What is the difference between Expand and Deep modes?

Expand mode iteratively refines queries based on previous results for exploration, while Deep mode performs parallel multi-query synthesis for exhaustive analysis.

What backends does the Deep Research skill use?

It primarily uses Tavily for AI-optimized search, while mandating the loading of Exa (neural/GitHub search) and Valyu (recursive research) for comprehensive discovery.

Does this skill require specific API keys?

Yes, it requires a TAVILY_API_KEY and an EXA_API_KEY to be configured in your environment or configuration files.

How does the smart stopping feature work?

The agent automatically stops searching when sources consistently agree (confidence threshold), when follow-up queries return redundant results, or when the maximum iteration limit is reached.

Deep Research Agent

Name: Deep Research Agent
Author: lev-os

bylev-os

0•

생산성 및 워크플로

Conducts multi-mode web research and synthesis using Tavily, Exa, and Valyu for comprehensive information discovery.

The Deep Research skill transforms Claude into a sophisticated research agent capable of complex, iterative information gathering. By leveraging multiple backends like Tavily, Exa, and Valyu, it goes beyond simple search results to provide multi-query synthesis, automated query refinement, and smart stopping based on confidence thresholds. Whether performing a quick fact-check or a deep dive into technical documentation and market trends, this tool orchestrates parallel searches to deliver structured markdown reports complete with source analysis and research traces.

주요 기능

01Smart stopping logic that halts research upon reaching confidence thresholds or information saturation.

02Mandatory integration with Exa and Valyu backends for neural and recursive research depth.

03Parallel search orchestration that analyzes a topic from multiple angles simultaneously.

040 GitHub stars

05Three specialized modes: Quick Search, Iterative Expansion, and Deep Multi-Query Synthesis.

06Structured markdown output featuring synthesized summaries, source relevance scores, and research traces.

사용 사례

01Automating the learning process for complex new domains through iterative query refinement.

02Performing technical due diligence and architectural pattern research for software projects.

03Conducting deep-dive competitive analysis and market trend discovery with multi-source verification.

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add lev-os/agents deep-research

For use in Claude.ai and ChatGPT

Download Skill