Snyk: How Claude Skills Work: Progressive Disclosure Explained

Name: Snyk: How Claude Skills Work: Progressive Disclosure Explained
Uploaded: 2026-08-09T07:10:45-04:00
Duration: 44 s
Description: Watch the full video here: https://youtu.be/EbLfnh0vJlk Use Snyk for free to find and fix security issues in your applications today! https://snyk.co/ugLYn ✍️ Resources ✍️ - Snyk Blog Post: https://snyk.io/articles/top-claude-skills-developers/ - Snyk ...

Snyk

08/09/2026

0 (0%)

Report Like Favorite

TL;DR

A Claude skill is a folder with a skill.md file containing YAML metadata and plain markdown instructions, optionally bundled with scripts and reference docs.
Claude uses progressive disclosure at startup, reading only the name and description of each skill — around 100 tokens — before deciding which to load fully.
Full skill instructions and supporting assets are only loaded when a task match is found, meaning unused skills cost virtually nothing in token overhead.

Summary

This short explainer breaks down exactly what a Claude skill is and why the architecture makes it practical to install many of them without degrading AI performance. A Claude skill is simply a folder containing a skill.md file, which includes a small YAML header with a name and description, followed by plain markdown instructions. Skills can also bundle supporting assets such as shell scripts, templates, and reference documentation alongside the core instruction file. The key engineering insight is the progressive disclosure loading model: when Claude starts up, it reads only the name and description of each installed skill — approximately 100 tokens per skill — and matches the user's task against those lightweight descriptors. Only when a match is found does Claude load the full instruction set, and supporting scripts or references are deferred even further, loading only if actually required during execution. This tiered loading strategy means that having dozens of skills installed carries virtually no token cost for skills that go unused in a given session, making the approach highly scalable for developer workflows that span multiple domains and task types.

Chapters

0:00 - What Is a Claude Skill
0:10 - Bundled Assets and Structure
0:14 - Progressive Disclosure Loading
0:33 - Cost of Unused Skills

Key Quotes

0:00 "A Claude skill is a folder with a file called skill.md inside it."
0:14 "Here's the clever part, and the reason you can install dozens of these without slowing Claude down. It uses what's called progressive disclosure."
0:23 "At startup, Claude only reads the name and description of each skill, roughly 100 tokens each. It matches your task against those descriptions, and only when it finds a match does it load the full instructions."
0:37 "So the cost of having a skill installed but unused is basically nothing."

FAQ

Does installing many Claude skills slow down performance or increase token usage?

No. Claude only reads the name and description of each skill at startup — roughly 100 tokens each. Full instructions and supporting files are only loaded when a skill is actually matched to a task, so unused skills cost virtually nothing.

What files make up a Claude skill?

At minimum, a skill is a folder containing a skill.md file with a YAML header (name and description) and plain markdown instructions. It can optionally include scripts, templates, and reference documentation that load only when needed.

Categories:

» Cybersecurity » Application Security
» Data Protection

Tags:

Show more Show less

TL;DR

Summary

Chapters

Key Quotes

FAQ

Becoming Agent Ready with Cyera: Essential Strategies and Insights

Harnessing AI for Secure Innovation in the Enterprise with Netskope & Omada