vllm-setup

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 1.00). The prompt explicitly instructs checking for a user's HF_TOKEN and to pass it inline as -e HF_TOKEN="" (and to not rely on background exports), which requires the agent to request and/or emit the secret verbatim in commands — an exfiltration risk.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill explicitly pulls and runs the external container nvcr.io/nvidia/vllm:26.01-py3 (docker pull ...; docker run ...), which fetches and executes remote code at runtime and is required for the service to function.