Deep Research Methodology - Transform vague topics into high-quality research reports
npx claudepluginhub wshuyi/deep-researchDeep Research Methodology (8-step method): Transform vague topics into high-quality research reports with systematic fact extraction, framework comparison, independent Agent verification, and verifiable conclusions.
Claude Code marketplace entries for the plugin-safe Antigravity Awesome Skills library and its compatible editorial bundles.
Production-ready workflow orchestration with 80 focused plugins, 185 specialized agents, and 153 skills - optimized for granular installation and minimal token usage
Agent skills for building and maintaining promptfoo evaluations
Share bugs, ideas, or general feedback.
A Claude Code Skill that transforms vague topics into high-quality, deliverable research reports using a systematic 8-step methodology with independent verification.
When researching complex topics, you often face:
Traditional research is time-consuming, and conclusions are often unreliable.
Deep Research provides a systematic 8-step methodology that:
git clone https://github.com/wshuyi/deep-research.git
cp -r deep-research/skills/deep-research ~/.claude/skills/
/plugin marketplace add wshuyi/deep-research
/plugin install deep-research@wshuyi/deep-research
After the repo gets 2+ stars, it will be automatically indexed by SkillsMP. Search for "deep-research" there.
Natural language triggers:
Step 0: Problem type identification
Step 0.5: Time-sensitivity assessment (BLOCKING for AI/tech topics)
Step 1: Problem decomposition & boundary definition
Step 2: Source tiering & authority locking
Step 3: Fact extraction & evidence cards
Step 4: Build comparison framework
Step 5: Reference alignment
Step 6: Fact → Conclusion derivation chain
Step 6.5: Independent Agent verification (BLOCKING) ← NEW in v2.1
Step 7: Use case validation (sanity check)
Step 8: Deliverable formatting
~/Downloads/research/<topic>/
├── 00_问题拆解.md # Problem decomposition
├── 01_资料来源.md # Source documentation
├── 02_事实卡片.md # Fact cards
├── 03_对比框架.md # Comparison framework
├── 04_推导过程.md # Derivation process
├── 05.5_校验记录.md # Independent Agent verification records (NEW)
├── 05_验证记录.md # Validation records
└── FINAL_调研报告.md # Final deliverable
User: 深度调研 REST API 和 GraphQL 的区别
Claude: [Executes 8-step methodology]
- Identifies as: Concept Comparison type
- Creates fact cards from official specs
- Uses 8-dimension comparison framework
- Validates with real-world scenarios
- Outputs structured report with citations
| Feature | Description |
|---|---|
| Source Tiering | L1-L4 hierarchy ensures conclusion traceability |
| Time-sensitivity | Auto-detects fast-moving fields (AI, crypto, etc.) |
| Fact Cards | Every claim has source, confidence level, applicability |
| Independent Verification | Separate Agent validates facts & logic before final report |
| Explicit Derivation | No "I feel like" - only mechanism-based conclusions |
| Quality Checklists | Boundary guard, scope creep prevention, risk-level distinction |
| Deliverable Output | One-line summary + structured chapters + citations |
Q: How is this different from just asking Claude to research something?
A: This skill enforces a systematic methodology with intermediate artifacts, source verification, independent Agent verification (Step 6.5), and explicit reasoning chains. Results are traceable and verifiable.
Q: Does it work for non-technical topics?
A: Yes, the methodology applies to any research topic. The 5 problem types (concept comparison, decision support, trend analysis, problem diagnosis, knowledge organization) cover most use cases.
Q: How does it handle outdated information?
A: Step 0.5 assesses time-sensitivity. For high-sensitivity fields (AI, blockchain), it enforces 6-month time windows and requires version number citations.
Q: What is Independent Agent Verification (Step 6.5)?