Google rolled out a feature this week that, on the surface, looks like a productivity gimmick — save your Gemini prompts as reusable "Skills" and trigger them with a slash command. But look closer and you'll see the skeleton of something much more interesting: a native, cross-tab AI automation layer baked directly into the world's dominant browser.

What Chrome Skills Actually Does

Starting April 14, Chrome desktop users (English-US, signed in) can save any Gemini prompt as a Skill directly from their chat history. Once saved, you trigger it by typing / in the Gemini sidebar or clicking the + button. Your prompt runs against whatever page you're viewing — and here's the part that matters — against any additional tabs you select simultaneously.

Google also shipped a library of 50+ pre-built Skills covering common workflows: comparing products, summarizing documents, breaking down recipe nutrition. You can customize any library Skill's underlying prompt after saving it. Everything syncs across your signed-in Chrome desktops automatically.

The UX borrows heavily from developer tooling. The slash-command trigger is straight out of Slack, Discord, and every code editor with an AI feature. The save-from-history flow is genuinely smart — you don't design these upfront, you capture the prompts that already worked.

Multi-Tab Dispatch Is Where It Gets Interesting

The headline feature isn't saved prompts. It's cross-tab execution.

When you fire a Skill, you can select multiple open tabs, and Gemini processes all of them in a single pass. Think about that concretely. You're evaluating five cloud hosting providers. Before this, you'd either manually copy-paste specs into a spreadsheet, build a quick scraping script with Puppeteer, or resign yourself to flipping between tabs for an hour. Now you open five tabs, trigger a comparison Skill, and get a synthesized side-by-side breakdown pulling from all five pages at once.

This is the first time a browser has shipped native multi-page AI reasoning as a built-in capability. Not an extension you have to trust with your browsing data. Not a third-party SaaS tool. Just a feature that ships with Chrome updates, protected by the same automated red-teaming Google applies to all Gemini interactions.

The security model makes the right tradeoffs. Read-only page analysis runs without friction. Any "write" action — sending an email, adding a calendar event — requires explicit confirmation first. That's the correct default for something running across arbitrary web pages.

I ran a few cross-tab workflows during the initial rollout and the results were surprisingly coherent. Gemini handles structured content (product specs, documentation pages, pricing tables) well. Unstructured long-form content gets shakier, especially when you're asking it to find differences across pages that are written in completely different formats. Your mileage will depend heavily on how specific your Skill prompt is.

The Extension Ecosystem Problem

Here's what nobody in the Slashdot thread is saying directly: a decent chunk of Chrome extensions just got commoditized. Price comparison tools, content summarizers, recipe converters, product research helpers — a well-crafted Skill replaces most of these without any installation.

The extensions that survive will be the ones doing things the automation layer can't: deep external API integrations, persistent state across sessions, and complex UI overlays. Pure "read the page and reformat the content" extensions are squarely in the blast radius.

The Actual Gap: No Developer Surface

Here's my honest frustration with this launch: there's no API.

You create these through the UI. You share them by... telling someone what prompt to type. There's no manifest format. No Chrome Web Store integration. No way to package a collection for a team or distribute them programmatically. You can't chain them, trigger them conditionally, or feed one Skill's output into another.

The obvious next step is exactly that platform layer. Picture your company distributing a Skills pack — "Summarize this Jira ticket," "Extract action items from this meeting recording," "Compare these AWS pricing pages against our current contract" — deployed the same way you'd push a managed Chrome extension today. That's clearly where this is headed. Google shipped the consumer feature first and left the developer surface for later.

Google's official position, per The New Stack, is to use the Gemini API with function calling if you want programmatic workflows. Which is fine, but that's a completely different paradigm — server-side API calls versus client-side browser-native automation. They're solving different problems.

What's Worth Doing Right Now

The actionable bits for developers are limited but real:

Test multi-tab workflows against your actual daily work. Open your monitoring dashboards, your CI logs, your Sentry error pages. Fire a Skill that synthesizes across them. You'll quickly find where Gemini handles your domain well and where it falls apart — and that knowledge will be valuable when the developer API inevitably lands.

Audit your team's internal extensions. If you maintain Chrome extensions that mostly read and summarize page content, Skills might already cover the use case. That's engineering time you could reclaim.

Watch for the platform announcement. Google hasn't given timelines, but the infrastructure is there — Gemini's function-calling system, Chrome's extension distribution pipeline, the enterprise managed-browser stack. When the developer surface ships, having built intuition around what Skills can and can't do will matter.

Chrome has been a document viewer bolted onto a JavaScript runtime for two decades. Google just started turning it into a prompt-driven automation platform. Whether that platform opens up to developers will determine if this is a neat consumer feature or a genuine shift in how browser tooling works.