codex-review

Route plans or code diffs to OpenAI Codex for independent review.

36|1|Updated Mar 24, 2026

Search Tags:#codex #review #code-review #cross-model #plan-review #iterative-feedback

Ready to run

npx skills add https://github.com/jiahao-shao1/sjh-skills --skill codex-review-jiahao-shao1

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: codex-review
Source: https://github.com/jiahao-shao1/sjh-skills/tree/main/skills/codex-review
Command: npx skills add https://github.com/jiahao-shao1/sjh-skills --skill codex-review-jiahao-shao1

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Cross-model review is needed to catch blind spots and validate plans or code changes by querying Codex for independent feedback, enabling safer architectures and more robust implementations.

Core Features & Use Cases

Auto-detects plan vs code review and routes to Codex
Iterative feedback: Claude revises plans/code based on Codex input up to 5 rounds
Triggers and workflows for architecture decisions, non-trivial refactors, config changes

Quick Start

Run the codex-review skill to have Codex review your plan or code changes and iterate until a verdict is reached.

Frequently Asked Questions about codex-review

High-intent search queries and answers about installing and using this skill.

FAQPage Schema

How do I get a cross-model code review on my code diffs?▼

Get cross-model code review by routing your code diffs to OpenAI Codex for independent verification of correctness and security. The process iterates up to five rounds with revised submissions until a robust verdict is reached.

Can I use Codex for plan review to catch architecture risks?▼

Yes, you can use Codex for plan review to catch architecture risks. The system auto-detects plan submissions and evaluates them for completeness and risks, providing iterative feedback to enable safer architectures.

What do I need to configure before running a Codex CLI review?▼

Before running a Codex CLI review, you need to ensure the Codex CLI is installed and your OpenAI credentials are configured. These prerequisites allow the system to route plans or code diffs for independent verification.

How does iterative feedback work when reviewing code changes?▼

Iterative feedback for reviewing code changes works by routing diffs to Codex and allowing Claude to revise the code based on Codex input. This cycle repeats up to five rounds to ensure correctness and security.

When should I use a cross-model second opinion for refactoring?▼

You should use a cross-model second opinion for non-trivial refactors, architecture decisions, or configuration changes. This approach validates plans and code changes by querying Codex to catch blind spots and ensure robust implementations.

What is the maximum number of rounds for Codex review iterations?▼

The maximum number of rounds for Codex review iterations is five. During these iterations, Claude revises plans or code diffs based on Codex input until a final verdict on correctness, security, or completeness is reached.