What model does the Codex Reviewer use?

The skill utilizes OpenAI's specialized GPT-5-Codex model via the Codex CLI to provide an independent, third-party perspective on your codebase.

Do I need a separate subscription to use this?

You must have the Codex CLI installed and be authenticated via an OpenAI/ChatGPT account or API key to successfully execute the reviews.

Is it safe to let Codex access my code?

Yes. This skill is strictly configured to use a read-only sandbox mode. Codex can analyze and report on your files but lacks permission to modify any code in your workspace.

How does this differ from Claude's standard code analysis?

While Claude provides excellent context-aware analysis, Codex acts as a separate auditor with a different model bias. Combining both helps surface edge cases and errors that a single reviewer might miss.

Can this skill help with Pull Request reviews?

Yes, it includes specific patterns for running git diffs and providing detailed feedback on branch changes, focusing on breaking changes, performance, and security.

Codex Independent Reviewer

Name: Codex Independent Reviewer
Author: paskals

bypaskals

0•

Herramientas de Colaboración

Leverages OpenAI's Codex CLI to provide unbiased, independent code reviews and architectural validation directly within Claude Code.

The Codex Reviewer skill empowers developers to obtain a specialized 'second opinion' by invoking OpenAI's GPT-5-Codex model for deep code analysis. By executing Codex in a mandatory read-only sandbox, this skill identifies security vulnerabilities, performance bottlenecks, and architectural inconsistencies that might be overlooked by a single AI agent. It is designed for high-stakes scenarios like pull request validation, security audits, and comparing implementation code against technical specifications, providing a synthesized report that highlights consensus and divergent insights between Claude and Codex.

Características Principales

01Automated diff analysis for pull request reviews

02Architecture and specification validation patterns

03Read-only sandbox for safe, non-destructive reviews

040 GitHub stars

05Independent GPT-5-Codex model analysis

06Structured JSON output for programmatic feedback processing

Casos de Uso

01Getting an unbiased second opinion on critical security-sensitive code

02Comparing technical design specifications against final implementations

03Validating complex architectural decisions before merging pull requests

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add paskals/ai-assist codex-reviewer

For use in Claude.ai and ChatGPT

Download Skill