Chrome Skills: How to Save Gemini Prompts as One-Click Tools (Complete Guide)

Chrome Skills launched on April 14, 2026 as part of Gemini in Chrome. The pitch is simple: any prompt you find yourself typing more than twice should be saved as a Skill and triggered with a single click. A set of pre-built Skills ships with it, and you can save your own. This guide covers how to actually use it -- setup, the workflows worth saving first, the multi-tab feature most readers miss, and the limits to plan around.

For what Google does and does not document about the pre-built library, see Chrome Gemini Skills Library: What Google Documents. For where Chrome Skills fits in Google's broader AI stack, see our guide to Gemini tool definitions, plugin patterns, Chrome Skills, and CLI workflows.

Key Takeaways

• Open the Skills picker by typing / in the Gemini side panel.
• Manage the library at chrome://skills/browse -- the URL alone is worth bookmarking.
• Free, desktop-only, English (US) for now. No mobile, no other languages announced.
• Skills run on the page you are viewing "along with any other tabs you select", in Google's words.
• Save any chat as a Skill from the chat history; edit any pre-built Skill to fit your workflow.
• The killer pattern is multi-tab aggregation -- prices, summaries, classifications across many tabs in one click.
• Skills can take actions, with a confirmation prompt -- Google names adding a calendar event and sending an email.
• Managed and Enterprise accounts are gated until an administrator enables the feature.

Setup: 30 Seconds to Your First Skill

Chrome's help page lists these requirements for Gemini in Chrome, and they gate Skills with it:

1. Chrome desktop -- "a Chromebook Plus, Mac, or Windows computer", on the latest version of Chrome. Note the drift: Google's April launch post said the rollout covered "Mac, Windows and ChromeOS", while the current help page specifies a Chromebook Plus. On an older Chromebook, check the help page rather than the launch coverage.
2. Browser language set to English (United States). System Settings does not matter -- this is Chrome's own language setting at chrome://settings/languages.
3. Signed into a personal Google account in Chrome, aged 13 or older (or the applicable age in your country), in a supported region.

One gate that decides whether a team rollout is possible at all: Chrome's help page states that "Skills are not currently available for managed or Enterprise accounts unless your organization's administrator has enabled the feature." If you are on a work profile and the / picker has no Skills in it, that is the likeliest reason.

Open the Gemini side panel: click the sparkle icon in the top-right corner of Chrome, or use the keyboard shortcut. The panel slides out from the right edge.

To run your first Skill: type / into the prompt box, or click the plus (+) button. The Skills menu opens; scroll to browse what is available and click one. In Google's words, "your Skill will run on the page you're viewing, along with any other tabs you select."

To browse the full library: navigate to chrome://skills/browse directly, which Chrome's help page describes as "an expanded list of all available skills". Your own saved Skills live at chrome://skills/yourSkills. Google does not publish how many pre-built Skills ship or a fixed list of categories, so your own library page is the only accurate inventory -- see what Google does and does not document.

That is the entire setup. There is no install step, no extension, no configuration file.

Save Your Own Skill From Any Prompt

The most underused feature: any prompt you have already run can be saved as a Skill from the chat history.

The flow:

1. Run a normal Gemini prompt in the side panel: "Extract every email address from this page into a markdown list."
2. After Gemini answers, click the menu icon (three dots) on your prompt in the chat history.
3. Choose "Save as Skill".
4. Name it ("Extract Emails"), optionally edit the prompt to be more general ("Extract every email address from the active page into a markdown list, sorted alphabetically"), and save.

Your new Skill lands in the library and is available from the / picker forever after. The next time you need to extract emails from a page, you click once instead of typing the prompt.

The non-obvious move: edit the prompt before saving to make it more reusable. The prompt you ran was specific to one page; the Skill should work on any page. Add words like "the current page" or "the active tab" to anchor it. Add output format hints ("as a markdown list", "as a bulleted list", "as a CSV") so the Skill produces consistent output every time.

Edit Pre-Built Skills to Fit Your Style

Every pre-built Skill in the library is editable. This matters more than it sounds. The defaults are reasonable, but your team probably has conventions the defaults do not match. The fix is two clicks:

1. Open chrome://skills/browse.
2. Find the pre-built Skill you want to adapt.
3. Click "Edit". (Chrome's help page reaches the same page via / then "Browse Skills"; the pencil icon is a separate control inside the Edit skill popup, described as a way "to make your instructions more robust".)
4. Edit the underlying prompt.
5. Save -- it replaces the original in your library.

Examples of useful edits:

• "Summarize this article" → add "in 5 bullet points, under 100 words total, in second-person voice".
• "Translate this page" → specify the target language and the dialect.
• "Extract action items" → specify your team's task format ("for each action item: assignee, deadline, priority").
• "Find similar products" → add "from these specific retailers: ...".

This pattern -- start with a pre-built Skill, edit the prompt to match your conventions -- is faster than writing a Skill from scratch and produces better-shaped prompts because the Library defaults handle the boilerplate.

