There are two good ways to start with Codex on this site. The main browse page is Codex skills. This guide is for the follow-up questions: how to evaluate Codex-oriented skills, how portable they are, and when AGENTS.md matters more than the skill list itself.
If your question is “what Codex skills can I browse right now,” go straight to the Codex platform collection. This guide is here for the next question: how do Codex users evaluate skills once they understand what is available?
That distinction matters because not every search with codex skills is asking for the same thing. Some users want a list. Others want a workflow explanation.
The strongest Codex-oriented skills usually do three things well:
That is why GitHub skills and Browser Automation skills can still be relevant even when the starting query is brand-specific.
Codex users often search for both skills and AGENTS.md because the two concepts live next to each other. Skills package reusable workflows. AGENTS.md helps define project-level expectations, boundaries, or conventions. The two are complementary, not interchangeable.
If that is your actual question, continue with the Codex AGENTS.md guide after this page.
Open the Codex skills collection for live browsing, compare adjacent workflows in GitHub skills, and use Agent Skills vs MCP only if your problem is architectural rather than operational.