@timo2026

adversarial-engine orchestrates four llm personas (architect, engineer, security officer, arbitrator) in iterative debate loops with code sandbox validation and vector-enhanced retrieval. convergence detection prevents infinite cycles.

多模型对抗引擎 - 四模型真实对抗辩论系统。架构师+工程师+安全官+仲裁者协作，代码沙箱验证，向量检索增强，收敛判断自动熔断。