Agent Skills Mastery: Create and Use SKILL.md
Learn Agent Skills from zero — the open SKILL.md format for packaging expertise into portable capabilities for Claude, Claude Code, and any compatible AI agent. Free beginner course.
Course Content
Agent Skills Mastery
Every AI agent hits the same wall: it’s a generalist. It doesn’t know your team’s deployment runbook, your company’s brand voice, or the exact sequence of flags your data pipeline needs. You can re-explain that context in every conversation — or you can package it once, as a skill, and let any compatible agent load it on demand.
That’s what Agent Skills are for. Originally developed by Anthropic and released as an open format, Agent Skills are now supported across a growing set of agent products — Claude Code, Claude.ai, VS Code, Cursor, OpenAI Codex, Gemini CLI, and more. A skill is nothing exotic: a folder with a SKILL.md file, optionally bundling scripts, reference docs, and templates. What makes the format powerful is progressive disclosure — agents load a skill’s name and description at startup, and only pull the full instructions into context when a task actually needs them.
By the end of this course you’ll be able to say, plainly: “I know how to create and use SKILL.md.” That’s the whole goal — no prior assumptions, no jargon left unexplained.
What You’ll Build
- A working skill that an agent can discover, trigger, and execute correctly on the first try
- A small trigger test set — a handful of labeled prompts that check whether your skill activates when it should
- Five real, working skills covering five different jobs a skill can do — the capstone project
How This Course Is Organized
You’ll start with the why, then go layer by layer through the format itself: the frontmatter fields, the folder anatomy, how agents find a skill, and how they load it. Once you understand the shape of a skill, you’ll build one by hand, bundle files into it, test it, and publish it for others to use. The capstone has you build five small skills rather than one large one — breadth first, so you leave this course having actually made things, not just read about them.
What’s Next After This Course
This is the free, foundational half of a two-course path. Once you’re comfortable creating and using skills, Production Agent Skills Engineering picks up where this course stops — design patterns, testing and evaluation at scale, security, versioning, and everything else that separates a skill that works on your laptop from one thousands of agents can rely on.
📋 Prerequisites
- Basic familiarity with an AI coding assistant or chat agent (Claude, ChatGPT, Copilot, or similar)
- Comfortable editing YAML/Markdown files and running commands in a terminal
🎯 What You'll Learn
- Explain what Agent Skills are and how progressive disclosure keeps agents context-efficient
- Write valid SKILL.md files that follow the open Agent Skills specification
- Explain where agents discover skills and how the three-tier loading model works
- Bundle scripts, references, and assets into a well-organized skill
- Manually test a skill and build a small query set to check it triggers correctly
- Publish a skill for a team and keep it portable across agent clients
- Build five small, working skills covering five distinct jobs a skill can do