dispatching-parallel-agents

Dispatch one agent per independent problem domain using the Task tool, giving each a focused prompt with specific scope, constraints, and expected output. All agents run concurrently, then you review summaries and run the full test suite to verify integration.

Use parallel agents when failures are independent, such as different test files or subsystems with different root causes and no shared state. Avoid parallel dispatch when failures are related, require full system context, or agents would edit the same files.

A good agent prompt is focused on one problem domain, self-contained with all needed context like error messages and test names, and specific about constraints and output. Include what not to change and request a summary of root cause and fixes.

Yes, agents can conflict if they edit the same files or share state. Prevent this by assigning disjoint scopes per agent, and after completion check for overlapping edits, review each summary, and run the full test suite to catch integration issues.

Parallel dispatch does not fit related failures where one fix resolves others, exploratory debugging where the problem is unknown, or investigations needing full system context. Agents can also make systematic errors, so spot-check their work.