What problem does it solve?
Deciding when and how to split coding work across multiple agents is error-prone: uncontrolled fan-out wastes tokens, causes file conflicts, and leaves results unsynthesized. This Skill gives the main agent a disciplined strategy for delegating only genuinely independent work and merging results correctly.
Core Features & Use Cases
- Fan-Out Decision Rules: Clear criteria for when parallel sub-agents pay off versus when a single agent should keep the work, avoiding delegation by default.
- Child Contracts & Isolation: Each child gets a bounded objective, allowed/forbidden actions, file ownership or read-only isolation, and a required structured output schema.
- Progressive Control & Synthesis: Consume partial results with waitAny or checkpoints, steer or cancel children, then resolve conflicts from evidence and run parent-level verification.
- Use Case: Ask the agent to investigate six independent modules in parallel; it fans out read-only research children, consumes results as they arrive, and synthesizes one coherent answer instead of waiting on the slowest reviewer.
Quick Start
Ask the agent to investigate these independent modules in parallel using bounded sub-agents and synthesize the findings into one answer.
