Deep Researcher

Deep Researcher

mangowhoiscloud Updated 15 repo stars
Crumb + Crumb Studio — 개발 과정과 라이브 시연

File contents

Deep Researcher

Systematically investigates a given topic and generates a structured report.

Workflow

  1. Topic decomposition: Break the user request into 3-5 search queries
  2. Parallel collection: Multi-angle search via web_search (Korean + English)
  3. Deep collection: Fetch body text from top 3-5 URLs via web_fetch
  4. Cross-validation: Only adopt information confirmed by 2+ sources
  5. Report generation: Write a structured markdown report
  6. Save: Record insights to project memory via memory_save

Report Format

## [Topic] Research Report
> Date: YYYY-MM-DD | Sources: N

### Key Findings
- Finding 1 (source: ...)
- Finding 2 (source: ...)

### Detailed Analysis
[Organized by section]

### Sources
- [Title](URL) — Summary

Guidelines

  • Perform at least 3 searches with different keywords
  • No single-source dependency — cross-validation required
  • Prioritize recent sources for time-sensitive information
  • Clearly distinguish speculation from facts
  • Write report in English (keep original-language citations as-is)

mangowhoiscloud/crumb commit f503d8dee3

Frequently asked questions

npx get-skill add mangowhoiscloud/deep-researcher-2