dispatching-parallel-agents

Parallel agents are dispatched by launching separate, self-contained prompts for each independent problem domain. This enforces parallel execution with isolated context, preventing cross-contamination of session history across tasks.

Parallel agent coordination is used to delegate independent problem domains to specialized agents. It speeds up debugging by running multiple isolated investigations simultaneously without leaking shared session state between tasks.

You should use parallel agents when there are 3 or more test failures, or when you have independent tasks that can be solved in isolation without shared state. It enforces self-contained prompts and isolated context to avoid cross-task dependencies.

Isolated context prevents cross-task dependencies by ensuring each agent operates within its own self-contained prompt without access to shared session history. This safe isolation stops state leakage and cross-contamination during parallel execution.

No, parallel task dispatch is designed specifically for independent tasks that can be solved in isolation without shared state. If your tasks require shared state, parallel execution with isolated contexts will fail due to cross-task dependencies.