ai-for-science-generator

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 loading models from Hugging Face (e.g., using AutoModelForCausalLM.from_pretrained with trust_remote_code=True and setting HF_ENDPOINT to mirror endpoints) which fetches and executes user-provided model/code from public third-party sources, allowing untrusted content to influence runtime behavior.