Deep Research Skill

Description

A multi-agent research orchestration skill that breaks complex research questions into parallel sub-investigations. Each sub-agent focuses on a specific aspect, and results are synthesized into a comprehensive report with citations and confidence levels.

Instructions

When the user asks you to research a topic in depth, follow this workflow:

Step 1: Decompose the Research Question

Break the question into 3-6 independent research tracks:

Copy
## Research: "Should we migrate from REST to GraphQL?"

### Sub-Investigations:
1. **Current API analysis** — Catalog existing endpoints, payload sizes, over-fetching patterns
2. **GraphQL benefits assessment** — Query flexibility, type safety, real-time subscriptions
3. **Migration cost analysis** — Tooling, learning curve, client-side changes required
4. **Performance comparison** — Caching strategies, N+1 problems, batching
5. **Team readiness** — Current skill set, training requirements, hiring market
6. **Case studies** — Companies of similar size that migrated, outcomes

Step 2: Dispatch Parallel Sub-Agents

Launch each sub-investigation as an independent subagent with clear scope:

Subagent 1: Analyze src/routes/ and src/controllers/ for API patterns
Subagent 2: Research GraphQL benefits via web search and documentation
Subagent 3: Estimate migration effort based on codebase size
Subagent 4: Benchmark current API response times, identify bottlenecks

Step 3: Synthesize Findings

Combine all sub-agent results into a structured report:

Copy
## Research Report: REST to GraphQL Migration

### Executive Summary
[2-3 sentence conclusion with recommendation]

### Findings by Track

#### 1. Current API Analysis
- 47 REST endpoints across 12 controllers
- Average response payload: 4.2KB (23% unused fields)
- 8 endpoints require multiple sequential calls
- **Conclusion:** Moderate over-fetching problem

#### 2. GraphQL Benefits
...

### Recommendation
| Factor | REST (stay) | GraphQL (migrate) |
|--------|------------|-------------------|
| Development speed | Current | +30% after ramp-up |
| Performance | Adequate | Better for mobile |
| Migration cost | $0 | ~3 person-months |
| Risk | Low | Medium |

### Confidence Level: HIGH (based on 6 completed sub-investigations)

### Sources
1. Codebase analysis: src/routes/, src/controllers/
2. GraphQL Foundation: graphql.org/learn
3. ...

Rules

• Always decompose into at least 3 sub-investigations
• Each sub-agent should have a clear, bounded scope
• Include confidence levels (HIGH/MEDIUM/LOW) for each finding
• Cite sources — file paths for code analysis, URLs for web research
• Present trade-offs objectively, then give a clear recommendation
• If sub-investigations contradict each other, highlight the conflict explicitly
• Time-box research: spend no more than 2-3 minutes per sub-investigation
• Always include an "Executive Summary" at the top for busy readers

Examples

User: Research whether we should use Postgres or DynamoDB for our new service Action: Dispatch sub-agents for: data model analysis, cost comparison, scaling requirements, team expertise, operational overhead

User: Investigate why our build times have doubled Action: Dispatch sub-agents for: dependency tree analysis, webpack/bundler config, CI pipeline steps, test suite duration, cache effectiveness

User: Research authentication approaches for our API Action: Dispatch sub-agents for: JWT vs sessions, OAuth providers, security requirements, existing infrastructure, migration path