gptq

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

• Third-party content exposure detected (high risk: 0.90). This skill explicitly loads and uses public, user-contributed models and datasets (e.g., AutoGPTQForCausalLM.from_quantized("TheBloke/...") and load_dataset("c4") / "anon8231489123/ShareGPT_Vicuna_unfiltered"), so the agent ingests untrusted third-party content (models/texts) as part of calibration and inference, enabling potential indirect prompt injection.