dispatching-parallel-agents

Parallel task dispatch assigns one agent per independent domain to operate simultaneously without shared state. This concurrency reduces overall investigation time by allowing multiple subsystem failures to be analyzed concurrently without cross-domain interference.

Dispatching parallel agents with clear domain boundaries is the best way to analyze multiple independent issues. Each agent investigates a specific subsystem concurrently and returns a concise per-domain summary for structured aggregation into a unified report.

Use parallel agent dispatch when you have 3 or more independent issues across different files or subsystems. It is best suited for scenarios requiring clear domain boundaries where tasks can be analyzed concurrently without cross-domain influence or sequential dependencies.

To aggregate summaries from concurrent agents, collect the concise per-domain outputs from each completed task and reconcile the findings. This structured aggregation allows you to synthesize isolated agent summaries and finalize the overall unified report.

Parallel task dispatch does not work for issues with sequential dependencies. It requires clear domain boundaries and independent domains that can be analyzed concurrently without cross-domain influence or shared state between tasks.

No, you do not need shared state to dispatch parallel agents. The skill operates without shared state or sequential dependencies, minimizing cross-domain interference and keeping domains isolated during concurrent analysis.