transformers-js

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 directly as the model input (e.g., user-entered prompt/text passed to extractor(text, ...) or generator(prompt, ...) in the provided Express/Browser/React examples), with no step that selects a specific trusted item first.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill includes runtime browser module imports from the jsDelivr CDN that fetch and execute remote JavaScript (e.g., https://cdn.jsdelivr.net/npm/@huggingface/transformers and https://cdn.jsdelivr.net/npm/@huggingface/transformers@4) and an ONNX WASM path (https://cdn.jsdelivr.net/npm/onnxruntime-web/dist/) which are fetched at runtime and execute remote code.