Where are my research results and sessions stored?

All research data, including raw artifacts, queries, and final synthesis files, are stored in the XDG-compliant directory `~/.config/LEV/research/` for easy retrieval and auditing.

How does the synthesis component work?

The system uses the Oracle backend (powered by GPT-5 class models) to fuse results from multiple search providers into a single, grounded, and cited markdown report.

What is the difference between 'quick' and 'deep' modes?

Quick mode performs a fast web lookup using Brave or Firecrawl for immediate answers, while Deep mode utilizes recursive loops via Valyu to verify information until it meets a high confidence threshold.

Can I resume a previous research session?

Absolutely. You can use the `research resume ` command to pick up exactly where a previous investigation left off.

Does this skill support social media and academic searches?

Yes, it includes specialized routes for 'social' (Grok and Hacker News) and 'academic' (Arxiv and neural search) to find trending topics or peer-reviewed papers.

LEV Deep Research Intelligence

Name: LEV Deep Research Intelligence
Author: lev-os

bylev-os

0•

ウェブスクレイピングとデータ収集

Orchestrates multi-layered web research and data synthesis using recursive confidence loops and specialized search engines.

LEV Research is a sophisticated intelligence system for Claude Code designed to streamline complex information gathering. By routing queries through an optimized hierarchy of backends—including Brave, Firecrawl, Valyu, and Exa—it eliminates decision fatigue and provides the most efficient path to accurate data. Whether you need a lightning-fast web lookup or a deep, recursive OSINT investigation, the skill manages sessions, tracks citations, and uses multi-perspective AI synthesis to deliver high-confidence reports directly into your workflow.

主な機能

01Adaptive routing across Quick, Full, Deep, and Max research modes

02Specialized search adapters for Social (X/Grok), Academic (Arxiv), and OSINT (Sherlock)

03Recursive confidence loops that iterate until research reaches 85% certainty

040 GitHub stars

05Unified session management with XDG-compliant storage for queries and artifacts

06Structured data extraction and multi-source synthesis using Oracle AI models

ユースケース

01Conducting deep-dive technical research on emerging software frameworks and libraries

02Automating the collection and synthesis of market data into structured markdown reports

03Performing OSINT investigations and real-time social sentiment analysis

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add lev-os/agents lev-research

For use in Claude.ai and ChatGPT