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

1. Decompose: Break research question into specific areas
2. Locate: Use locator agents to find relevant files/docs
3. Analyze: Use analyzer agents on most promising findings
4. 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.