dispatching-parallel-agents

To fix multiple failing tests in parallel, you dispatch separate agents to address each independent failure. This approach requires defining distinct agent prompts with clear scope, constraints, and expected outputs for each unrelated test.

Parallel agent orchestration is a technique to solve multiple independent tasks or failures simultaneously without shared state. It involves dispatching focused agents to handle separate subsystem bugs concurrently and collecting their summaries for integration.

You should dispatch separate agents when multiple unrelated problems can be solved without shared state, such as fixing several failing tests or addressing distinct code issues across separate files. Ensure each task has independent domains before parallel execution.

After running parallel agents, you integrate fixes by reviewing each agent's summary to verify no conflicts. You then run the full test suite and merge the resolved code issues to ensure the independent changes work together correctly.

The limitation of dispatching parallel agents is that it requires tasks to be completely independent without shared state. If failures are related or affect the same files, concurrent agents may produce conflicting fixes that require extensive manual review.