tao-finetune-huggingface-model

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

• Third-party content exposure detected (high risk: 0.85). Step 3/4 runtime ingests outsider-authored free text from live-fetched HuggingFace model cards / repo example scripts / task docs (public web content) into the agent’s LLM context to derive the training recipe.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill explicitly fetches HuggingFace model/dataset repos at runtime (e.g., AutoConfig.from_pretrained(model_id, trust_remote_code=True) and load_dataset(..., trust_remote_code=True)) which can load and execute remote repo code from the Hugging Face Hub (https://huggingface.co/<model_id>), so this is a runtime external dependency capable of executing remote code.