Codex Review

You've written some code and you want a second pair of eyes before it ships — or the user asked you to put Codex on it. Codex, the OpenAI coding agent, has a built-in review mode (codex exec review) that diffs the repo, reasons about the changes, and reports findings. This skill is how you run it and what to do with what comes back.

A review from a different model is worth the round-trip: it reasons from a different prior and catches the bug or edge case you've stopped being able to see in your own work. But a review is input, not a verdict — you triage the findings, you don't rubber-stamp them.

When to use

Self-triggered (your own initiative):