What Are Claude Skills? How Agent Skills Work and Where to Find Them

Claude skills, which Anthropic calls Agent Skills, are folders containing a SKILL.md file of instructions, plus optional scripts and reference files, that Claude loads automatically when your task matches the skill’s description. You write a workflow down once, and Claude applies it whenever it is relevant, without you pasting the same prompt into every conversation.

This guide covers what goes inside a skill, how Claude decides to use one, where skills work, how to add one to Claude Code, and how to find good skills without relying on made-up rankings. Every claim about how skills work comes from Anthropic’s own documentation, listed in the sources at the end.

What is inside a Claude skill?

A skill is a directory. The only required file is SKILL.md, which starts with YAML frontmatter and continues with plain Markdown instructions. A minimal example:

---
name: release-notes
description: Writes release notes from merged pull requests. Use when the user asks for a changelog or release notes.
---

# Release notes

## Instructions
1. List the merged pull requests since the last tag.
2. Group them into Added, Changed and Fixed.
3. Write one plain sentence per change.

Anthropic’s Agent Skills overview lists two required fields and these limits:

• name: up to 64 characters, lowercase letters, numbers and hyphens only, no XML tags, and it cannot contain the reserved words “anthropic” or “claude”.
• description: non-empty, up to 1,024 characters, no XML tags. It should say both what the skill does and when Claude should use it.

Claude Code is a little looser: its docs treat every frontmatter field as optional, default the name to the folder name, and recommend a description. It also adds optional fields such as allowed-tools and disable-model-invocation, which stops Claude from loading a skill on its own, useful for something like a deploy workflow.

Beyond SKILL.md, a skill can bundle extra Markdown files (a forms guide, an API reference), executable scripts, and resources such as templates or schemas.

How does Claude decide when to use a skill?

Skills use what Anthropic calls progressive disclosure: Claude loads information in stages instead of reading everything up front. The figures below come from the Agent Skills overview.

The practical consequence: the description does the matching. If yours is vague, Claude will not know when to reach for the skill. Because unused skills cost only their metadata, you can keep many installed without filling the context window.

Where do Claude skills work?

Skills run across Claude’s agent products, but each surface handles them differently.

Custom skills do not sync between these surfaces. A skill you upload to claude.ai is not available in the API, and Claude Code skills live on your filesystem, separate from both. Anthropic also notes the Claude Agent SDK supports skills, and in December 2025 it published Agent Skills as an open standard at agentskills.io.

How do you add a skill to Claude Code?

This follows the steps in the Claude Code skills documentation.

1. Create the folder. For a skill available in all your projects, run mkdir -p ~/.claude/skills/release-notes. For one that belongs to a single repository, use .claude/skills/release-notes and commit it so your team gets it.
2. Write SKILL.md. Add frontmatter with a description that says what the skill does and when to use it, then the instructions in Markdown.
3. Move detail into supporting files. The docs suggest keeping SKILL.md under 500 lines and putting long reference material in separate files.
4. Test it. Ask for something that matches the description, or run it directly by typing /release-notes at the start of a message.
5. Check it is listed. Run /skills to see every skill Claude Code can use. Edits to skill folders are picked up during the session.

To install someone else’s skills packaged as a plugin, add their marketplace and install from it. For Anthropic’s example repository that is /plugin marketplace add anthropics/skills, then /plugin install example-skills@anthropic-agent-skills. Plugin skills show up namespaced, as /plugin-name:skill-name.

Is there a Claude Code skills marketplace?

In Claude Code, a marketplace is a source of plugins, and plugins can carry skills. Anthropic runs an official one, claude-plugins-official, which hosts the skill-creator plugin for building and evaluating skills. Its anthropics/skills repository can also be added as a marketplace.

For browsing across many authors, directories help. The FoundrList skills directory lists SKILL.md playbooks that people publish for Claude Code and other coding agents. You can filter by agent and by category, sort by newest, trending, most installed or most upvoted, and open any skill page to read its SKILL.md, README, files, changelog and install steps before you add it.

What are the best Claude skills?

There is no official ranking, so be wary of lists that claim one. A better question is which kinds of skill pay off. Anthropic groups its example skills into creative and design, development and technical, enterprise and communication, and document skills for Word, Excel, PowerPoint and PDF. Those are a good reference for structure, though Anthropic labels the repository as demonstration material.

When you compare skills, check:

• The description. Does it say clearly when the skill should trigger?
• The size. A focused SKILL.md with reference files beats one enormous file.
• The scripts. Can you read every one and understand what it does?
• The source. Do you trust the author, and is the code public?

What about marketing skills for Claude?

Marketing is a natural fit, because so much of it is a repeatable checklist. Anthropic itself cites applying brand guidelines as a use case. Typical marketing skills encode a brand voice, a content brief format, or an SEO review routine. On FoundrList you can filter the directory by the Marketing, SEO and Content categories. If your skill does search work, pair it with our guide to getting indexed in Google and AI search and the free llms.txt generator.

Are Claude skills safe to install?

Treat a skill like software. Anthropic advises using skills only from trusted sources and auditing every bundled file, including scripts and images, before use. A malicious skill can direct Claude to call tools or run code that does not match its stated purpose. In Claude Code this matters more, because skills have the same network access as any other program on your computer.

How do you publish your own skill?

Once a skill works for you, share it. On FoundrList, go to publish a skill and paste the public GitHub repository URL. FoundrList reads the repository details and its SKILL.md to pre-fill the listing, and you choose the agents and category it fits.

Frequently asked questions

Sources

• Agent Skills (Claude Docs), platform.claude.com/docs/en/agents-and-tools/agent-skills/overview, accessed October 27, 2026.
• Extend Claude with skills (Claude Code Docs), code.claude.com/docs/en/skills, accessed October 27, 2026.
• Introducing Agent Skills (Anthropic), claude.com/blog/skills, accessed October 27, 2026.
• anthropics/skills (GitHub), github.com/anthropics/skills, accessed October 27, 2026.