claude-review-loop

Why we included this project

Claude Code users who want an independent check on their work before merging can get it without leaving the terminal. You describe a task, Claude implements it, and a stop hook then hands the diff to OpenAI's Codex, which runs up to four parallel review agents: one on the git diff for code quality, test coverage, and security, one on project structure and architecture, plus optional Next.js and browser UX checks when the project calls for them. Codex deduplicates the findings into a single review file, and Claude reads and addresses them, so the code gets a second opinion from a different model rather than a self-review. The plugin is small and shell-based, easy to read and tweak, and it works as a lightweight quality gate for teams already on Claude Code.