model-download-user

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

• Third-party content exposure detected (medium risk: 0.30). Outsider-authored free text is accepted into the workflow via the REST API request body to POST http://localhost:8200/api/v1/models/download (model name, hub, type, config, etc.), which the service then uses to fetch/convert models from external hubs.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill explicitly fetches and executes remote code at runtime: it instructs cloning and then running scripts from https://github.com/open-edge-platform/edge-ai-libraries.git and downloads+sources the shell script https://raw.githubusercontent.com/open-edge-platform/edge-ai-libraries/main/microservices/model-download/scripts/get_model.sh, so external content is fetched and executed.