Mitigate prompt injection risks in LLM-based systems. Use when designing, building, or reviewing AI systems that accept user prompts, or when evaluating model safety for deployment.
