Skills — Teach Claude New Tricks

What Are Skills?

In the last module, you learned about CLAUDE.md — a file that tells Claude how to behave in your project. Skills take that idea further. A skill is a reusable set of instructions that Claude can follow on demand, like a recipe card you hand to a chef.

Instead of typing the same complex instructions over and over, you write them once in a special file called SKILL.md. Then you trigger that skill whenever you need it — by typing a simple slash command like /fix-issue or /explain-code.

Think of it like this: CLAUDE.md is the employee handbook that applies to everything. Skills are specialized training manuals for specific tasks.

Personal vs Project Skills

Skills live in one of two places, depending on who should use them:

Personal skills are yours alone. They live in your home directory and follow you across every project:

~/.claude/skills/fix-issue/SKILL.md
~/.claude/skills/explain-code/SKILL.md
~/.claude/skills/morning-report/SKILL.md

Project skills are shared with your team. They live inside the project folder and get committed to version control:

.claude/skills/deploy/SKILL.md
.claude/skills/review-pr/SKILL.md
.claude/skills/generate-docs/SKILL.md

The rule is simple: if it’s useful for you personally (like your preferred writing style or your morning routine), make it a personal skill. If it’s useful for the project (like how to deploy or review code), make it a project skill.

Pro Tip

Personal skills are great for things like “summarize this in my writing style” or “generate a LinkedIn post.” Project skills are great for team workflows like “run the test suite and explain any failures.”

Anatomy of a SKILL.md File

Every SKILL.md file has two parts: frontmatter (metadata at the top) and instructions (the actual recipe).

Here’s what a real SKILL.md looks like:

Let’s break down the frontmatter fields:

• name — The slash command name. This becomes /explain-code.
• description — A short summary that appears when you list available skills.
• allowed-tools — Which tools Claude can use when running this skill. This is a safety guardrail.
• disable-model-invocation (optional) — If set to true, the skill can only use tools, not generate text. Useful for pure automation tasks.

Pro Tip

The allowed-tools field is your safety net. If a skill should only read files and never modify them, set it to Read only. Claude will follow this restriction.

Using $ARGUMENTS

Skills become truly powerful when you add parameters. The special $ARGUMENTS placeholder lets you pass information to a skill when you invoke it.

Here’s a skill that fixes GitHub issues:

Now you can use it like this:

Claude reads the SKILL.md, replaces $ARGUMENTS with 42, and follows the instructions. It fetches issue #42 from GitHub, reads the code, implements the fix, runs tests, and commits — all from one command.

/fix-issue 42

Real Examples

Here are skills that real people use every day. You can copy any of these and customize them:

/explain-code src/components/PaymentForm.tsx

/morning-report

/generate-docs src/utils/helpers.js

/security-review src/auth/

Creating Your First Skill

Let’s walk through creating a skill from scratch. Say you want a skill that explains any error message in plain English and suggests a fix.

Then write:

Now you can use it immediately:

/explain-error "ECONNREFUSED 127.0.0.1:3000"

Pro Tip

You can also ask Claude to create skills for you! Just say “Create a skill called /weekly-summary that generates a summary of all commits from the past week.” Claude will create the folder and SKILL.md file automatically.

Try It Yourself