dispatching-parallel-agents

Dispatch parallel agents to isolate and solve independent failures efficiently. Apply this approach when 3+ test files or subsystems fail independently with no shared state and parallel investigation is beneficial.

Dispatch parallel agents when 3+ test files or subsystems fail independently with no shared state. This approach enables faster remediation of multiple failures compared to sequential investigation.

Prevent context leakage by constructing exactly what each agent needs instead of letting them inherit your session history. Provide clear constraints and isolated contexts so agents stay focused on their specific independent failure.

Coordinate independent subsystem failures by dispatching one agent per independent problem domain. Let agents work concurrently while preserving your own context for coordination work rather than individual investigations.

No, you should not have shared state when diagnosing multiple failing test files with parallel agents. This technique requires independent failures with no shared state so each agent can investigate in isolation without cross-talk.