Explore advanced security strategies for AI agent workflows in cloud-native environments through this 25-minute conference talk. Learn to move beyond hard-coded credentials by implementing trusted, granular identities for agents acting on behalf of users. Discover how to establish traceability, enforce least privilege principles, and enable auditable decision-making within zero-trust architectures. Examine shared agents and tool-calling patterns while understanding how SPIRE's workload identity integrates with user identity to support secure delegation and dynamic, context-aware authorization. Master techniques for safeguarding agent interactions with external tools and data sources through identity propagation and policy enforcement. Gain practical insights through a real-world case study utilizing Llama Stack and the extended Model Context Protocol (MCP) to build secure, identity-aware agentic platforms ready for production deployment.