Autoreview

Runs structured code review from Codex (default) or five other engines before you commit or ship. The contract is blunt about what matters: verify every finding yourself, reject speculative edge cases and broad rewrites, fix the bug class not just the instance, and keep running until nothing actionable comes back. It handles dirty work, branches, PRs, and single commits, supports parallel test runs during review, and can take up to 30 minutes on large bundles. The guidance on regression provenance and automerge bot attribution is unusually detailed. Think of it as a paranoid reviewer that reads docs and checks dependencies, but you still own the accept/reject call on every line.

Install to Claude Code

npx -y skills add openclaw/agent-skills --skill autoreview --agent claude-code