Questions
Common questions.
What is Agent Sandboxing?
Agent sandboxing is the practice of running an AI agent in an isolated environment with restricted access to systems, networks, and data.
How is Agent Sandboxing used in production?
It reduces the blast radius when an agent is compromised, misconfigured, or behaving unpredictably.
Which terms are related to Agent Sandboxing?
Closely related concepts include Agent Context Isolation, Agent Scope, Ambient Authority, Capability Control (AI). Each is defined in the Prefactor glossary.
Keep reading