rtvi-cv-customize-model

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

• Potentially malicious external URL detected (high risk: 0.90). The skill instructs cloning and using remote deployment code (git clone https://github.com/NVIDIA-AI-Blueprints/video-search-and-summarization.git) which fetches scripts/Dockerfiles (e.g., dev-profile.sh, ds-start.sh) that are then built/executed at runtime (and also references pulling the base image nvcr.io/nvidia/vss-core/vss-rt-cv:3.2.1), so these external resources directly supply and execute required runtime code.

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

• Attempt to modify system services in skill instructions detected (high risk: 0.90). This skill instructs editing runtime startup scripts and compose mounts and explicitly runs sudo mkdir/chown to modify host-mounted model directories and remount profile scripts, which changes the host system state and requires elevated privileges.