ptq

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

• Third-party content exposure detected (high risk: 0.90). The skill's required workflow (references/unsupported-models.md Step A and related steps) explicitly instructs downloading and inspecting public Hugging Face model checkpoints and their modeling_*.py files (and even cloning transformers from GitHub) and to use trust_remote_code, meaning arbitrary third‑party model READMEs and source files from the open web are fetched and interpreted to determine dependencies and runtime actions.

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

• Potentially malicious external URL detected (high risk: 0.80). The skill includes runtime instructions to clone and install code from https://github.com/huggingface/transformers.git (git clone → pip install) and to pull/run a remote container image nvcr.io/nvidia/tensorrt-llm/release:, which fetch and execute remote code that can directly affect the agent's runtime behavior.