Launch multiple sub-agents in parallel with intelligent model selection, Zero-shot CoT reasoning, independence verification, and batch operation scheduling

Single-task execution with LLM-as-Judge verification in an iterative loop, supporting auto-retry and strict orchestrator role separation