amc-setup-calibration-stack

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 1.00). The skill instructs the agent to ask for NGC and HuggingFace tokens (with a chat fallback) and then embed those secrets directly into shell commands (e.g., echo "<NGC_API_KEY>" | docker login ... and HF_TOKEN="<HF_TOKEN>" "$HF_BIN" ...), which requires the LLM to handle/output secret values verbatim.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill performs runtime operations that fetch and execute remote artifacts — notably contacting nvcr.io to pull and run NGC container images (docker login nvcr.io) and referencing the GitHub repo (https://github.com/NVIDIA-AI-IOT/auto-magic-calib.git) which may be cloned and used to run docker-compose — so these external endpoints are runtime dependencies that can result in executing remote code.

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 privileged state changes (sudo usermod to add the user to the docker group and sudo chown -R 1000:1000 on project/model directories, plus suggesting sudo apt install), which ask the agent (or user acting on the agent's instructions) to modify system ownership and user/group membership — actions that alter the host machine's security state.