davidleitw/xreview
12 stars · Last commit 2026-04-02
Agent-native code review engine for Claude Code — delegates review to Codex for independent three-party consensus
README preview
# xreview Agent-native code review engine for Claude Code and Codex CLI, powered by Codex. xreview delegates code review to Codex (a separate AI model) so your coding agent gets an independent second opinion. It orchestrates a three-party review loop: **Codex reviews, your agent verifies, you decide.** **[中文版 README](docs/README.zh-TW.md)** Curious why this exists? I wrote a post about the motivation and how it works: [Why I built xreview](https://davidleitw.github.io/posts/xreview-en) ## How It Works When you ask your coding agent to review your code, the xreview skill takes over: 1. **Codex reviews** your code and reports findings (bugs, security issues, logic errors) 2. **Your agent verifies** each finding independently — reads the actual source code, confirms or challenges false positives by discussing with Codex 3. **Your agent presents** a Fix Plan with only verified findings — trigger, impact, cascade, and fix options 4. **You decide** — approve all recommended fixes, pick by severity, or adjust per finding 5. **Your agent fixes** strictly per your approved plan 6. **Codex verifies** the fixes in a follow-up round, may find new issues or reopen dismissed ones