Agent sandbox docs

[isidorn]

We need agent sandbox docs in the next couple of weeks.

Goal - make it clearer to users/enterprises what is going on, and how to control / disable.

Example docs from other products
https://cursor.com/docs/reference/sandbox
https://code.claude.com/docs/en/sandboxing

@ntrogh should this be separate or part of a larger doc? I am happy that to quickly sync next week so we quick start this.

@alexdima @dileepyavan

[ntrogh]

@isidorn For now we have instructions to configure sandboxing embedded in our Agent Tools and MCP docs:

• https://code.visualstudio.com/docs/copilot/agents/agent-tools#_sandbox-terminal-commands
• https://code.visualstudio.com/docs/copilot/customization/mcp-servers#_sandbox-mcp-servers

We also have more conceptual content in our Security doc: https://code.visualstudio.com/docs/copilot/security#_terminal-sandboxing-experimental

Suggestion (let's discuss next week):

• Include in our Enterprise/AI Settings article
• Include in our Concepts/Trust & Safety article

[isidorn]

I like your suggestions. Also we should frame this as "Agent Sandboxing", and terminal sandboxing is just an implementation detail of what we are actually sandboxing. Related issue microsoft/vscode#303421

Let's discuss next week - ping me when you have time. Thanks

[isidorn]

Thank you Nick 🙏