Google's Prompt Gems: Turn Your Best AI Ideas Into Chrome Tools

Google’s latest Chrome AI move is either the most practical thing they’ve shipped this year or a solution desperately searching for a problem. Probably both.

“Skills in Chrome” lets you take any AI prompt you keep reusing — your go-to email rewriter, your meeting notes summarizer, your “explain this to me like I’m not a lawyer” button — and pin it as a one-click tool directly in the browser. No more digging through chat history. No more maintaining a sticky note full of prompt templates. The prompt becomes a verb. Something you do rather than something you type.

That’s a genuinely useful idea. Whether Google’s execution earns it a place in your workflow is a different question entirely.

What Google Actually Shipped

Skills in Chrome is a browser-native prompt management layer, sitting between you and whatever AI you’re talking to. The core mechanic: you craft a prompt you’re happy with, save it as a “Skill,” and it appears as a persistent tool you can invoke on any selected text, any page, any context without rebuilding the prompt from scratch.

The demo shows the kind of workflows this unlocks — highlight a paragraph of dense legalese, click your “plain English” Skill, get a clean summary. Paste in a block of meeting notes, hit your “action items” Skill, get a formatted list. It’s prompt engineering made domestic.

The integration is Chrome-level, which means it’s not locked to a single Google product. It’s not just a Gemini thing. It lives in the browser, which is where most knowledge work already happens. That’s the right place to put it.

Skills also appear to be shareable, which opens a more interesting second act: a prompt marketplace where people trade and distribute their best tools. More on that in a moment.

Why Prompt Management Has Been A Real Problem

Here’s something that rarely gets said plainly: prompt engineering has a terrible UX problem. The best prompts people actually use are locked away in chat histories, docs, Discord servers, and Reddit comment threads. The person who figured out the perfect prompt for extracting competitor insights from earnings calls probably uses it every week — but they’re re-typing or copy-pasting it every time, hoping they got it right.

Third-party tools have tried to solve this. PromptLayer, PromptBase, various ChatGPT extensions — they’ve all taken a crack at prompt versioning and storage. Custom GPTs from OpenAI are probably the most mainstream attempt: package a prompt and some configuration into a reusable bot. But Custom GPTs are destinations you navigate to, not tools you reach for. There’s still friction.

What Chrome Skills is doing is collapsing that friction almost entirely. The tool lives at the browser level, which means it’s available everywhere. That’s architecturally smarter than building another app or another tab.

The Competitive Landscape Is About to Get Crowded

Google isn’t operating in a vacuum here, and the timing matters. Microsoft has been shipping Copilot into Edge for over a year. Safari has AI features on the roadmap. Arc from The Browser Company has been experimenting with AI-native browsing since before it was trendy. Brave has Leo, their built-in AI assistant.

But none of them have shipped persistent, user-defined prompt tools as a first-class browser feature. Microsoft’s Copilot in Edge is more about ambient assistance than reusable, custom workflows. Arc’s approach has been clever but niche. Brave’s Leo is competent but hasn’t moved the needle on workflow tooling.

Google Chrome has two things none of those competitors can match: market share and distribution. Chrome sits on roughly 65% of the global browser market. If Skills ships to everyone, it immediately becomes the largest deployment of browser-level prompt tooling in history. That’s not a small thing.

The question is whether Google will fumble the execution the way they tend to fumble things that require sustained product attention.

What’s Actually Interesting: The Shareability Layer

The part of Skills that deserves more attention than the marketing is making clear is the sharing component. If users can package and distribute Skills — whether through the Chrome Web Store, through a curated directory, or through some future Skills marketplace — then Google has essentially built an ecosystem for prompt distribution.

This changes the game in a subtle way. Right now, the value of a well-crafted prompt lives with the person who wrote it. It doesn’t compound. It doesn’t spread unless someone takes the explicit effort to share it. A Skills sharing layer turns good prompts into distributable tools.

Imagine legal teams sharing a “contract risk flag” Skill across an organization. Journalists sharing a “source credibility check” Skill. Developers sharing a “code review” Skill that works on highlighted code in any browser tab. The individual prompt becomes infrastructure.

That’s what Custom GPTs tried to be, but they required people to consciously navigate to a different tool. Skills are ambient. They’re always there.

The Honest Skepticism

Let me push back on the obvious hype, because this needs it.

First, AI features in Chrome have a bad track record. Google has added AI capabilities to Chrome before. Tab grouping summaries. The page summarizer. The help me write feature. How many of those have become staples of your workflow? For most people, the answer is zero. Google has a consistent pattern of shipping AI features that demo well and disappear into the browser settings menu.

Second, this is still tethered to Google’s AI ecosystem. Skills in Chrome, as shown, connects to Gemini. If you’ve built your AI workflow around Claude, ChatGPT, or any other provider, you’re not getting the same integration. You can bet the Skills that surface in the UI will be Gemini-optimized. Google is building a better on-ramp to their own AI, which is fine, but it’s worth naming clearly.

Third, prompt portability is an open question. If you spend a year refining your Skills library in Chrome, what happens when you switch browsers? What happens if Google deprecates the feature or pivots? The best personal prompts are valuable intellectual property in a low-key way — locking them into a browser-native format without clear export paths is a real risk.

Fourth, the demo shows simple use cases. Summarize this. Rewrite this. Explain this. These are one-hop tasks that most power users have already figured out how to handle with bookmarklets, Alfred workflows, Raycast extensions, or just muscle memory. The users who would benefit most from Skills — people doing complex, multi-step AI workflows — will quickly outgrow what a one-click prompt tool can offer.

Who This Actually Helps

Strip away the skepticism and the answer is pretty clear: Skills in Chrome is aimed at the 80% of Chrome users who have never seriously engaged with AI tools and wouldn’t know how to write a good prompt if their job depended on it.

For those users — and there are hundreds of millions of them — having pre-built, community-curated Skills available in the browser is genuinely transformative. They don’t need to understand prompt engineering. They just need to click. Google is abstracting the skill (lowercase) of using AI behind a product called Skills (uppercase), and that’s not a bad trade for most people.

For power users, developers, and anyone already deep in the AI tools ecosystem, this is a nice-to-have that probably won’t change much. You’ll check it out, maybe add a few Skills, and then go back to whatever workflow you’ve already optimized.

The Verdict

Skills in Chrome is a real feature solving a real problem, and Google deserves credit for putting it at the browser level rather than inside another Google product that half the world doesn’t use. The architecture is right. The idea is right.

The execution risk is Google-shaped: they might not give it the sustained product attention it needs to become essential. And the Gemini dependency is a real constraint for anyone who’s already made different AI choices.

But if Google follows through — if the Skills library becomes a place where you can actually find high-quality, domain-specific tools built by people who know what they’re doing — this could be the most mainstream AI workflow improvement since the first browser extensions. The distribution is there. The concept is sound.

Whether Chrome Skills becomes infrastructure or another abandoned experiment is a Google product question, not an AI question. And on Google product execution, the record speaks for itself.