Multi-agent research orchestration for gathering technical, architectural, UX, security, and competitive insights. Use before brainstorm, during planning, or for troubleshooting.
Execute parallel research agents to gather comprehensive insights on a topic, then synthesize into actionable reports.
This skill builds on shared orchestration patterns. See:
includes/orchestration/parallel-execution.md - Agent dispatch and concurrencyincludes/orchestration/result-synthesis.md - Collecting and merging resultsincludes/orchestration/context-management.md - Context folding and memoryincludes/orchestration/checkpoint-patterns.md - Save/resume workflowsThe patterns below are research-specific extensions.
Dispatch selected agents with:
After all agents complete:
| Query Contains | Agents to Dispatch |
|---|---|
| "how to implement" | technical, architectural |
| "best practice" | technical, security |
| "UI", "UX", "user" | technical, ux |
| "secure", "auth" | technical, security |
| "compare", "vs" | technical, competitive |
| "market", "trend" | competitive |
| Default | technical, architectural |
docs/.claude/research/{topic-slug}/
├── README.md # Executive summary
├── technical.md # Technical findings
├── architectural.md # Architecture analysis
├── ux.md # UX considerations (if applicable)
├── security.md # Security analysis (if applicable)
├── competitive.md # Market context (if applicable)
└── sources.md # All citations
ALL web searches MUST include current year:
WebSearch(query: "{topic} best practices 2026")
Every finding must include:
Every finding must cite:
/brainstorm "add feature X"
-> Automatically triggers /research "feature X" first
-> Brainstorm receives research context
Error encountered
-> /research:debug "error message"
-> Technical + architectural agents analyze
| Command | Agents | Use Case |
|---|---|---|
/research "topic" | Auto-selected | General research |
/research:quick "topic" | technical only | Fast answer |
/research --agents tech,security "topic" | Specified | Targeted |
/research:debug "error" | tech, arch | Troubleshooting |
After synthesis completes, run consensus review:
documentation-reviewer - Technical accuracy, completenesscode-quality-reviewer - Code example correctness (if applicable)| Requirement | Threshold |
|---|---|
| Consensus Score | >= 70% |
| Critical Findings | 0 |
| High Findings | 0 |
If consensus < 70% OR critical/high findings exist:
See includes/orchestration/consensus-review.md for detailed patterns.