dispatching-parallel-agents

Debug multiple failing test files concurrently by dispatching independent agents to each isolated problem domain. This parallel investigation approach works when failures have no shared state, allowing agents to proceed concurrently without sequential dependencies.

Avoid parallel debugging agents when failures are related, require full-system context, or carry shared-state interference risks. Parallel dispatch is only effective for independent failures across separate files or subsystems where concurrent progress is possible without sequential dependencies.

Integrate parallel debugging agent results by collecting their outputs, checking for overlap or conflicts, and verifying through a full test-suite run before merging. This conflict-aware integration ensures independent fixes do not interfere with each other.

Scope prompts for parallel debugging agents by giving each agent only the context needed for its specific failure set. Self-contained instructions with constrained outputs ensure agents stay within their single problem domain during concurrent investigation.

Use parallel agents for root cause analysis across different subsystems when those subsystems are broken independently. Dispatch one agent per isolated problem domain to investigate failures concurrently, provided there is no shared state between the investigations.