This gives Claude direct access to Tavily's search API when you need live web data or LLM-friendly extraction. You get four main endpoints: search for queries with optional AI-generated answers, extract for pulling content from specific URLs, map for discovering site structure, and crawl for traversing and extracting in one pass. There's also a research mode that does multi-step investigation with citations, though that's async and takes longer. The skill defaults to conservative settings like basic search depth and five results, which is smart since you can always dial it up. Honestly most useful when you need structured web data with sources rather than just raw HTML scraping.
npx skills add https://github.com/intellectronica/agent-skills --skill tavily