Vibe Coding Discovergarak
View on GitHubthe LLM vulnerability scanner
A command-line LLM vulnerability scanner that runs static, dynamic, and adaptive probes against hosted or local models. Tests for issues such as prompt injection, jailbreaks, data leakage, hallucinations, and toxic outputs, then reports detected failures.
Use Cases
Probe LLMs for prompt injection vulnerabilitiesTest resistance to jailbreaksAssess potential data leakageDetect hallucinations and misinformationEvaluate toxicity generationCompare model vulnerability ratesRed-team hosted or locally run language models
Built With
- Language
- Python
- Frameworks
- Hugging Face Transformers · LiteLLM · llama.cpp · AWS Bedrock · OpenAI API · Replicate
Tags
LLM security · red teaming · vulnerability scanner · model evaluation · adversarial testing · prompt injection · jailbreak detection · hallucination · data leakage · toxicity