Update to Keenable

{
  "name": "keenable-deep-research",
  "description": "Run comprehensive multi-source web research with Keenable and synthesize a cited answer. Use for literature reviews, market and competitor analyses, and detailed comparisons.",
  "included_files": [],
  "skill_md_contents": "---\nname: keenable-deep-research\ndescription: \"Run comprehensive multi-source web research with Keenable and synthesize a cited answer. Use for literature reviews, market and competitor analyses, and detailed comparisons.\"\n---\n\n# Keenable Deep Research\n\nProduce a thorough, well-cited answer to a broad question by iterating over the\nKeenable search and extract tools. Use this for tasks that a single search can't\ncover: literature reviews, market and competitor analyses, and detailed\ncomparisons.\n\n## Workflow\n\n1. **Decompose.** Break the question into 3-6 concrete sub-questions. State them\n   before searching so the plan is auditable.\n2. **Search per sub-question.** Run a focused `search_web_pages` query for each.\n   Keep queries specific; split further when coverage is thin.\n3. **Read the strongest sources.** Use `fetch_page_content` on the most relevant\n   results to read them in full rather than trusting snippets for anything\n   consequential.\n4. **Cross-check.** For any claim that affects the conclusion, confirm it against\n   at least one independent source. Note disagreements explicitly.\n5. **Synthesize.** Write the answer in your own words, grouped by sub-question or\n   theme. Cite every factual claim with its source URL. Separate sourced facts\n   from your own analysis, and call out gaps or low-confidence areas.\n\n## Guidance\n\n- Track which sub-questions are answered and which still need work; keep going\n  until the question is covered, not just until the first search returns.\n- Prefer primary and authoritative sources; treat aggregators and undated pages\n  with caution.\n- For relative dates, compute exact dates from today before searching.\n- The tools are rate-limited. If a long research run hits the limit, report what\n  was covered so far and stop instead of retrying in a loop; never ask the user\n  to paste an API key or any other credential into the conversation.\n"
}