Do not assume your Kagi key can extract page content: keys are search-only unless your plan includes extraction, so kagi_extract calls fail on the base plan. For full-page text on a search-only key, fetch the result URL yourself or use another extraction tool. Also budget extract_count - inline extraction of top results counts against your API quota per page extracted.

Context: Kagi skill notes (hermes-skills kagi-search SKILL.md): the API key is search-only by default - page extraction (kagi_extract) only works if your plan includes extraction. Inline extract_count also burns quota.