Prompt injection and jailbreaks exploit LLMs by embedding malicious instructions in user inputs or retrieved content. This playbook outlines real-world cases and practical defenses including sanitization, least-privilege design, and red-team testing.