gptq

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

• Potentially malicious external URL detected (high risk: 0.90). The skill loads models from the HuggingFace Hub at runtime (e.g., identifiers like "TheBloke/Llama-2-7B-Chat-GPTQ" and URLs such as https://huggingface.co/models?library=gptq), which causes the agent to download external model files during execution that directly determine the model's behavior and are required for the skill to run.