China Telecom Multi-agents systems (MAS) increasingly cooperate through workflow, orchestrated, and mesh communication patterns. While existing Internet protocols provide confidentiality and endpoint authentication, they were not designed for agent-native semantics such as dynamic identity, computation-bounded requests, context integrity, and intermediary trust. This document analyzes security risks in agent communication and identifies limitations in widely deployed protocols (TLS, HTTP, MQTT,A2A and etc.) and developer framework (AutoGen and etc.). This document intend to establish a common problem statement and gap analysis to inform future IETF standardization.