physical-ai-infrastructure-setup-and-resilient-scaling
Pass
Audited by Gen Agent Trust Hub on Jun 12, 2026
Risk Level: SAFE
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill fetches infrastructure tools, GPG keys, and Helm charts from trusted organizations and well-known services including HashiCorp's official repository, NVIDIA's NGC Helm registry, and Microsoft's GitHub repositories. It also facilitates downloading model weights from Hugging Face.
- [COMMAND_EXECUTION]: Provisioning involves the execution of standard infrastructure management tools such as Terraform, Helm, Kubectl, and Azure CLI. Root privileges ("sudo") are utilized for legitimate system-level tasks like package installation (Terraform) and managing local Kubernetes environments (MicroK8s).
- [PROMPT_INJECTION]: The skill includes instructions for specialized sub-agents. These instructions partition responsibilities between submission and monitoring tasks. While sub-agents are directed to return control after submission, this is explicitly designed to ensure the main agent remains responsible for providing real-time status updates and logs to the user, enhancing rather than concealing system state.
- [CREDENTIALS_UNSAFE]: Security best practices are enforced throughout the instructions, specifically requiring secrets (like API keys) to be stored in local environment files (".env") and prohibiting the echoing or logging of raw credentials. Sensitive Terraform outputs are properly marked to prevent accidental exposure.
Audit Metadata