Open source security toolkit that scans LLM prompts and outputs for prompt injection, PII leakage, and toxic content.