How do I trigger the review workflow?

You can start the process by using natural language triggers like 'start codex review' or 'with codex review' within the Claude Code terminal.

Does Codex have access to my local project files?

Yes, Codex operates within the same directory as Claude, allowing it to inspect the codebase directly during the review process.

Can I use this on multiple git branches?

Yes, the skill is designed with branch isolation, meaning review states and session IDs are managed independently for each git branch or worktree.

How does the escalation process work?

If Claude and Codex cannot reach an agreement after a set number of iterations, the skill escalates the decision to the user with a summary of the conflict.

What is a cross-agent review?

It is a workflow where one AI agent (Claude) handles the implementation while a second agent (Codex) acts as a reviewer to catch errors and suggest improvements.

Codex Code Reviewer

Name: Codex Code Reviewer
Author: artwist-polyakov

byartwist-polyakov

•

Collaboration Tools

Implements a multi-agent review workflow where Claude develops and Codex validates logic, planning, and code quality.

The Codex Review skill establishes a sophisticated cross-agent development environment by pairing Claude's implementation capabilities with Codex's specialized review logic. It automates the end-to-end development lifecycle—from initial architectural planning to final code implementation—ensuring all changes pass rigorous technical validation. By utilizing a structured feedback loop, it handles 'accept or argue' scenarios and provides automated state management across different git branches, making it an essential tool for maintaining high standards in AI-assisted software engineering.

Key Features

01Configurable iteration limits with intelligent escalation to the human user

02Cross-agent review loop between Claude and Codex for enhanced quality control

03Two-stage validation covering both high-level technical plans and final implementation

0442 GitHub stars

05Automated state management with branch-specific isolation and review history

06Point-by-point feedback processing for handling complex implementation disputes

Use Cases

01Automating peer-review processes for AI-generated code to catch logical or architectural flaws

02Validating complex system changes like JWT authentication or database migrations before merging

03Maintaining architectural consistency across multiple git branches and worktrees simultaneously

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add artwist-polyakov/polyakov-claude-skills codex-review

For use in Claude.ai and ChatGPT

Download Skill