How to sandbox AI agents with OpenShell | AgentOps Unlocked

Publicado em

Vídeo em Inglês, publicado pelo canal Red Hat.

Assistir ao vídeo original no YouTube

Assuntos: Agentes de IA, Assistente pessoal. Publicado no YouTube em 2026-10-20.

Seleção automática por assunto e idioma. O vídeo pertence ao canal de origem.

About OpenShell

OpenShell is an open source runtime from NVIDIA that runs AI agents in isolated sandboxes. What each sandbox can do is set in a YAML policy: which files processes can read or write, which user they run as, and which network destinations each program can reach. Anything the policy does not allow is denied. File and process rules are fixed when the sandbox starts, while network rules can be changed while it runs, according to NVIDIA's documentation.

According to Red Hat, credentials are never stored inside the sandbox: OpenShell injects them at the network boundary, so a compromised agent has nothing to steal. Its OpenShift driver runs each agent as a Kubernetes pod, and OpenShell is planned for integration into Red Hat AI. The video's description adds that the demo uses the OpenCode coding agent and Kata Containers on Red Hat OpenShift.

Fact sheet

  • Project: OpenShell
  • Developer: NVIDIA (open source)
  • Policy format: YAML, deny by default
  • File system enforcement: Landlock (Linux kernel)
  • Red Hat integration: OpenShift driver; planned for Red Hat AI
  • Presenter (per the description): Legare Kerrison, Red Hat Developer Advocate

Read also


Sources and references: NVIDIA OpenShell – Sandbox Policies · Red Hat Blog – Red Hat AI and OpenShell · YouTube description of the video by the Red Hat channel.

Comentários

Postagens mais visitadas deste blog

How to build a JARVIS AI Voice Agent for FREE | Full Livekit Tutorial (2026)

How Does AI Actually Work? 🤖 | AI, Machine Learning & Generative AI Explained