Frequently Asked Questions

A SKILL.md is a plain Markdown file with YAML frontmatter that teaches Claude domain expertise, conventions, or workflows. It is the foundation of Claude Agent Skills.

Every skill on skill-dot-md.com is analyzed by Claude API before it appears in the catalog. Skills are scored as Safe, Review needed, or Dangerous. Dangerous skills are held for admin review and never shown publicly. That said, always read a skill before installing — treat it like open-source code.

Skills are designed primarily for Claude Code (the CLI). You can also paste SKILL.md content directly into a system prompt when using the Claude API. Support in other Claude interfaces varies.

There is no hard limit, but skills use context. Lazy loading mitigates this — Claude only reads the full file when relevant. Five to fifteen focused skills per project is a practical guideline.

Yes. A skill is just a Markdown file. Edit it, extend it, or fork the original repository — respecting the author's license.

Delete the SKILL.md file from .claude/skills/ in your project. For globally installed skills, remove it from ~/.claude/skills/.

The AI analysis may have flagged prompt injection patterns, requests for sensitive file access, or external network calls. Review the safety notes on your skill's page, address the concerns, and re-submit.

No. This is an independent community catalog. The Claude Agent Skills format is defined by Anthropic, but this catalog is community-built and not officially endorsed.

Skills provide knowledge — conventions, workflows, domain expertise — as plain text. MCP tools provide functionality — connecting Claude to APIs, databases, and external systems. They are complementary: a skill can describe how to use an MCP tool.

Yes. Include the SKILL.md content in your system prompt when calling the Claude API. The format is designed for Claude Code, but the knowledge works anywhere you inject it as context.