end with one verdict:
readyneeds-fixblockedIf there are no issues, say so clearly and mention any residual risk or unavailable proof.
npx claudepluginhub neonwatty/apple-agent-kitCreates structured, bite-sized implementation plans from specs or requirements before writing code. Useful for breaking down multi-step tasks into testable steps with file structure and task boundaries.