majiayu000 Deep Research Skill
Perform thorough research using web search, documentation, and intelligent synthesis to inform development decisions.
Capabilities
- Web search via WebSearch tool
- Documentation analysis
- Technology trend research
- Competitive analysis
- Best practices discovery
- Academic/technical paper review
Research Methodology
Phase 1: Broad Search (10-15 sources)
- Query multiple search engines
- Scan for credibility (check date, author, domain)
- Filter by relevance score
- Prioritize official docs, established blogs, GitHub repos
Phase 2: Deep Dive (Top 5 sources)
- Read thoroughly
- Extract key insights
- Identify patterns and trends
- Note contradictions or debates
- Look for code examples and real-world applications
Phase 3: Synthesis
- Combine findings into cohesive narrative
- Create actionable recommendations
- Document all sources
- Generate summary report
Output Format
Research saved to: temp/research/{topic}-{timestamp}.md
# Research: {Topic}
## Executive Summary
[3-5 bullet points - key findings]
## Key Findings
### 1. {Finding Title}
- **Source**: [Link](url)
- **Insight**: What was learned
- **Actionable**: How to apply this
- **Code Example**: (if applicable)
### 2. {Finding Title}
...
## Recommendations
1. **Immediate Action**: What to do now
2. **Best Practice**: Pattern to follow
3. **Avoid**: What not to do
## Implementation Plan
- [ ] Step 1
- [ ] Step 2
## Sources
- [Title](URL) - Brief description
- [Title](URL) - Brief description
Usage Examples
Technology Research
"deep research on LangGraph supervisor pattern for production systems
Focus on: state management, error handling, scalability
Save to: temp/research/langgraph-supervisor.md"
Competitive Analysis
"research competitors in AI code generation space
Analyze: features, pricing, tech stack, user feedback
Identify: gaps we can fill, unique angles
Output: temp/research/competitive-analysis.md"
Best Practices
"research React Server Components best practices for Next.js 14
Include: when to use vs client components, data fetching patterns, common pitfalls
Find: code examples from Vercel and community
Save: temp/research/rsc-best-practices.md"
Integration with Build Process
Research findings automatically:
- Update Learning: Add insights to
directives/learning.json - Create Specs: If features found → add to backlog
- Improve Docs: Suggest updates to INSTRUCTIONS.md
- Inform Architecture: Use findings in technical decisions
Research Quality Checklist
Before completing research:
- At least 5 credible sources
- Checked for recency (prefer <1 year old info)
- Included official documentation
- Found real-world examples/code
- Synthesized conflicting information
- Created actionable recommendations
- Documented all sources with working links
Advanced Research Patterns
Comparative Research
"research and compare:
Option A: Using Prisma ORM
Option B: Using raw SQL with Postgres
Option C: Using Drizzle ORM
Compare: performance, DX, type safety, migrations, community support
Recommend: Best option for Next.js 14 + Supabase stack"
Trend Analysis
"research current trends in AI agent orchestration frameworks
Analyze: LangGraph, CrewAI, AutoGPT, LangChain, Semantic Kernel
Identify: Which is gaining traction, production-ready, best for SaaS
Timeline: Last 6 months only"
Remember: Great research leads to better decisions. Invest time in deep research before implementation!