parallel-agent-dispatch

Dispatch one agent per independent problem domain to investigate failures concurrently. This preserves your main session context for coordination while agents work concurrently with isolated context on specific scopes without shared state.

Orchestrating parallel agents requires grouping failures by domain and creating focused tasks with strict scoping and clear goals. You construct exact instructions and isolated context for each agent, then synthesize their findings to verify no conflicts exist.

Use parallel agent dispatch when you have 3+ unrelated failures across different files or subsystems that can be investigated concurrently without shared state, avoiding the wasted time of sequential investigation for independent bugs.

Parallel agent dispatch requires isolated context rather than shared session history. Agents should never inherit your session context; you construct exactly what they need to stay focused and succeed at their specific task scoping.

The best way to scope independent tasks is enforcing clear prompts, strict task scoping, and isolated context per agent. Assign specific scope and constraints for each domain, then review and integrate outputs summarizing findings and recommended fixes.

Dispatching parallel agents is limited to independent problems without shared state. If failures are related or require shared context across subsystems, parallel investigation risks conflicts and sequential debugging becomes necessary.