AI security review

Use when reviewing LLM, agent, RAG, tool-calling, memory, or AI-assisted workflows for prompt injection, unsafe agency, sensitive-data exposure, insecure output handling, or unbounded resource use.