This chapter provides a practical framework for managing security throughout the lifecycle of autonomous, Agentic AI systems. It introduces a detailed 10-stage security lifecycle, beginning with secure agent design and intent specification, progressing through model integrity, tool access control, memory security, and identity management, and concluding with secure decommissioning and incident response. For each stage, specific risks—such as goal misalignment, prompt injection, tool poisoning, and behavioral drift—are analyzed, and corresponding mitigation strategies are presented. Beyond technical controls, the chapter emphasizes organizational governance by presenting a RACI (Responsible, Accountable, Consulted, Informed) matrix to assign clear roles and responsibilities. Furthermore, it outlines a practical approach for extending existing DevSecOps and MLSecOps pipelines to accommodate Agentic AI and provides a CISO-level checklist for strategic oversight. Ultimately, the chapter argues that securing Agentic AI is a continuous, multi-layered discipline that requires embedding security-by-design principles into every phase of development, operation, and governance.