How do I create a new custom skill?

You can use the $skill-creator command within Codex to bootstrap a new skill directory containing the required SKILL.md and metadata structure.

What is the safest sandbox mode for autonomous coding?

The 'read-only' mode is the safest as it prevents all writes; however, 'workspace-write' is the standard for development as it restricts file changes to the specific project directory.

How can I add a new MCP server to my configuration?

You can add an MCP server via the CLI using the command 'codex mcp add -- ' or by manually adding a [mcp_servers] block to your config.toml.

Where is the Codex configuration file located?

The primary configuration file is located at ~/.codex/config.toml on your local system.

Can I use local LLMs with this configuration?

Yes, by setting the model_provider to 'oss' in the configuration, you can integrate local providers like LM Studio or Ollama.

Codex CLI Configuration Guide

Name: Codex CLI Configuration Guide
Author: GGPrompts

byGGPrompts

•

開発者ツール

Streamlines the configuration and optimization of the OpenAI Codex CLI for advanced autonomous development workflows.

概要

This skill provides comprehensive guidance for managing the OpenAI Codex CLI, acting as a bridge for developers who need to configure sandbox policies, security approvals, and tool integrations within Claude. It offers structured patterns for editing config.toml, managing Model Context Protocol (MCP) servers, and developing custom skills, ensuring that your AI-driven development environment is both secure and highly performant.

主な機能

Simplified MCP server integration via CLI and configuration files
Support for local LLM provider integration like Ollama and LM Studio
Comprehensive config.toml management for model and provider settings
Security policy configuration for sandbox access and command approvals
1 GitHub stars
Framework for creating and bootstrapping custom Codex skills

ユースケース

Extending Claude's capabilities with custom MCP servers and shell tools
Establishing secure 'workspace-write' sandbox boundaries for autonomous agents
Tuning model reasoning effort and verbosity for specific coding tasks

概要

主な機能

Simplified MCP server integration via CLI and configuration files
Support for local LLM provider integration like Ollama and LM Studio
Comprehensive config.toml management for model and provider settings
Security policy configuration for sandbox access and command approvals
1 GitHub stars
Framework for creating and bootstrapping custom Codex skills

ユースケース

Extending Claude's capabilities with custom MCP servers and shell tools
Establishing secure 'workspace-write' sandbox boundaries for autonomous agents
Tuning model reasoning effort and verbosity for specific coding tasks