Review Loop

This runs a bounded review-fix-re-review cycle on either plans or implementations. You can specify reviewers (code, security, accessibility, etc.), set a severity gate to control what gets fixed versus recorded, and cap the rounds. The interesting bit is zero trust: reviewers stick around until they individually see no critical issues, so you can't game your way out early. Recurring findings escalate in severity automatically. It exits when everyone's satisfied or you hit the round cap, then gives you a verdict table showing what was fixed and what got punted to issues. Use it when you want structured iteration without babysitting each round yourself.

Install to Claude Code

npx -y skills add camacho/ai-skills --skill review-loop --agent claude-code