Skill

codex-review

Runs OpenAI Codex CLI (`codex review`) for AI code reviews on git staged, unstaged, uncommitted, or branch changes before committing. Uses custom prompts, 10min timeout, saves output to .codex-review-output.md.

Git

OpenAI

npx claudepluginhub delexw/claude-code-misc

Tool Access

This skill uses the workspace's default tool permissions.

Preview

Use the `codex review` command to get an AI-powered code review of the current changes.

SKILL.md

Similar Skills

codex-cli-review

147

Executes Codex CLI for code reviews with full disk access, enabling deep codebase exploration and uncommitted change analysis. Outputs severity-grouped findings and merge gate.

1 file1 tool

sd0x-dev-flow

codex-review

Runs official Codex code review on code changes, diffs, branches, or commits, then double-checks each finding by reading source code and classifying agree/disagree/nuance.

4 tools

codex-advisor

diff-review

Reviews git diffs with Codex for bugs, security vulnerabilities, style issues, and performance problems. Useful for second opinions after AI edits or pre-commit checks.

codex

Stats

Parent Repo Stars25

Parent Repo Forks5

Last CommitMar 19, 2026

Actions

View Source View Plugin View on GitHub View README

Help us improve

Share bugs, ideas, or general feedback.

Codex CLI Code Review

Use the codex review command to get an AI-powered code review of the current changes.

Workflow

1. Check codex CLI availability

Run which codex to confirm it's installed. If not found, tell the user: "The codex CLI is not installed. Install it with npm install -g @openai/codex or see https://github.com/openai/codex"

2. Determine what to review

Pick the right flag based on context:

Staged changes only (default before a commit): codex review
All local changes (staged + unstaged + untracked): codex review --uncommitted
Changes against a branch: codex review --base <branch>
A specific commit: codex review --commit <sha>

If the user doesn't specify, check git diff --cached --stat first. If there are staged changes, review those. Otherwise fall back to --uncommitted.

3. Run the review

IMPORTANT: Use a 10-minute timeout (600000ms) for the Bash command — codex streams from the OpenAI API and large diffs can take several minutes to complete. The default 2-minute timeout will cause the stream to disconnect mid-review.

The prompt should describe what to focus on. If the user provided $ARGUMENTS, use that as the prompt. Otherwise use a sensible default like "review the changes for bugs, style issues, and potential improvements".

Command construction rules:

Staged changes (no flag): pass the prompt as a positional argument.
```
codex review "<prompt>"
```
With --base, --commit, or --uncommitted: the CLI does not allow a positional prompt argument alongside these flags. Pipe the prompt via stdin using - instead:
```
echo "<prompt>" | codex review --base <branch> -
echo "<prompt>" | codex review --uncommitted -
echo "<prompt>" | codex review --commit <sha> -
```

If the command fails due to environment issues (missing API key, network error, etc.), try these in order:

Check if OPENAI_API_KEY is set
Try running with --config model="gpt-5.2" as a fallback model
If still failing, report the error and skip the review

4. Save the review output

Save the full codex review output to .codex-review-output.md in the repository root.

5. Handle the review output

No issues found: Report that the review passed clean and proceed.
Minor suggestions: Summarize them for the user. These are informational — the user can decide whether to act on them.
Critical issues (bugs, security vulnerabilities, logic errors): Summarize the critical findings clearly. Ask the user whether they want to fix these before proceeding. If yes, make the fixes but do NOT commit — present the changes for human review first.

6. Environment troubleshooting

The codex CLI depends on:

Node.js runtime
OPENAI_API_KEY environment variable
Network access to OpenAI API

If running through asdf/nvm, the shim path may need to be resolved. Try $(which codex) or the full path if the shim doesn't work in the current shell context.