Is it safe to run Codex on my local files?

The skill includes mandatory sandbox flags such as 'read-only' for analysis and 'workspace-write' for automated editing, ensuring you control the level of access granted to the CLI.

How do I continue a previous Codex session?

You can simply ask to 'resume' or 'continue'. The skill uses a specialized resume syntax that pipes your new instructions into the last active session while maintaining configuration.

What happens if Claude and Codex disagree on a solution?

The skill is designed to treat Codex as a colleague. Claude is instructed to evaluate Codex's output critically, push back on inaccuracies, and discuss disagreements with Codex using its identity as a peer AI.

Can I control the reasoning depth of the models?

Yes, the skill allows you to specify reasoning effort levels including xhigh, high, medium, and low to balance processing power and task complexity.

How does this skill handle long-running tasks?

It utilizes background execution via the bash tool, allowing Claude to continue with independent work while Codex processes complex tasks that may take 15 minutes or more.

Codex CLI Bridge

Name: Codex CLI Bridge
Author: integralmedia1

byintegralmedia1

0•

开发者工具

Delegates complex coding tasks, analysis, and refactoring to OpenAI Codex with advanced reasoning and background execution.

The Codex skill allows Claude to leverage the OpenAI Codex CLI as a specialized peer agent for deep code analysis, automated editing, and multi-directory refactoring. It provides a seamless interface to run long-duration tasks in the background with configurable reasoning efforts and granular sandbox safety levels, ranging from read-only analysis to workspace-wide edits. By treating Codex as a collaborative colleague, this skill enables sophisticated AI workflows where Claude can delegate heavy lifting, review outputs, and even debate technical implementations with Codex to ensure precision and quality.

主要功能

01Support for gpt-5.4 and reasoning-optimized model variants

02Asynchronous background execution for tasks lasting 5-15+ minutes

03Flexible sandbox environments including read-only and workspace-write modes

040 GitHub stars

05Configurable reasoning effort levels (xhigh to low) for complex logic

06Persistent session management with seamless resume capabilities via stdin piping

使用场景

01Analyzing UI screenshots to generate or repair frontend code using image context

02Performing large-scale project refactoring across multiple directories in background mode

03Executing deep code reviews and security audits using high-reasoning models

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add integralmedia1/skill-codex codex

For use in Claude.ai and ChatGPT