Vigil: LLM Prompt Injection Detection and Defense Toolkit

Abstract

Open-source scanner for detecting prompt injections using vector similarity, YARA rules, text classifiers, and canary tokens.

Categories

Framework mappings

@misc{deadbits2024vigil,
  title = {{Vigil: LLM Prompt Injection Detection and Defense Toolkit}},
  author = {DeadBits},
  year = {2024},
  url = {https://github.com/deadbits/vigil-llm}
}