Zero Trust for AI Agents? Google Workspace CLI's Design Philosophy

Greetings from Japan. Every now and then, you stumble upon a technical blog post that disguises itself as a how-I-built-my-CLI walkthrough, only to quietly unfold into something far more interesting. Justin Poehnelt, a Senior DevRel at Google, recently released a CLI for Google Workspace, and wrote about its design. I expected implementation details. What I got was a Zero Trust design philosophy for AI agents, dressed in Rust and JSON. It's the engineering equivalent of ordering a simple bowl of ramen and discovering the chef has been quietly perfecting the broth for thirty years. By the end of this article, you'll see why the principles behind this CLI matter well beyond the command line, and why they might reshape how you think about designing anything that involves AI agents. You Need to Rewrite Your CLI for AI Agents Human DX optimizes for discoverability. Agent DX optimizes for predictability. What I learned building a CLI for agents first. justin.poehnelt.com googleworkspace / cl