What is Review Codex?

It is a specialized skill for Claude Code that leverages the Codex CLI to perform a final, independent review of your code or implementation plans to ensure high quality and standard compliance.

What standards does it use for the review process?

The skill references a standards.md file within your plugin root, allowing you to define custom organizational or project-specific rules that Codex must enforce.

Can it review plans before code is written?

Yes, it specifically detects plan files to perform a feasibility and completeness review, helping you catch logic or architectural flaws before implementation starts.

Does it support iterative code changes?

Yes, it tracks active sessions to support 'resume' functionality, allowing it to verify that issues identified in previous review rounds have been correctly addressed.

How does it differ from standard Claude reviews?

It acts as an external validator, using specific JSON schemas and a dedicated standards document to provide a more structured, rigorous, and objective assessment than standard conversational reviews.

Review Codex

Name: Review Codex
Author: Z-M-Huang

byZ-M-Huang

•

协作工具

Orchestrates automated code and plan reviews using Codex as a final validation gate for development tasks.

Review Codex serves as the ultimate quality assurance layer in the development workflow, invoking the Codex CLI to perform independent assessments of implementation plans and final code. By acting as a specialized third-party reviewer after standard LLM passes, it ensures that all changes adhere to predefined organizational standards, identifies potential security vulnerabilities, and enforces JSON-structured feedback for consistent, actionable results. This skill is ideal for teams requiring a rigorous, automated 'final gate' before code is approved or merged.

主要功能

01Automated checking against custom organizational standards and documentation

0211 GitHub stars

03Independent assessment layer providing a final gate before deployment or merge

04Structured JSON output via schema enforcement for predictable review results

05Dual-mode validation for both technical implementation plans and final codebases

06Persistent session management for iterative re-reviews and addressing feedback

使用场景

01Automating the final sign-off for complex code implementations before merging to main branches

02Validating technical implementation plans for feasibility and completeness before work begins

03Ensuring code style, security, and architectural standards are consistently met across a project

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add z-m-huang/claude-codex review-codex

For use in Claude.ai and ChatGPT

Download Skill