A new framework called Agent-BOM aims to enhance security auditing for LLM-based agentic systems by providing a unified graph representation that captures both static capabilities and dynamic states. This innovation addresses critical gaps in existing auditing methods, potentially improving security measures in complex autonomous environments.