Skip to content
Get packs

Free course

AI agents & agent skills, from zero to building your own.

What agents are, how they use tools and skills, how to install skills safely, how to write and publish your own, and how to build a simple agent — in plain English, with links to the real tools.

14 chapters · about 65 minutes in total · free, no sign-up

Start chapter 1

Get each new chapter and the best new skills by email

One email a week: new course chapters, skills worth installing, and the fastest-growing open-source AI projects. Unsubscribe anytime.

Part 1 · Understand

  1. 01What an AI agent is: model, tools, loop and contextAn AI agent is a language model that can take actions in a loop until a job is done. Learn the four parts every agent has and how it differs from a chat.
  2. 02How agents use tools, MCP and the context windowTool calls, the Model Context Protocol and context windows explained: how an agent reaches the outside world and why its working memory is precious.
  3. 03What agent skills are and how progressive disclosure worksA skill is a folder with a SKILL.md file. Learn how agents load its name and description first, the body when relevant, and extra files only on demand.
  4. 04Skills vs prompts vs MCP vs subagents vs plugins: when to use whichPrompts, skills, MCP servers, subagents and plugins each solve a different problem. A side-by-side comparison and a simple way to choose the right one.

Part 2 · Use

  1. 05Install and use skills in Claude Code, the Claude apps and the APIWhere skill folders go in Claude Code, how to add skills in the Claude apps, and what to know about using skills through the API and Agent SDK.
  2. 06Choosing good skills and reading them safelyHow to tell a useful skill from a weak one, and how to review its instructions, scripts, network access and secrets handling before you install it.
  3. 07Everyday workflows with skills for developers, designers and marketersPractical skill-powered workflows for coding, design and marketing work, with examples of what to ask, which skills help and how to combine them.

Part 3 · Create

  1. 08Write your first agent skill, step by stepCreate a working skill from scratch: the folder, the SKILL.md frontmatter, clear instructions and examples, then install it and test it with a real request.
  2. 09Writing skill descriptions that trigger reliablyThe description decides whether your skill is ever used. A practical formula, before-and-after examples and a way to test that it fires at the right time.
  3. 10Adding scripts, templates and reference files to a skillWhen to move detail out of SKILL.md into reference files, templates and scripts, how to point to them, and how to write scripts an agent can run safely.
  4. 11Testing, iterating and versioning a skillHow to check that a skill triggers, follows its steps and improves results, how to iterate without breaking it, and how to version it with Git.
  5. 12Sharing and publishing your skillsPackage skills for others: a clear GitHub repository layout, an open-source licence, a helpful README, and distribution through plugins and marketplaces.

Part 4 · Build agents

  1. 13Build a simple agent with toolsThe core of every agent in a page of code: define tools, run the loop, feed results back and stop safely. An SDK-agnostic walkthrough you can adapt.
  2. 14Agent safety and good practiceLeast-privilege permissions, human approval, prompt-injection defences, secret handling and cost control: the habits that keep agents useful and safe.