Borrow a skill
You don't have to write every skill yourself
Where: Claude Code · Folder: any folder — this one installs globally
Start a new Claude Code session
before you begin.
In Exercise 2.1 you wrote a skillSkillA reusable bundle of know-how that Claude loads on demand. Lives as a folder of markdown on your machine; kicks in when the conversation matches what the skill is for.. That's the expensive way to get one.
People publish skills the same way they publish libraries: on GitHubGitHubThe website where most people host git repositories. Git is the tool, GitHub is one popular place that stores the result. Owned by Microsoft, free for public projects., free, one command to install. This is the shortest exercise in the course — two commands and a question — and it changes how you think about the other ten.
Step 1 — Install two skills
Paste these one at a time. They're commands, not prompts, but Claude Code will run them for you if you paste them in.
Prompt:
npx skills add https://github.com/2ykwang/agent-skills --skill html-visual
Prompt:
npx skills add https://github.com/langchain-ai/deepagents --skill blog-post
Say yes when it asks to install. html-visual turns things into visual HTMLHTMLThe markup language for web pages — defines the structure: headings, paragraphs, buttons, images. CSS styles it; JavaScript makes it interactive. pages. blog-post writes long-form posts that don't read like a press release.
Step 2 — Ask what you just gained
Prompt:
List every skill I now have available, where each one is installed, and — in
one line each — when you would decide to use it. Include the ones I wrote
myself.
You should see the skill you built in 2.1 sitting next to two you didn't. The agent treats them identically.
Stuck?Debrief — what should have happened
- Two new skills on your machine, installed globally.
- A list showing your own skill and the borrowed ones side by side, with no distinction between them.
Interlude B in Exercise 2.2 put it this way: a loose file is a sticky note on a colleague's desk, a skill is a section in the company handbook. What just happened is that you added two sections somebody else wrote.
The instinct most people bring from software — check what exists before you build it — barely exists yet for agent skills, because the ecosystem is eighteen months old. It's about to be the default. Search before you write.
Stuck?FAQ for this exercise
Q: Is it safe to install a skill from a stranger's GitHub? A: Treat it exactly like an npmnpmJavaScript's package manager — ships with Node. Downloads libraries other people have written and lets you use them in your code (or as standalone command-line tools). package or a browser extension: a skill is instructions your agent will follow, so a malicious one is a real risk. Read the SKILL.md before you install from anywhere you don't recognise — ask Claude to summarise what it does and flag anything that touches credentials, sends data, or runs shell commands. The two above are public, popular, and readable.
Q: npx asked me to install something first. A: That's normal — npx fetches the tool on demand. Say yes. If it fails entirely, ask Claude: "npx isn't working, diagnose it and fix it."
Q: Where did they actually go? A: ~/.claude/skills/, the same folder your own skill from 2.1 lives in. Ask Claude to open it if you want to read them.
Q: How do I get rid of one? A: Ask — "remove the blog-post skill." It's a folder; deleting it is the uninstall.
OptionalGo deeper
Go shopping.
Search GitHub for published Claude agent skills. Find me five that would
actually be useful for someone with my job — read my person.md first — and for
each one tell me what it does, who made it, and whether the SKILL.md looks
trustworthy. Don't install anything yet.
Then install the one that made you say "oh." That reflex — someone has probably already built this — is worth more than any single skill.