agentsclimarketplace

Codex review

Skill lil-lon/.claude/skills/codex-review

Review staged changes with Codex by piping `git diff --cached` to `codex exec` and asking for P1-P4 labeled findings. Use when the user asks for a Codex-specific review on staged work ("codex review", "review with codex"). Do NOT auto-invoke for generic "review this" requests — those belong to the built-in `/review` skill.From its SKILL.md

Install
npx -y skills add lil-lon/.claude --skill codex-review

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

2 things to look at

  • 2 stars2 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.
  • runs commandsInstructs the agent to run 1 command, including `bash "$CLAUDE_CONFIG_DIR/skills/codex-review/review.sh"`.

SKILL.md

1.6 KB, 307 tokens by cl100k_base, as published. Nobody here has run it

Run the review script that lives next to this SKILL.md, from the user's current working directory (so git diff --cached operates on their repo):

bash "$CLAUDE_CONFIG_DIR/skills/codex-review/review.sh"

The script aborts with exit 1 when no staged changes exist — surface that message and stop.

Print the script's stdout verbatim. Do not paraphrase, trim, or add a summary on top.

Handling findings

  • P1 findings must always be fixed. Never commit when codex reports a P1.
  • P2 findings should be fixed, unless they have been previously discussed in this conversation and explicitly accepted as a policy trade-off. Treat an accepted P2 as resolved.
  • The iteration loop terminates when codex no longer reports any P1, AND each remaining P2 has been either fixed or accepted as a known trade-off in this conversation.
  • P3/P4 findings are non-blocking; surface them but proceed.

Sandbox

The Claude Code sandbox blocks codex exec from accessing its session, so run this script unsandboxed. It is safe because the script itself passes --sandbox read-only to codex exec.

TODO: narrow the sandbox.excludedCommands glob in settings.json so that this wrapper script is matched and the explicit unsandbox flag is no longer needed.

What ships with it: 1 file

426 B alongside SKILL.md, 1 of them executable

Keep looking

Skills are one crate of 325,949. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.