Standalone Ruby library for completed AI input and output safety. Detects and redacts PII, identifies prompt injection attacks (including multi-turn patterns), scans for leaked secrets, enforces reusable custom restricted-content policies, and provides explicit Rails boundary adapters and sanitized application-defined notifications — runs entirely in-process with one lightweight bundled-gem dependency. Ships with a provider-agnostic llm_provider: hook for LLM-backed semantic evaluation.