huggingface-local-models

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

• Third-party content exposure detected (low risk: 0.10). The workflow’s required runtime path reads first-party provider metadata/pages from Hugging Face (search results, the ?local-app=llama.cpp repo page, and the /api/models/<repo>/tree/main?recursive=true JSON) rather than any outsider-authored free text submitted into the skill’s sources without first selecting a specific HF repo/file.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill explicitly instructs cloning and building remote code from https://github.com/ggml-org/llama.cpp at runtime (git clone + make) and to fetch and parse Hugging Face pages (https://huggingface.co/?local-app=llama.cpp and https://huggingface.co/api/models//tree/main?recursive=true) whose page/tree content is treated as the authoritative snippet to construct launch commands, so these runtime fetches can execute remote code or directly control agent instructions.