Are you an LLM? Read llms.txt for a summary of the docs, or llms-full.txt for the full context.
Skip to content

Research Skill

Enables agents to research current best practices, design patterns, and implementation approaches for technologies and product domains.

What It Provides

Research Methods
  • Context7 MCP integration patterns (preferred)
  • WebSearch fallback strategies
  • Query formulation techniques
  • Result synthesis methods
Research Workflows
  • Tech stack analysis procedures
  • Domain pattern research
  • Pre-implementation research
  • Trend analysis approaches
Quality Criteria
  • Result evaluation standards (recency, authority, specificity)
  • Source prioritization rules
  • Actionability assessment

Used By

/agentful-generate Command:
  • Researches best practices for detected tech stack
  • Informs domain agent generation
  • Guides skill creation with current patterns
Orchestrator Agent:
  • Before implementing complex features
  • When architectural decisions needed
  • For comparing implementation approaches
Agent Generation:
  • Researching framework-specific patterns
  • Understanding domain implementation approaches
  • Informing specialized agent creation

Tools Available

  • Context7 MCP (if connected)
  • WebSearch (always available)
  • Read (tech stack detection, product specs)
  • Write (research synthesis reports)

Research Methods

1. Context7 MCP (Preferred)

Note: Context7 is an optional MCP dependency. If not installed, the skill falls back to WebSearch.

Advantages:
  • Curated, accurate results
  • Up-to-date framework documentation
  • Language-specific patterns
  • Faster than web search
  • No noise or outdated content
Check Availability: 
function hasContext7() {
  return typeof mcp__context7__search !== 'undefined';
}
Usage: 
const results = await mcp__context7__search({
  query: 'Next.js 15 app router best practices',
  limit: 5
});

2. WebSearch (Fallback)

Advantages:
  • Always available
  • Broader coverage (blogs, forums)
  • Real-world examples
  • Community insights
Limitations:
  • May include outdated content
  • Requires filtering and verification
  • More noise
Usage: 
const results = await WebSearch({
  query: 'Next.js 15 app router best practices 2026',
  limit: 5
});
// Always include current year for freshness

Research Workflow

Step 1: Determine Research Method

const researchMethod = hasContext7() ? 'context7' : 'websearch';
console.log(`Research method: ${researchMethod}`);

Step 2: Formulate Queries

Good queries:
  • "Next.js 15 server components best practices"
  • "React 19 concurrent rendering patterns"
  • "PostgreSQL connection pooling with Prisma"
  • "Authentication implementation patterns 2026"
Bad queries:
  • "Next.js" (too broad)
  • "How to code" (too generic)
  • "Best JavaScript framework" (subjective, not actionable)
Query Template: 
[Technology/Framework] + [Specific Topic] + [Year]

Step 3: Execute Research

Research 5-10 targeted queries to balance coverage vs speed.

Step 4: Synthesize Findings

Extract actionable insights:

  • Tech stack patterns
  • Domain-specific patterns
  • Best practices
  • Common pitfalls
  • References

Research Use Cases

Use Case 1: Tech Stack Research

After tech stack detection, research current best practices:

Tech Stack: Next.js 15, TypeScript, PostgreSQL, Prisma
 
Queries:
1. "Next.js 15.1.0 best practices"
2. "Next.js project structure patterns"
3. "Prisma with PostgreSQL optimization"
4. "TypeScript configuration for Next.js"
 
Synthesis:
- Framework patterns for Next.js 15
- Project structure recommendations
- Database optimization approaches
- TypeScript configuration best practices

Use Case 2: Domain Research

Research implementation patterns for product domains:

Product: Task management app
Domains: authentication, tasks, collaboration
 
Queries:
1. "Task management app architecture patterns"
2. "Authentication implementation Next.js"
3. "Task management data model design"
4. "Real-time collaboration implementation"
 
Synthesis:
- Domain-specific implementation patterns
- Database schema design approaches
- API design recommendations

Use Case 3: Pre-Implementation Research

Before implementing complex features:

Feature: Real-time notifications
Technologies: WebSockets, Redis, React
 
Queries:
1. "WebSocket implementation Next.js"
2. "Redis pub/sub patterns notifications"
3. "React real-time updates best practices"
4. "Scaling WebSocket connections"
 
Synthesis:
- WebSocket implementation patterns
- Redis pub/sub approaches
- React real-time update strategies
- Scalability considerations

Quality Criteria

When evaluating research findings:

  1. Recency: Prefer results from current year
  2. Authority: Official docs > verified blogs > forums
  3. Specificity: Specific patterns > general advice
  4. Actionability: Code examples > theory
  5. Relevance: Exact version match > general version

Examples

Tech Stack Research Flow

1. Detect tech stack (Next.js 15, TypeScript, Prisma)
2. Check Context7 availability
3. Research queries:
   - "Next.js 15 best practices"
   - "Next.js project structure"
   - "Prisma patterns"
4. Synthesize findings
5. Apply to agent generation

Domain Research Flow

1. Read product requirements
2. Extract domains (auth, tasks, collaboration)
3. Research each domain:
   - "Authentication implementation patterns"
   - "Task management data model"
   - "Real-time collaboration implementation"
4. Synthesize domain-specific patterns
5. Inform domain agent specialization

Architectural Decision Research

1. Identify complex feature
2. Research implementation approaches:
   - "Feature X implementation approach A"
   - "Feature X implementation approach B"
3. Present findings to user for decision
4. Proceed with chosen approach

Best Practices

  1. Always check Context7 availability first - More accurate results
  2. Include year in WebSearch queries - Ensure freshness
  3. Limit queries to 5-10 - Balance coverage vs speed
  4. Synthesize findings - Don't just dump raw results
  5. Cache research results - Avoid redundant queries
  6. Provide references - Link to source material
  7. Time-box research - Don't spend hours researching

Rules

DO

  • Prefer Context7 when available
  • Include current year in WebSearch queries
  • Formulate specific, targeted queries
  • Synthesize findings into actionable insights
  • Provide references to sources
  • Time-box research activities

DON'T

  • Use overly broad queries
  • Skip synthesis step
  • Ignore result quality criteria
  • Spend excessive time researching
  • Present raw results without context

Implementation Details

Location: .claude/skills/research/SKILL.md

Model: Sonnet

MCP Dependencies:
  • Context7 (optional, preferred)
  • WebSearch (built-in fallback)
Tools Used:
  • Context7 MCP (if available)
  • WebSearch (fallback)
  • Read (tech stack, product specs)
  • Write (research synthesis)

See Also