Problem: Claude Code edits code without independent validation. A second AI reviewer (OpenAI Codex) can catch bugs, security issues, and style problems before they're committed — acting as an automated "second pair of eyes."
Goal: Integrate Codex CLI (v0.99.0, already installed) into every Claude Code session as a global review agent with three trigger points: task completion, pre-commit blocking, and on-demand /codex-review command.
Outcome: Claude's changes get automatically reviewed by Codex, and Claude fixes issues before they reach git history.