// the find
anthropics/skills
Public repository for Agent Skills
Anthropic's official repository of 'skills' for Claude — structured folders containing SKILL.md files with YAML frontmatter and markdown instructions that Claude loads to handle specialized tasks. Covers document manipulation (docx/pptx/pdf/xlsx), design, MCP server generation, and enterprise comms. Aimed at developers and teams wanting to extend Claude's behavior for specific workflows without writing code.
- The production document skills (docx, pptx, pdf, xlsx) are the real value here — they include actual Python scripts with XSD schema validation against ISO-IEC29500, LibreOffice integration, and redline/change-tracking logic that reveals how Anthropic handles complex Office formats internally.
- The skill format itself is minimal and sane: just a YAML frontmatter block and markdown. No framework lock-in beyond the Claude platform. Creating a custom skill is as low-friction as writing a README.
- The mcp-builder skill includes an evaluation framework (connections.py, evaluation.py, example_evaluation.xml) and a visual HTML eval viewer, which is a genuinely useful pattern for testing AI-generated code.
- Mixed licensing is handled honestly — Apache 2.0 for examples, source-available for the production document skills — rather than pretending everything is open source.
- The skills system is entirely Claude-platform-specific. There's no way to run or test skills locally outside of Claude.ai or the Claude API, so you're debugging blind unless you're paying for API access.
- Schema duplication is wasteful and maintenance-unfriendly: the full ISO-IEC29500 XSD set appears to be copied verbatim under both skills/docx and skills/pptx, making the repo unnecessarily large and divergence-prone.
- The disclaimer ('implementations and behaviors you receive from Claude may differ') is doing a lot of work. There's no versioning or compatibility contract between skills and Claude model versions, so a skill that works today may silently degrade after a model update.
- The star count (138k) is almost certainly inflated by hype and the Anthropic brand rather than real adoption signals — the actual community contribution depth is thin, with no third-party skill PRs or substantive issue tracker activity visible.