Agent Skills (Overview) — Claude Docs
by Anthropic (Claude Docs)
The canonical reference for how Agent Skills work, level by level.
Overview
The official Agent Skills overview in the Claude docs defines Skills as modular capabilities that package instructions, metadata, and optional resources (scripts, templates) that Claude uses automatically when relevant. It explains the filesystem-based architecture that enables progressive disclosure across three levels: Level 1 metadata (name and description in YAML frontmatter, ~100 tokens, always loaded), Level 2 instructions (the SKILL.md body, under 5k tokens, loaded when triggered), and Level 3 resources and code (bundled files read or executed via bash only as needed, with effectively unlimited size). It documents the required SKILL.md frontmatter fields and their validation rules (name max 64 chars, lowercase/numbers/hyphens, no reserved words 'anthropic'/'claude'; description non-empty, max 1024 chars), where Skills work (Claude API with three beta headers, Claude Code as filesystem directories, and claude.ai), pre-built Skills (pptx, xlsx, docx, pdf), open-source Skills in the anthropics/skills repo, security considerations, and cross-surface, sharing, and runtime limitations.
At a Glance
- Topic
- Skills
- Level
- Beginner
- Format
- Documentation
- Cost
- Free
- Duration
- ~12 min read
- Provider
- Anthropic (Claude Docs)
- Hands-on
- No
- Certificate
- None
What You’ll Learn
- ✓The three levels of Skill content and when each loads
- ✓The required SKILL.md YAML frontmatter fields and their validation rules
- ✓How Skills work across the Claude API, Claude Code, and claude.ai
- ✓The API beta headers and skill_id/container parameters needed to use Skills
- ✓Security considerations and runtime constraints per surface
Highlights
- •Authoritative, always-current reference documentation
- •Clear token-cost table for the three loading levels
- •Worked example of loading a PDF-processing skill step by step
Who It’s For
Best For
- ✓Developers implementing or debugging Skills
- ✓Anyone needing the exact SKILL.md field rules
Prerequisites
- •Basic understanding of Claude and agents
- •Comfort reading technical documentation
FAQ
What is Agent Skills (Overview) — Claude Docs?
The official Claude documentation overview of Agent Skills: the three-level progressive-loading model, SKILL.md structure, where Skills run, and their security and runtime constraints. Best as the reference every skill author returns to.
Is Agent Skills (Overview) — Claude Docs free?
Agent Skills (Overview) — Claude Docs is free to access.
What level is Agent Skills (Overview) — Claude Docs for?
Agent Skills (Overview) — Claude Docs is aimed at a beginner audience. Recommended background: Basic understanding of Claude and agents, Comfort reading technical documentation.
How long does Agent Skills (Overview) — Claude Docs take?
Expect roughly ~12 min read. Most learners work through it at their own pace.
What will I learn from Agent Skills (Overview) — Claude Docs?
You'll learn: The three levels of Skill content and when each loads; The required SKILL.md YAML frontmatter fields and their validation rules; How Skills work across the Claude API, Claude Code, and claude.ai; The API beta headers and skill_id/container parameters needed to use Skills; Security considerations and runtime constraints per surface.