hugging-face-community-evals

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

• Third-party content exposure detected (high risk: 0.90). The skill's scripts explicitly fetch and run models from the public Hugging Face Hub (e.g., model IDs like "meta-llama/..." and the "hf-inference-providers/{args.model}" usage in scripts/inspect_eval_uv.py) and include a --trust-remote-code option in scripts/inspect_vllm_uv.py and lighteval_vllm_uv.py, which allows executing untrusted, user-provided repository code from the Hub that can materially change runtime behavior.

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

• Potentially malicious external URL detected (high risk: 0.80). The scripts fetch models from the Hugging Face Hub at runtime (e.g., https://huggingface.co/microsoft/phi-2) and expose a --trust-remote-code option which allows executing code from those remote model repositories, so remote Hugging Face model repos can execute arbitrary code when used.