orchestrator

You start multi-agent orchestration by routing an intent through route-intent.mjs. This triggers the Lead Agent to plan tasks, delegate to Worker Agents, and manage parallel execution across long-running operations.

Multi-agent orchestration handles long-running operations through session recovery and archiving. This allows the Lead Agent to resume interrupted tasks, maintain secure isolation, and aggregate results reliably across extended execution periods.

Yes, you can automate API key rotation during parallel subtask execution. The orchestration framework supports rotating API keys natively to distribute load and prevent rate limits when multiple Worker Agents operate concurrently.

Task decomposition in a lead-worker architecture is the process where a Lead Agent breaks down a complex intent into smaller subtasks. These subtasks are distributed to specialized Worker Agents for parallel execution and traceable result aggregation.

Yes, multi-agent orchestration works without external dependencies. The skill operates entirely using internal scripts to manage task planning, routing, session recovery, and secure isolation without requiring additional packages or libraries.

You should not use multi-agent orchestration for simple, single-step tasks that do not require parallel execution or complex planning. Overhead from Lead-Worker coordination and session management makes it unsuitable for lightweight, isolated operations.