quantizing-models-bitsandbytes

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (high risk: 0.80). The skill explicitly loads public datasets and models from third-party hubs (e.g., references/qlora-training.md calls datasets.load_dataset("timdettmers/openassistant-guanaco") from the Hugging Face hub and the docs link to GitHub/HuggingFace), which ingests user-generated/untrusted content that the agent is expected to read and use for training.