codex-cli

What This Skill Does

The codex-cli skill empowers your OpenClaw agent with the ability to interface directly with the OpenAI Codex CLI, a powerful terminal-based coding assistant. This skill transforms your agent into a professional software engineer, granting it the capability to read project files, execute terminal commands, implement features, and conduct rigorous code reviews. By delegating complex coding tasks to the specialized gpt-5-codex model via this interface, the OpenClaw agent acts as a supervisor, while the Codex CLI handles the low-level logic, file manipulation, and debugging processes required for high-quality software development.

Installation

To integrate this skill into your environment, use the OpenClaw package manager. Execute the following command in your terminal:

clawhub install openclaw/skills/skills/phucanh08/codex-sub-agents-1

After installation, ensure you have the base Codex CLI package installed via npm (npm i -g @openai/codex). Complete your authentication by running codex login to securely link your OpenAI subscription, allowing the agent to perform actions on your behalf.

Use Cases

• Feature Implementation: Automatically stub out new classes, functions, or API endpoints based on natural language requirements.
• CI/CD Troubleshooting: Diagnose failed pipeline logs and automatically apply fixes to your codebase.
• Code Review: Audit uncommitted changes or compare branches for potential bugs, security vulnerabilities, or style violations before pushing to production.
• Legacy Refactoring: Ask the agent to explain complex, undocumented legacy code and suggest modern alternatives or performance optimizations.
• Contextual File Editing: Execute multi-file refactors where the agent must traverse directories to perform consistent updates across a large project.

Example Prompts

1. "Codex, look at the project in ~/apps/payment-gateway and refactor the auth middleware to use the new session token structure."
2. "Review the current changes in my git repository and suggest improvements for better error handling before I commit."
3. "The CI build is failing on test-suite/integration-01. Please investigate the logs, identify the cause, and provide a fix."

Tips & Limitations

For maximum efficiency, use the --full-auto flag only when working within a sandbox or a version-controlled environment where you can easily revert changes. If you are dealing with critical infrastructure, set your approval mode to 'Read Only' to inspect all suggested edits before allowing the agent to write to your filesystem. Remember to use the /compact command periodically during long sessions to manage token usage and keep the agent's context window focused on the most relevant parts of your codebase.