What is Claude Skills?
TL;DR
Anthropic's reusable AI workflow system officially released in late 2025. Defined in a single Markdown file and auto-triggered by description matches. The major Claude Code differentiator.
Claude Skills: Definition & Explanation
Claude Skills is Anthropic's officially released (late 2025) mechanism for defining 'reusable AI workflows' for Claude Code. Skills are written entirely in Markdown with YAML frontmatter, and Claude auto-triggers them when user input matches the natural-language description. As Claude Code's biggest differentiator vs Cursor / GitHub Copilot / Cline, Skills enable engineering organizations to standardize workflows organization-wide. Anatomy: (1) YAML frontmatter (name/description/allowed-tools/env), (2) Markdown body with the Step 1-N procedure, (3) optional combinations of subagents, MCP servers, hooks, and Bash commands. Locations: <code>.claude/skills/</code> (project-shared), <code>~/.claude/skills/</code> (personal), Plugin format on npm/GitHub for enterprise distribution. Examples: PR review, DB migration safety check, incident postmortem, security scan - any repeatable engineering work. Unlike Slash Commands (manual), Skills auto-trigger on intent match. Combine with Subagents to offload heavy work and protect main context. Anthropic plans an official Skills Marketplace for Q1 2026; OSS communities (awesome-claude-code) already publish Skill collections. Adoption metrics: Shopify reports +40% PRs/engineer, Stripe -30% on-call duration, Anthropic-internal new-hire onboarding shrunk to 1 day. Anyone who can write Markdown can build Skills, so PM/design/support automation is also expanding.