dispatching-parallel-agents

You can debug multiple unrelated test failures by dispatching parallel agents, assigning one focused agent per independent failure to investigate concurrently while preserving coordinator context and avoiding cross-contamination of state.

Orchestrate parallel agents by providing each with self-contained prompts, isolated context, and clear success criteria, then consolidating their summaries, checking for conflicts, and running full verification before integrating returned fixes safely.

Yes, you can use parallel agents to fix separate subsystems when the problem domains can be analyzed independently without shared runtime state, allowing concurrent investigation and faster resolution than sequential debugging.

Integrate concurrent agent fixes by reviewing their summaries, checking for conflicts across the codebase, and running full verification to validate integration safely before applying the returned fixes.

Avoid parallelization for debugging when problems are not independent, require shared runtime state, or lack clear success criteria, as these conditions prevent isolated agent context and safe concurrent investigation.