deepstream-import-vision-model

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

• Third-party content exposure detected (medium risk: 0.30). The skill’s runtime workflow ingests outsider-authored free text from HuggingFace model assets by calling the HF tree API and downloading config.json/labels during Step 1–3 (via scripts/model/hf-list-files.sh and scripts/model/hf-download-config.sh, then parsing config.json content in model-acquire.md).