modelscope-cli

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

• Third-party content exposure detected (high risk: 0.80). This skill's required workflows (e.g., SKILL.md and scripts/run_ms_model_download.sh, scripts/run_ms_datasets_download.sh, and reference/WIKI.md) explicitly download models/datasets from the public ModelScope hub (https://modelscope.cn) and run tools (modelscope download / snapshot_download, curl/openssl checks) that ingest community-hosted, user-contributed content which the scripts then read/interpret (file names, sizes, metadata) to drive decisions (downloads, parameter estimation, diagnostics), exposing the agent to untrusted third-party content.

MEDIUM W013: Attempt to modify system services in skill instructions.

• Attempt to modify system services in skill instructions detected (high risk: 1.00). The skill explicitly instructs installing a CA certificate into system trust stores using sudo (e.g., sudo cp ... into /etc/... and update-ca-certificates) and offers disabling SSL verification, which requires elevated privileges and modifies system security files.