What problem does it solve?
It solves the problem of deciding when and how to delegate parts of a task to subagents, selecting an orchestration topology that improves outcomes, and supervising parallel or staged execution without wasting coordination effort.
Core Features & Use Cases
- Dynamic decomposition decisions: Determines whether extra children beyond a mandatory task-owning child are warranted based on task shape and coordination tradeoffs.
- Topology selection & supervision: Classifies task patterns (e.g., generator→selector, solver→verifier, map→reduce) and selects a fitting subagent orchestration graph.
- Clear contracts & integration rules: Enforces distinct responsibilities, non-overlapping success conditions, staged barriers, and parent narration boundaries so children own substantive workspace actions.
- Adaptation to new evidence: Re-plans the topology when new evidence suggests the current split is wasteful or missing necessary roles.
- Use Case: When building a complex research-to-delivery workflow, it helps a runtime launch specialists (e.g., evidence collectors and verifiers), run independent branches in parallel, and then integrate bounded returns into a single coherent final result.
Quick Start
Load this module when your agent runtime needs to decide on additional subagents and an orchestration plan during an ongoing task, and then use its policy to guide how child stages are launched, waited on, and integrated.
