transformers

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

• Third-party content exposure detected (high risk: 0.80). The skill loads models/tokenizers from the public Hugging Face Hub (e.g., AutoTokenizer.from_pretrained / AutoModelForCausalLM.from_pretrained and pipeline("model-id")), which pulls user-contributed, untrusted third-party artifacts that the agent will read/execute as part of its workflow.