modal

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

• Third-party content exposure detected (high risk: 0.80). The skill's workflow explicitly downloads and runs public model artifacts (e.g., AutoModel.from_pretrained in the "Download model at build time" example), which pulls untrusted, user-hosted content from the open model hub (Hugging Face) that the agent will load/interpret and which can materially affect runtime behavior.