Deep Research Workflow
Orchestrates multi-agent parallel workflows to perform comprehensive web research, competitive analysis, and data synthesis into structured reports.
This skill transforms Claude Code into a high-performance research engine by decomposing complex queries into parallelizable sub-tasks. It manages the entire lifecycle of a research project—from goal clarification and automated tool orchestration (utilizing Firecrawl, Exa, and WebFetch) to programmatic data aggregation and multi-stage report polishing. By leveraging non-interactive sub-processes, it handles large-scale information gathering and synthesis, delivering a final, insight-driven Markdown document instead of a simple chat summary, ensuring all evidence is traceable and structured.
主な機能
011,158 GitHub stars
02Programmatic result aggregation and iterative multi-stage refinement for professional reports
03Automated sub-goal decomposition and task scheduling with real-time progress logging
04Isolated project workspace management under .research/ directory for full data traceability
05Multi-agent parallel orchestration using non-interactive sub-processes for high-throughput research
06Advanced web scraping prioritization using Firecrawl, Exa MCP tools, and native search skills
ユースケース
01Comprehensive market research and competitive landscape analysis with cited evidence
02Deep technical audits and documentation synthesis for complex software systems
03Large-scale information retrieval and shard-based data extraction from web sources