What's in this lesson: A definitive guide to creating custom "Skills" for Claude—specialized folders that teach Claude workflows so you never have to re-explain them.
Why this matters: Teaching Claude once means you benefit every single time. It turns raw AI capabilities into reliable, optimized workflows.
The Complete Guide to Building Skills for Claude

Teach Claude Once

A skill is a set of instructions—packaged as a simple folder—that teaches Claude how to handle specific tasks or workflows.

Instead of re-explaining your preferences, processes, and domain expertise in every conversation, skills let you configure Claude's behavior permanently. They are incredibly powerful for repeatable workflows like generating frontend designs from specs, conducting research, or writing documents that strictly follow a team style guide.

The Anatomy of a Skill

At its core, a skill is just a folder. But inside that folder lies a specific, modular structure that Claude understands perfectly.

Interactive Folder Tree

Click a folder or file to learn about its purpose.

📄 SKILL.md
📁 scripts/
📁 references/
📁 assets/
SKILL.md (Required)

The brain of your skill. It contains instructions written in Markdown, paired with a YAML frontmatter block at the top that tells Claude exactly when and how to use this skill.

Knowledge Check

Which component of a Claude skill is absolutely required for the skill to function?

Fundamentals

The Secret Sauce: Progressive Disclosure

To optimize performance and save tokens, skills operate on a principle called Progressive Disclosure. Claude doesn't load every single file in your skill folder at once. Instead, it uses a three-level system.

Click each level to expand

Level 1: Always Loaded
The YAML frontmatter inside SKILL.md. It is always loaded in Claude's system prompt, providing just enough info for Claude to know when the skill is relevant.
Level 2: Loaded when relevant
The body of SKILL.md. Claude only loads this full instruction set when it decides the skill is actually needed for the current user task.
Level 3: Discovered as needed
The Linked files (scripts, references, assets). Claude navigates to and loads these files entirely on-demand during execution.

Composability

Claude isn't limited to using one skill at a time. The system is designed for composability.

Because of Progressive Disclosure, Claude can hold the Level 1 YAML data for dozens of skills simultaneously. When you give it a complex task, it can dynamically load and mix multiple skills. For instance, it can combine a "Research Methodology" skill with a "Brand Tone formatting" skill to produce a perfectly styled, accurately researched document in one pass.

Pro tip: Ensure your skills do one thing extremely well, rather than trying to make a single "do everything" skill folder.

Key Takeaways

  • Skills are Folders: They teach Claude workflows permanently so you don't have to repeat yourself.
  • SKILL.md is the Brain: The only strictly required file is `SKILL.md`, containing YAML frontmatter and Markdown instructions.
  • Progressive Disclosure saves tokens: Only the YAML is always loaded. Instructions and files are loaded dynamically when relevant.
  • Composability: Claude can dynamically mix and match multiple skills for complex tasks.

Final Assessment

You've completed the tutorial on Building Skills for Claude. Let's verify your understanding.

  • There are 3 questions.
  • You must score 100% to earn the completion certificate.
  • Once you answer and proceed, your choices are locked.

Question 1 of 3

What is the primary benefit of building a Claude Skill instead of using a standard chat prompt?

Question 2 of 3

In the "Progressive Disclosure" model, what specific piece of information is ALWAYS loaded into Claude's system prompt to help it decide when to use a skill?

Question 3 of 3

You are building a skill and want Claude to automate data processing by running Python. Where should you place your Python code in the skill structure?

Protocol Results

--%