# Sandbox

Concept · [Building with AI](https://ecosystem.samcarlton.com/ai)

Also called: sandboxing, sandboxed, sand box, isolated environment, agent sandbox

A walled-off space where an [agent](https://ecosystem.samcarlton.com/ai-agent) can run commands and change files without reaching the rest of your computer, your accounts or the open internet. It limits the damage from a mistake or a [prompt injection](https://ecosystem.samcarlton.com/prompt-injection), which is what makes it reasonable to let an agent run without approving each step.

**For example:** A team runs its overnight [agents](https://ecosystem.samcarlton.com/ai-agent) in a sandbox that holds a copy of the code but none of the company's [passwords](https://ecosystem.samcarlton.com/authentication).

## Connects to (9)

- [Permission mode](https://ecosystem.samcarlton.com/permission-mode): The rules that decide which actions an agent may take on its own and which need your approval first.
- [AFK](https://ecosystem.samcarlton.com/afk): Short for away from keyboard.
- [Agent environment](https://ecosystem.samcarlton.com/agent-environment): Everything outside the agent's own software that it can look at or change: your files, terminal, browser, databases and connected services.
- [Prompt injection](https://ecosystem.samcarlton.com/prompt-injection): A security weakness in AI products.
- [Secrets](https://ecosystem.samcarlton.com/secrets): Passwords, API keys and other credentials that software needs to run.
- [AI safety](https://ecosystem.samcarlton.com/ai-safety): The work of making AI systems behave safely and of measuring how safe they are.
- [Environment](https://ecosystem.samcarlton.com/environment): A separate copy of a system used for one purpose.
- [Agent mode](https://ecosystem.samcarlton.com/agent-mode): A named setting in a coding agent that changes how freely it acts.
- [Filesystem](https://ecosystem.samcarlton.com/filesystem): The files and folders on a computer.

Map view: https://ecosystem.samcarlton.com/sandbox