orchestrator

Multi-domain planning is managed by decomposing a request into subtasks, dispatching specialist agents in parallel, and synthesizing their outputs into a unified result. A shared findings scratchpad handles sequential dependencies across domains.

Parallel research and chained execution across engineering and UX roles is achieved by orchestrating multi-role work, running independent tasks simultaneously, and chaining dependent tasks via a shared findings scratchpad to merge outputs.

Yes, you can use agent dispatch when you do not know which specialist to use. The orchestrator plans the work, delegates subtasks to the right experts across product, engineering, data, UX, QA, and GTM, and merges their outputs.

Yes, multi-role orchestration requires a Phase 0 clarifying-question gate and an announced Phase 3 work plan. It requests confirmation before dispatching actions that write files, run commands, or access external resources to ensure safety and control.

Confidence signaling works by requiring CONFIDENCE-signal handling during the orchestration process. This mechanism ensures that synthesized results from sequential task execution maintain reliability across the shared findings scratchpad.

You should not use a multi-role orchestration approach for simple, single-domain tasks that lack sequential dependencies or parallel research requirements, as the overhead of planning subtasks and synthesizing specialist agents would be unnecessary.