Learn to build with AI
Short, practical guides on Claude Skills, MCP servers and the rest of the developer AI stack. No fluff.
What are AI workflows, and how do you install one?
In this directory, an AI workflow is a self-hosted platform or framework for wiring automations and LLM agents together, not a single file you drop in. Here's the difference between the two shapes it comes in, and how to run one.
What are GitHub Copilot agents, and how do you install one?
A Copilot agent is a markdown file that gives GitHub Copilot Chat a fixed persona and rulebook for one job, selectable right from the mode picker in VS Code. Here is what's actually in one, and how to get one running.
What are Cursor Rules, and how do you install one?
A Cursor Rule teaches the editor your project's own conventions instead of relying on the model's general knowledge. Here's what's inside one, and how to drop one into a project.
What makes a good AI prompt, and how to write one
A good prompt isn't a magic phrase, it's enough of the right information that the model doesn't have to guess. Here's what actually goes into one, with three worked examples and where the same discipline shows up inside the skills and agents in this directory.
What is an agent harness, and how do you run one with DeepSeek?
A harness is the loop that turns a model into a coding agent. Here's what that means, plus a step-by-step setup you can actually follow and check as you go.
How to install a Claude Skill
Installing a skill means something different depending on where you use Claude: a folder copy in Claude Code, a zip upload on claude.ai, an API call through the Skills API.
Claude Skills vs MCP servers
Both make an assistant more capable, but they solve different problems. One is know-how, the other is access, and most real setups use both together.
What are Claude Skills?
A skill is a folder of instructions you hand to Claude so it already knows your task. Here is what is actually inside one, and what people mix it up with.