Hey Product Hunt! 👋 Instead of relying on clever prompts (which can be bypassed), SafeBrowse enforces a hard security boundary between untrusted content and your LLMs. Key features: • 50+ prompt injection patterns detected • Policy engine for login/payment forms • RAG sanitization for vector DBs • Python SDK with sync/async support • Fail-closed: security cannot be bypassed We'd love your feedback! What security challenges are you facing with AI agents?