This handles fetching dbt documentation in a format that's actually useful for Claude. The key insight is appending .md to any docs.getdbt.com URL to get clean markdown instead of HTML. It includes a search script that filters through the full docs by keyword, which beats trying to load everything at once. Start with the llms.txt index to find pages by title, then grab individual pages as needed. The explicit warnings about treating fetched content as untrusted are worth noting, since documentation sites can contain all sorts of embedded examples that might look like instructions.
npx skills add https://github.com/dbt-labs/dbt-agent-skills --skill fetching-dbt-docs