dispatching-parallel-agents

Group failures by independent problem domain, then dispatch one agent per domain with a focused prompt containing the failing test names, error messages, and constraints. Run the agents concurrently and integrate their fixes after reviewing each summary.

A good agent prompt is focused on one problem domain, self-contained with all needed context such as error messages and test names, explicit about constraints like not changing unrelated code, and specific about the expected output such as a summary of root cause and fixes.

Avoid parallel dispatch when failures are related and fixing one might fix others, when understanding requires full system context, during exploratory debugging where the cause is unknown, or when agents would share state such as editing the same files.

Read each agent's summary to understand what changed, check whether agents edited the same code, run the full test suite to confirm all fixes work together, and spot-check results since agents can make systematic errors.

Isolated context keeps each agent focused on its narrow scope without distraction from unrelated history, and preserves the coordinator's own context for integration work. You construct exactly the instructions and context each agent needs.