superagent

Why we included this project

LLM-backed apps face attacks that ordinary security tooling does not catch, and Superagent is built for exactly that gap. It is a lightweight SDK you embed directly in your application: it screens runtime calls for prompt injections and unsafe tool use, redacts PII and secrets before they reach the model, and scans repositories for poisoned instructions aimed at AI agents. A red-teaming mode lets you probe your own assistant for weak points before someone else does. With TypeScript and Python bindings and an open-weight guard model you can self-host for low latency, it is a practical safety layer for teams that do not want to bolt on a heavy governance platform.