Multi-source search + evidence compilation + structured report.
Call
python3 skills/deep-research/run.py --topic '研究主题' --queries '["关键词1","关键词2"]' --max_results 5 --depth basic
Parameters
| Parameter | Type | Required | Description |
|---|---|---|---|
| topic | string | Yes | Research topic |
| queries | string[] | No | Search keywords, 3 are automatically generated by default |
| max_results | int | No | Number of results per query, default 5 |
| depth | string | No | "basic" or "advanced" |
| fetch_urls | string[] | No | Additional URLs to fetch full text from |
Output
Returns JSON, containing searches(search results),fetched_pages(fetched pages),summary_prompt(synthesis prompt).
After the LLM receives the results, organize them in a structured way following "question → findings/evidence → confidence → impact/recommendations".