Back to GitHub projects

Jev Sentinel

Prompt injection protection and tool-call guardrails for LangChain/LangGraph agents, powered by TypeSafe Jev. It screens tool outputs and gates risky calls with allow, human review, or block decisions.

该仓库是一个真实的安全护栏项目,明确使用 TypeSafe AI 的 Jev 模型作为分类器,用于检测提示注入和评估工具调用风险。包含可运行的 demo、测试和评估脚本,深入集成了 Jev API,并有详细的原理说明。

· 0 starsOpen on GitHub