benchmark-model-kernels

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

• Third-party content exposure detected (medium risk: 0.30). Outsider-authored text from a user-supplied model reference (Hub ID/local directory/config.json) is ingested at runtime by scripts/benchmark_model.py via AutoConfig.from_pretrained(...) and AutoModelForCausalLM.from_config(...), which reads provider metadata and (optionally) remote code to construct a Transformers module for shape derivation.