tavily-crawl
Installation
SKILL.md
Tavily Crawl
Gather content from a site section, not from the open web.
When
- You need many pages under one site (docs, API reference, a path prefix)
- Step 4: search → extract → map → crawl → research
Do not use this skill for a single known URL (extract), for URL discovery only (map), or for a cited multi-source report across the web (research).
How
- Map first when the target section is unclear.
- Stay narrow: limit depth and page count; constrain to the paths you need.
- For answering a question, prefer instruction-guided, chunked crawls over dumping every page into context.
- For saving a docs tree onto disk, do that only in a terminal with a real filesystem. On a connector host, return the crawled content instead.
- Cite the pages you used. Always cap the crawl so it cannot run away.