Anthropic Official Skills

Anthropic publishes a set of reference skills — artifacts-builder, pdf, docx, and others — in the anthropics/skills repo. These are the canonical reference implementations of the agentskills.io standard: the right place to read when you want to learn the SKILL.md anatomy, frontmatter conventions, and progressive-disclosure patterns by example.

When to use

• You're learning how to write a skill — read these first
• You need a known-good starting point for a specific output format (PDF, DOCX, structured artifacts)
• You want to fork and adapt a reference skill for your own use case

When not to use

• You want a third-party opinionated framework — see Superpowers instead
• You need a skill for a domain Anthropic hasn't published (these are format-oriented, not domain-oriented)

How it works

Each skill in the repo is a self-contained folder with a SKILL.md and (where applicable) scripts and reference materials. They demonstrate the standard's six frontmatter fields, bounded allowed-tools declarations, and tiered progressive disclosure. Because they're written by the standard's authors, they're the lowest-risk installs in the ecosystem — but "lowest-risk" is not "no risk"; still re-vet before installing in any environment with real credentials.

Install

npx skills add anthropics/skills

Full SKILL.md

Full source at the canonical repo: github.com/anthropics/skills. Browse the skills/ directory for the individual skills and their SKILL.md files. Embedded copy omitted at launch — the canonical repo is the authoritative source.

Verify it yourself

Quick static check — even reference implementations should be scanned. The pdf and docx skills in particular bundle scripts; confirm any scripts/*.sh or scripts/*.py is doing only the format conversion it claims and nothing more.

Two-model review — paste each SKILL.md you intend to install into two different models with the Vetting Guide prompt. Reference skills are usually LOW risk, but the pattern of "trusted because it's official" is exactly what an attacker would exploit by mirroring the repo.

Review transparency

Related

• Skills Collection Overview
• Vetting Guide
• Agent Skills — Open Standard