Glossary · Anthropic

Claude Skills

Claude Skills are reusable, modular instructions and capabilities you can drop into a Claude project. Definition, ecosystem, and use cases.

By Kadin Nestler · May 28, 2026 · Updated May 28, 2026
On this page

What a skill contains

A Claude skill is a folder with a SKILL.md file at the root and optional supporting files (scripts, templates, reference docs). The SKILL.md has frontmatter naming the skill and a body describing when to invoke it and how it works. Claude reads the skill's description at session start and invokes the full skill when the user request matches. Skills can be local (in your project), user-scoped (in your home directory), or distributed via plugin marketplaces.

Why skills exist

  • Reuse — write a workflow once, invoke it across projects and sessions.
  • Routing — Claude picks the right skill for the task instead of running everything off a generic prompt.
  • Distribution — skills can be shared, versioned, and installed like packages.
  • Composition — multiple skills can be chained or invoked from each other.

Skill ecosystems

Anthropic ships official skills with Claude Code (code review, debugging, security review). Community marketplaces (gstack, seo-geo, Shopify liquid-skills, Vercel agent-skills) distribute hundreds more. Vendor-specific skills exist for Shopify Liquid, Vercel deployments, Stripe, and many others. The pattern is becoming the standard way to package domain expertise on top of a frontier model.

How SMBs benefit indirectly

You probably will not write skills yourself, but the AI agencies and tools you hire do. A vendor with a well-developed skill library for your vertical (insurance, restaurants, legal, healthcare) can ship faster and more consistently than one starting from a blank prompt every engagement. Ask vendors whether they maintain reusable skills and which ones apply to your work.

What it means for your business

A vendor's skill library is essentially their durable IP. If they have one, your engagement gets cheaper and faster as they reuse it. If they do not, every project is greenfield and you pay for it.

  • Claude Agent SDK — The Claude Agent SDK is Anthropic's open-source framework for building production agents on top of Claude. Definition, capabilities, and when to use it.
  • Claude Code — Claude Code is Anthropic's terminal-based AI coding agent — reads your repo, runs commands, edits files, and ships PRs. Definition, pricing, and use cases.
  • Tool Use — Tool use is when an LLM calls external APIs, databases, or code on its own. Definition, function calling, and how it powers AI agents.
  • Model Context Protocol (MCP) — MCP is an open standard for connecting AI models to tools, data, and APIs. Definition, ecosystem, and why it matters for AI interoperability.
  • Prompt Engineering — Prompt engineering is the practice of writing instructions to LLMs to get reliable, structured output. Definition, techniques, and when to stop optimizing.
← Back to the full glossary