TypeScript library resk-llm-ts offers pipeline-based defense against LLM prompt injection

Prompt injection ranks as the top security risk for large language model applications, as attackers use natural language to manipulate AI systems in ways traditional tools like web application firewalls cannot detect. A TypeScript library called resk-llm-ts provides a zero-dependency SecurityPipeline featuring 11 detectors designed to identify direct injection attempts, jailbreaks, memory poisoning, and content framing attacks. Developers can integrate the library into an Express.js application by adding it as middleware or calling the pipeline manually within request handlers, enabling threats to be blocked before reaching the LLM. The tool flags suspicious input with a blocked boolean and exposes threat metadata such as severity and detector name to aid debugging. The library's authors caution that it is pattern-based and not foolproof, noting that novel attacks may slip through and legitimate inputs could occasionally trigger false positives.