| name | spawn-research-agents |
| description | Use when conducting codebase research to orchestrate specialized agents in parallel for comprehensive investigation. |
Spawn Research Agents
Use specialized agents in parallel to research different aspects of the codebase efficiently.
Agent Selection
Choose agents based on research needs:
Always start with locators:
codebase-locator - Find WHERE files/components are
thoughts-locator - Find WHAT documentation exists
Then use analyzers on findings:
codebase-analyzer - Understand HOW specific code works
codebase-pattern-finder - Find SIMILAR implementations
thoughts-analyzer - Extract insights from specific docs
Only if user explicitly requests:
web-search-researcher - External research
Parallel Execution Pattern
Spawn multiple agents at once using multiple Task calls in a single message:
Task(subagent_type="codebase-locator", prompt="Find all authentication files")
Task(subagent_type="thoughts-locator", prompt="Find auth-related documentation")
Then wait for both to complete before spawning follow-up agents.
Research Workflow
- Decompose: Break research question into specific areas
- Locate: Use locator agents to find relevant files/docs
- Analyze: Use analyzer agents on most promising findings
- Synthesize: Combine all results with file:line references
Example
Research question: "How does authentication work?"
Step 1 - Locate (parallel):
Task(subagent_type="codebase-locator", prompt="Find all authentication-related files including handlers, middleware, and services")
Task(subagent_type="thoughts-locator", prompt="Find any documentation about authentication implementation")
Step 2 - Analyze (parallel, after step 1 completes):
Task(subagent_type="codebase-analyzer", prompt="Analyze authentication flow from login endpoint through JWT generation")
Task(subagent_type="codebase-pattern-finder", prompt="Find examples of how other endpoints use authentication middleware")
Step 3 - Synthesize:
Combine findings with specific file paths and line numbers.