Does it require manual intervention?

While highly automated, the skill is programmed to stop and ask for user direction when encountering breaking changes or significant architectural shifts.

Which models are compatible with this skill?

It is optimized for gpt-5.2-codex, gpt-5.1-codex-mini, and gpt-5.1-codex-max to balance between reasoning power and cost efficiency.

How does this skill improve code quality?

By establishing a 'Cross-Review' phase, the skill forces a second AI evaluation of the logic, architecture, and security of the code before it is finalized.

What is the Codex Review Loop?

It is a specialized Claude Code skill that orchestrates two AI processes: one to handle planning and implementation, and another to act as a dedicated reviewer for validation.

Can I use this for existing codebases?

Yes, the skill can be used to review existing implementations or to plan and execute new features within an established repository.

Codex Review Loop

Name: Codex Review Loop
Author: tygwan

bytygwan

•

Security & Testing

Orchestrates a dual-AI engineering loop that validates plans and reviews code implementations through continuous cross-verification.

The Codex Review Loop skill implements a high-efficiency 'Dual-AI' workflow designed to maximize code quality and architectural integrity. By separating the roles of implementation and validation between two Codex instances, it ensures that every plan is scrutinized for logic errors and security flaws before execution. This skill is particularly useful for complex software engineering tasks where architectural consistency and bug prevention are paramount, providing a structured, phase-based approach to the development lifecycle.

Key Features

01Dual-AI orchestration for simultaneous planning and validation roles

02Automated logic and security auditing in a read-only sandbox environment

03Phase-based development loop: Plan, Validate, Implement, Review, and Fix

041 GitHub stars

05Seamless session resumption to maintain context during iterative debugging

06Support for gpt-5.2-codex and specialized mini/max model variants

Use Cases

01Validating complex architectural designs before implementation begins

02Iterative bug fixing through self-correcting AI-to-AI feedback loops

03Automating deep-dive code reviews for performance and best practices

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add tygwan/ultra-codex-init codex-review-loop

For use in Claude.ai and ChatGPT

Download Skill