dispatching-parallel-agents

To coordinate parallel debugging, you dispatch specialized agents to work concurrently on independent problem domains. This isolates context per agent, preventing cross-task interference while preserving your main session for orchestration.

Multi-agent orchestration is the coordination of specialized agents working concurrently. You should use parallel agents when you have 3 or more unrelated failures or subsystems where each issue can be analyzed without knowledge of the others.

You dispatch specialized agents by assigning one agent per independent problem domain. The system enforces this isolation to maximize parallelism, using clear self-contained prompts to maintain focus and reduce session history leakage.

Parallel agents are best suited for subsystems without shared state. If your subsystems share dependencies or require knowledge of one another to be analyzed, this approach is not recommended due to enforced context isolation.

The best way to prevent context leakage during independent investigations is to enforce one agent per domain with isolated contexts. Clear, self-contained agent prompts ensure each task maintains focus without leaking session history.

Dispatching parallel agents preserves your main session context because the specialized agents operate in isolation. By delegating independent tasks to these agents, your primary context remains free for high-level coordination work.