dispatching-parallel-agents

Dispatch parallel agents to investigate independent failure domains by grouping failures, creating focused agent tasks with defined scope and constraints, running them concurrently, and consolidating summaries for review without cross-domain interference.

The best way to run parallel agents for concurrent debugging is to identify distinct independent domains, create focused agent tasks with defined scope and expected output per domain, dispatch agents concurrently, and review summaries to consolidate fixes without cross-domain interference.

Define scope and constraints by identifying independent failure domains, then creating focused agent tasks for each domain that specify a clear scope, goal, constraints, and expected output before dispatching agents in parallel.

No, you should not dispatch parallel agents for subsystems that share state. This approach covers multiple simultaneous failures or subsystems that can be analyzed without shared state, ensuring fixes consolidate without cross-domain interference.

Consolidate results by reading summaries from all dispatched parallel agents, reviewing their findings, and integrating fixes without cross-domain interference after all concurrent domain-specific investigations complete.