Your AI Agents Are Lying to Each Other — And Your Architecture Is Why

Multi-agent AI systems can look coherent while failing underneath.