huggingface-local-models

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

• Third-party content exposure detected (medium risk: 0.30). The required workflow reads Hugging Face model pages and the models tree API (e.g., https://huggingface.co/<repo>?local-app=llama.cpp and https://huggingface.co/api/models/<repo>/tree/main?recursive=true) to extract quant/file names, and those pages/tree contents can be influenced by outsiders publishing/authoring repos.

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 0.90). The skill explicitly instructs fetching and parsing Hugging Face pages (e.g., https://huggingface.co/?local-app=llama.cpp and the tree API https://huggingface.co/api/models//tree/main?recursive=true) at runtime to choose quants and build launch commands, and also includes a git clone of https://github.com/ggml-org/llama.cpp followed by make, so remote content both controls agent behavior and results in executing fetched code.