Enforces a structured protocol for detecting and responding to prompt injection attempts in LLM applications. Use when building an input-sensitive app, auditing for security vulnerabilities, or operating a model that interacts with user-generated content.
Idioma do texto original: inglês