I built this because prompt injection is one of those risks everyone talks about but few people actually have tooling for. Gatekeeper runs entirely in your browser. Nothing gets sent anywhere. It matches against a rule set I put together from OWASP's LLM Top 10, published jailbreak research, and patterns I've studied in my AI security coursework.