Skills-Learning

Skills Learning is a workspace for building and testing "skills" for Codex, OpenAI's coding assistant. A skill is a folder of instructions that teaches Codex how to carry out a specialized task, so instead of typing out detailed instructions every time, you can just reference the skill and Codex knows what to do.

Each skill has a required file called SKILL.md, which starts with a short block describing the skill's name and a description that tells Codex when to use it, followed by the actual instructions for the task. A skill can also include optional folders: references for deeper documentation that only loads when needed, evals for test prompts that check whether the skill behaves correctly, scripts for small reusable programs the skill can call on, and assets for templates or other supporting files.

This repository currently ships one working example skill called seo-reviewer, which audits websites, landing pages, web apps, or source code for search engine optimization problems, such as missing metadata, broken sitemaps, or structured data issues. It comes with its own checklist document, a set of test prompts, and a small Python script that pulls a basic SEO snapshot from a web page or an HTML file.

To use a skill, you simply reference it by name or path in a prompt to Codex, or describe the task naturally and let Codex pick the right skill on its own if it recognizes the request. The repository also explains how to create a brand new skill from scratch, how to validate that a skill's files are well formed, and how to run its test prompts to compare behavior across changes.

A separate lock file keeps track of any external skills that were installed from other sources, along with hashes to verify their origin. The project is released under the MIT license, and its included seo-reviewer skill carries its own copy of that license as well.