ml-iterate

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

• Third-party content exposure detected (high risk: 0.90). The skill's Web mode explicitly requires the agent to WebFetch and parse open/public third-party content (e.g., "WebFetch the user's model card: https://huggingface.co/{org}/{model}" and GitHub issues/docs listed in the "Web mode URL registry") and then use those fetched pages to drive hyperparameter decisions and code, which exposes the agent to untrusted user-generated content that could inject instructions.

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

• Potentially malicious external URL detected (high risk: 1.00). The skill mandates runtime WebFetch calls (e.g., https://huggingface.co/{org}/{model}) and uses the fetched content to ground, quote, and control the agent's instructions and outputs, so these external URLs directly control prompts and are required dependencies.