dispatching-parallel-agents

To run parallel debugging tasks, dispatch a dedicated agent per independent problem domain to investigate concurrently. Each agent operates in isolation without shared history, preventing context leakage and ensuring rapid progress across unrelated subsystems.

Parallel agent orchestration coordinates multiple independent tasks by assigning a dedicated agent per problem domain. Use this approach when failures or work items are separate, non-interdependent, and can run concurrently without needing shared state or context.

Prevent context leakage by enforcing a self-contained prompt per agent with a specific scope, clear goal, and constraints against cross-domain interference. This isolation ensures agents run concurrently without sharing history or state.

Parallel agent dispatch does not work for debugging interdependent subsystem failures because it is designed for separate, non-interdependent work items. If failures require shared state or cross-domain history, agents cannot operate in true isolation.

Integrate findings after dispatching parallel agents by collecting each agent's specific output, validating consistency across results, and synthesizing actions. This ensures combined progress is achieved without collateral conflicts between domains.