codex-chat

Delegate development tasks between Claude Code and the OpenAI Codex CLI.

5|Updated Mar 27, 2026
One-click install
npx skills add https://github.com/barkleesanders/claude-code-starter --skill codex-chat
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: codex-chat
Source: https://github.com/barkleesanders/claude-code-starter/tree/main/skills/codex-chat
Command: npx skills add https://github.com/barkleesanders/claude-code-starter --skill codex-chat

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires tmux, codex, python3, and includes scripts (resource) components.

What problem does it solve?

This Skill removes the friction of switching between Claude Code and the OpenAI Codex CLI, allowing you to access Codex's specialized capabilities, get alternative perspectives on your work, and delegate tasks without interrupting your current workflow in Claude Code.

Core Features & Use Cases

  • Dual Interaction Modes: Use one-shot exec mode for quick, isolated tasks like code reviews or second opinions, or persistent tmux mode for multi-turn debugging conversations that preserve context across prompts.
  • Parallel Workflows: Run Codex investigations in the background while continuing work in Claude Code, then retrieve and compare results later to make more informed decisions.
  • Use Case: For example, when debugging a complex race condition, you can ask Codex to analyze the issue in parallel while you continue working on a proposed fix in Claude Code, then compare both perspectives to select the most robust solution.

Quick Start

Use the codex-chat skill to ask OpenAI Codex to review the recent git diff in your project for security issues and performance anti-patterns, then save the output to a file for your review.

Frequently Asked Questions about codex-chat

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

FAQPage Schema
How do I get a second opinion from OpenAI Codex without leaving Claude Code?

To get a Codex second opinion without leaving Claude Code, use the codex-chat skill to delegate tasks directly to the OpenAI Codex CLI, utilizing one-shot exec mode for isolated code reviews or persistent tmux mode for multi-turn debugging.

Can I run parallel codebase investigations using Claude Code and Codex?

Yes, you can run parallel codebase investigations by delegating tasks to Codex in a persistent tmux session while continuing your current work in Claude Code, allowing you to retrieve and compare alternative AI perspectives later.

What dependencies do I need to use Codex for task delegation within Claude Code?

To use Codex for task delegation within Claude Code, you need the OpenAI Codex CLI installed and authenticated, along with tmux for persistent interactive sessions and python3.

What is the best way to use Codex for debugging complex codebase issues alongside Claude?

The best way to use Codex for debugging complex issues alongside Claude is using persistent tmux mode, which preserves context across multi-turn prompts for parallel investigation of race conditions or algorithm analysis.

Does the codex-chat skill support saving code review outputs to a file?

Yes, the codex-chat skill supports saving code review outputs to a file by using the codex exec one-shot prompt execution mode, which includes optional file output and working directory specification.

When should I use one-shot exec mode instead of persistent tmux mode for Codex tasks?

Use Codex one-shot exec mode for quick, isolated tasks like code reviews or second opinions, and use persistent tmux mode when you need multi-turn debugging conversations that preserve context across prompts.