michaelschecht Tavily
Use this skill when the work requires API-backed web research with structured results, answer generation, or content extraction.
Prerequisites
- Tavily API key available as
TAVILY_API_KEY - Tavily client or helper script already installed in the target environment
When To Use
- Research current topics with source-backed summaries
- Filter searches to specific domains or exclude noisy sources
- Pull raw page content for deeper analysis
- Run image or news-oriented searches for time-sensitive topics
Search Strategy
- Use
basicdepth for quick fact-finding - Use
advanceddepth for broader research and multiple perspectives - Use news mode for queries involving latest, current, recent, or today
- Keep result counts tight unless the task genuinely needs breadth
Practical Workflow
- Pick
basicoradvanceddepth based on speed vs. coverage. - Decide whether the query is evergreen or news-focused.
- Apply include/exclude domain filters where source quality matters.
- Extract raw content from the strongest URLs if the summary is insufficient.
- Return synthesized findings with source links and recency context.
Notes
- Tavily is best for LLM-oriented retrieval, not exhaustive archival search
- Reduce result count and depth when cost or rate limits matter
- Prefer domain filtering for technical documentation and authoritative research