dispatching-parallel-agents

Run parallel agents by dispatching one agent per independent failure domain, allowing concurrent investigation of subsystem failures without shared state. This enables parallel execution where each domain is addressed independently and results are integrated into a summary.

Use parallel agents for triaging independent tasks when you have 2 or more failures across tests or subsystems that can be addressed without context from others. Dispatching parallel agents prevents sequential investigation bottlenecks and saves time.

Yes, you can dispatch parallel agents for multi-file or subsystem failures with no interdependencies. Assigning one agent per isolated domain minimizes cross-domain context interference and enables rapid concurrent triage.

The best way to coordinate parallel investigations is dispatching parallel agents, which requires a mechanism to assign one agent per independent failure domain. This enables parallel execution with summary integration to consolidate findings.

Parallel agents are not suitable for tasks with sequential dependencies or shared state because the mechanism requires each failure domain to be addressable without context from others. Interdependent subsystems need sequential investigation instead.