The Agent Skills Directory

[SAFE]: The skill provides operational guidance for managing NVIDIA RAG Blueprint deployments. It correctly identifies and uses official repositories and tools.
[REMOTE_CODE_EXECUTION]: The skill instructs the agent to install packages via pip and uv, and to deploy services using docker compose and helm. All identified sources, such as official package registries and NVIDIA's NGC catalog, are recognized as trusted organizations and well-known services.
[COMMAND_EXECUTION]: The skill utilizes numerous shell commands for environment analysis, service management, and health verification. These commands are standard for the described tasks (e.g., nvidia-smi, docker ps, kubectl get pods) and do not exhibit malicious intent.
[CREDENTIALS_UNSAFE]: The skill handles NGC_API_KEY and NVIDIA_API_KEY by checking for them in environment variables or .env files. It follows secure practices by instructing the user to provide these secrets through their shell environment rather than hardcoding them within the skill.
[DATA_EXFILTRATION]: No evidence of unauthorized data exfiltration was found. Network operations are limited to health checks on local services or interactions with official NVIDIA API endpoints required for the RAG pipeline's functionality.

rag-blueprint