Back to directory

Codex CLI Skills & Profiles

A curated set of skills, provider profiles and configuration templates for OpenAI's Codex CLI, adding deep research, image generation, browser debugging through Chrome DevTools MCP, and quick switching between model providers such as Azure OpenAI, OpenRouter and GitHub Copilot.

Codex CLI Skills & Profiles

This repository packages ready-to-run Codex CLI configuration aimed at individual developer productivity: preset provider profiles for the Copilot gateway, ChatGPT, Azure OpenAI, OpenRouter and LiteLLM/GitHub Copilot let a user switch models without hand-editing configuration, and skills add reusable workflows for requirements gathering, deep research, image generation, subtitle extraction and task handoff. Chrome DevTools MCP is included for debugging web pages, inspecting performance and running browser automation from Codex.

It can be installed as a full Codex home directory for everything at once, or added as a plugin marketplace so only its skills are installed without touching an existing config.toml, model providers or MCP settings; a separate npx-based skills installer is also documented as a fallback, though the README notes it may lag behind the repository's latest tagged release. The project's own Codex configuration is separate from a companion repository for Claude Code settings, skills and agents.

More in Agent skills

See all 581
  • Provenance gate

    github.com

    A deterministic authorization gateway that sits between an AI agent and its tools, tracing where each parameter of a side-effect call came from, and blocking the call or asking a human when the source does not check out.

  • A markdown framework of system rules that restates Isaac Asimov's Four Laws of Robotics as a set of prioritized loss calculations for large-language-model agents to follow, published as a discovery file for agents and crawlers to read.

  • Vitruvius

    github.com

    An engineering research agent that runs a discover, read, synthesize, verify and review loop over technical questions, reading sources directly rather than inferring from titles, and publishing its own scored verifier benchmark, weaknesses included.