Anthropic Agent Skills
The official Agent Skills collection: reusable SKILL.md folders that give Claude on-demand expertise.
About
Agent Skills are Anthropic's format for packaging expertise an agent can load on demand: each skill is a folder with a SKILL.md (YAML frontmatter plus Markdown instructions) and optional scripts, references, and examples. When the agent's task matches a skill's description, the skill content is injected into context — no fine-tuning, no prompt bloat for everything else.
The anthropics/skills repo is the official collection, and at 181,000+ stars it has become the de facto standard: document skills (DOCX, PDF, PPTX, XLSX), creative and enterprise skills, plus example skills, all installable in Claude Code via the anthropic-agent-skills marketplace. The format itself is open (agentskills.io hosts the spec) and third-party harnesses — OpenCode, Codex, community CLIs — have adopted it. Two caveats: individual skills carry their own licenses, so check before redistributing, and the install path is smoothest inside Anthropic's own tooling.
What it can do
- Official skill collection: document, creative, enterprise, and example skills
- SKILL.md format: Markdown instructions + YAML frontmatter + optional scripts
- On-demand loading — skills enter context only when relevant
- Claude Code marketplace install (`anthropic-agent-skills`)
- Skills API for programmatic creation and use
- Open spec at agentskills.io, adopted by third-party harnesses
Setup
- In Claude Code, add the marketplace:
/plugin marketplace add anthropics/skills. - Install a plugin, e.g.
/plugin install document-skills@anthropic-agent-skillsfor the DOCX/PDF/PPTX/XLSX skills. - The agent now loads matching skills automatically when relevant to the task.
- For the API: create skills via the Skills API (see the Skills API quickstart in the Anthropic docs) and reference them in requests.
- To author your own: write a SKILL.md with
name/descriptionfrontmatter, add optional scripts and references, and validate with theskills-reftool.
Prerequisites
Claude Code, the Claude API, or another harness that supports the Agent Skills format; nothing else for the pre-built skills.
Use cases
Works with
Trust score — 10/10
Scored 10/10 (verified tier, metrics-v1). Evidence: 181,354 GitHub stars (GitHub API), pushed Oct 9 2026, first-party Anthropic repo; the official docs and the agentskills.io spec corroborate the format and install paths; broad third-party adoption (OpenCode, Codex, community CLIs) confirms it is the de facto standard. Strengths: official, open format, enormous adoption, genuinely useful granularity (context only when relevant). Docked only for: (1) per-skill licenses vary — check before redistributing; (2) smoothest inside Anthropic's own tooling. License: various per skill.
★ 181,354 GitHub stars
How the trust score is built
Full methodology →- Adoption · stars/downloads, log-scaled vs peers
- 10.0/10
- Maintenance · recency, releases, issue load
- 10.0/10
- Openness · license + docs
- 9.0/10
- Manual review · sanity read + red-flag check
- 9.0/10
- · Adoption 10: starScore from 181,354 stars on anthropics/skills — exceptional even on the log scale
- · Maintenance 10: pushed 2026-10-09 (recency 10); first-party Anthropic repo under active development
- · Openness 9: per-skill licenses vary (0.7*8-9) + official docs and open spec at agentskills.io (0.3*10)
- · Manual 9: official docs, the agentskills.io spec, and broad third-party adoption (OpenCode, Codex, community CLIs) corroborate the format and install paths; docked only for per-skill license variance and smoothest UX inside Anthropic tooling
Sources
Every fact above was checked against these sources. Where a claim comes from the provider and couldn't be independently verified, the text says so.