Council
Multi-agent exploration workflow for complex codebase analysis.
Workflow
- •
Initial exploration: Dig around the codebase for the given area of interest. Gather keywords, architecture overview, and general context.
- •
Spawn task agents: Launch n=10 parallel task agents (unless specified otherwise) to explore deeper. Some agents should take unconventional approaches for variance.
- •
Synthesize results: Once agents complete, use gathered information to fulfill the user's request.
Usage
If the user is in plan mode, use the collected information to create the plan.
Example invocation:
code
Based on [area of interest], spawn 10 agents to explore: - Core implementation patterns - Edge cases and error handling - Integration points - Test coverage - Documentation gaps - Performance characteristics - Security considerations - Alternative approaches - Historical context (git history) - Related systems