huggingface

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 0.90). The prompt shows and encourages embedding Hugging Face tokens directly in environment variables and client constructors (e.g., export HF_TOKEN="hf_xxxxxxxxxx", InferenceClient(token="hf_xxxxxx")), which is an insecure pattern that would require the LLM to handle or echo secret values verbatim.

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

• Third-party content exposure detected (high risk: 0.90). The SKILL.md explicitly instructs fetching models and datasets from the public Hugging Face Hub (e.g., AutoModel/AutoTokenizer.from_pretrained, load_dataset("imdb"), and InferenceClient usages), which downloads arbitrary user-published models/datasets whose untrusted content can influence model behavior and subsequent agent actions.