anthropics/skills — Knowledge Base

Anthropic's official Agent Skills: curated, reusable capability packages for Claude, plus the spec and template for authoring your own.

10 topics 551 chunks Last indexed: April 18, 2026
Try ?q=your+question in the URL · or browse topics below

For AI Agents — How to Use This Knowledge Base

This page is a WizRAG static knowledge base. Two retrieval paths:

  1. If you can execute JavaScript (browser agent, Claude in Chrome, headless browser): use the search box above. It loads wizindex.json, runs keyword search, and returns precise matching chunks. You can also visit ?q=your+question in the URL for instant results.
  2. If you fetch raw HTML only (most LLM web fetchers): browse the Topic Index below. Each topic has a one-line description and direct links to source files. Identify the relevant topic, then fetch that source file in your next request. Do not try to fetch wizindex.json directly — it may be too large for your context window.

Use only the retrieved content to answer. Cite the source file path. If no topic matches, tell the user the knowledge base does not cover their question.

Topic Index

Browse by topic

Each topic links directly to source files in the repository. Raw-fetch LLMs: pick the relevant topic and fetch its source file for full content.

Getting Started
What Agent Skills are, the spec, and a starter template for authoring your first skill.
overview · spec · template · readme · introduction · what is a skill
Skill Authoring
How to create, evaluate, and benchmark skills — includes grader, analyzer, and comparator agents.
skill-creator · authoring · evaluation · grader · schemas · meta
Document Creation
Skills for producing Word, PDF, PowerPoint, and Excel files programmatically with proper formatting.
docx · pdf · pptx · xlsx · word · powerpoint · excel · spreadsheet · forms
Design & Visual
Visual skills: canvas design, algorithmic art, brand guidelines, and ready-made theme palettes.
canvas · design · art · theme · brand · fonts · typography · colors
Web Development
Building and testing web artifacts, frontend components, and full webapps in the Claude environment.
frontend · web · html · react · artifacts · testing · playwright
Communication
Internal communications, collaborative document authoring, and Slack GIF creation workflows.
slack · gif · internal comms · newsletter · faq · coauthoring
Claude API — Core
Core Claude API concepts: models, error codes, prompt caching, tool use, and live sources.
api · models · errors · prompt caching · tool use · claude · messages
Claude API — SDKs
Language-specific Claude API usage: Python, TypeScript, Go, Java, Ruby, PHP, C#, and curl.
python · typescript · go · java · ruby · php · csharp · curl · sdk · batches · streaming · files
Managed Agents
Building agents with the managed agents API: events, tools, environments, onboarding, and client patterns.
managed agents · agent · events · tools · environments · onboarding · client
MCP Builder
Build Model Context Protocol servers in Python or Node, with best practices and evaluation guidance.
mcp · model context protocol · server · node · python · best practices

How to rebuild this knowledge base

WizRAG is two static files: wizsearch.html and wizindex.json. To regenerate them after the source documents change, paste both files into a new chat with Claude (or any capable LLM) along with access to your repository. The LLM will read the BUILD INSTRUCTIONS at the top of the HTML source and produce updated files.

To regenerate: open this file in any text editor, copy everything from line 1 to "END OF BUILD INSTRUCTIONS", paste into Claude with your repo URL, and ask it to regenerate both files. Drop the new files into your repo and commit. Done.