The Multi-Tab Killer Feature

Most readers run Chrome Skills against one tab at a time. The feature that makes Skills genuinely powerful is multi-tab execution: select multiple tabs first, then trigger a Skill, and it runs against all of them with combined output.

Google documents the behaviour rather than the exact clicks: a Skill "will run on the page you're viewing, along with any other tabs you select." How to get there:

1. Select multiple tabs in the tab strip with Ctrl-click (Windows) or Cmd-click (macOS).
2. Open the Gemini side panel.
3. Type / and pick a Skill, or run a custom prompt.

The Skill then runs against the selected tabs' content as well as the active page. This is what changes Chrome Skills from a productivity feature to a workflow primitive. Patterns it enables:

• Price comparison across 10 product tabs. Run "Extract product name, price, shipping cost, and availability into a markdown table." Get a single table in seconds. Beats a spreadsheet.
• Summarize 5 article tabs. "Summarize each tab in 3 bullets, then write a one-paragraph synthesis comparing them." Reading time goes from 30 minutes to 3.
• Classify 8 ticket pages. "For each tab, classify as bug / feature request / question / spam, and extract the customer name."
• Compare 4 candidate resume pages. "Compare these candidates on years of experience, technical depth, and apparent culture fit."

Google's wording attaches tab selection to running a Skill in general rather than to particular Skills, so a prompt that names its scope explicitly ("for each selected tab") is the safest way to write one. Google has not published a cap on how many tabs a Skill can read.

The First Five Skills to Save (For Most Readers)

Five example prompts, written for this guide to show the shape a well-formed Skill takes: anchor the scope, name the output format, and say what to do about missing values. Adapt rather than copy.

1. "Summarize for a busy reader." Custom prompt: "Summarize this page in 5 bullets, no bullet longer than 15 words, in second-person voice. End with a one-sentence verdict on whether it is worth reading the full piece."
2. "Extract structured data." Custom prompt: "Find every {entity type} on this page and return them as a markdown table with columns: {column 1}, {column 2}, {column 3}. Where you have to infer a value, mark it as (inferred)." Replace {...} per use.
3. "Compare across selected tabs." Custom prompt: "For each selected tab, identify the same key dimensions and return a single comparison table with one row per tab and one column per dimension." Pair with multi-tab selection.
4. "Draft a Slack message about this." Custom prompt: "Draft a Slack message I could post to my team about the contents of this page. Include a one-line summary, the most important point, and a relevant link if one is present. Keep it under 80 words and in casual voice."
5. "Cite as a source." Custom prompt: "Generate a citation for this page in {APA / Chicago / MLA / your format} format, with all the metadata you can extract from the page. If author or date is missing, mark it as (unknown)."

Run each one a few times against different pages, refine the prompt based on what does and does not work, and re-save.

Skills Can Take Actions, With Confirmation

Worth correcting a common assumption, including one an earlier version of this guide made: Skills are not read-only. Google's launch post says a Skills prompt "will ask for confirmation before taking certain actions, such as adding an event to your calendar or sending an email." Google adds that Skills are "built on Chrome's foundation of security and privacy" and use "the same safeguards we apply to prompts in Gemini in Chrome."

So the thing to understand before you save an action-shaped Skill is the confirmation step, not an assumed inability to act. Google has not published which actions are gated beyond the two it names.

Where Chrome Skills Wins (and Where It Does Not)

Skills wins clearly:

• Repeatable web-content workflows. Anything you do the same way against different web pages is a Skill candidate.
• Multi-tab aggregation. Running one prompt across the tabs you have selected has no clean equivalent in the other browser AI surfaces we cover.
• Low-friction prompt automation for non-power-users. A teammate who would never write a custom Gemini prompt will happily click a Skill button.
• Browser-as-source-of-truth workflows. Skills lives where the content already is.

Skills does not win:

• Long agentic tasks. Skills produce one response per click, with a confirmation prompt in front of any action. For multi-step workflows that need planning and tool use, Gemini CLI, Claude Code, or Codex are the right shape. For driving desktop applications, see Perplexity Personal Computer or Codex computer use.
• Anything outside the browser. Skills lives in Chrome. If your work is in native apps, it is the wrong tool.
• Non-English (US) workflows. Skills is gated to English (United States). Google has not announced other languages or a timeline.
• Managed and Enterprise accounts, unless an administrator has enabled the feature.

Limits and Privacy

A few practical things to know before you go heavy:

• Page contents go to Google's servers to generate the response, the same as any other Gemini in Chrome interaction. Google describes Skills as using "the same safeguards we apply to prompts in Gemini in Chrome", so the privacy terms that govern Gemini in Chrome govern Skills too.
• Skills are tied to your Google account, not your device. In Google's words, "your saved Skills are available on any signed-in Chrome desktop device."
• Actions are confirmed, not blocked. See the section above.
• Google publishes no limits on prompt length, output length, or the number of tabs a Skill can read. If you need a hard number for a workflow, test it yourself rather than trusting a figure from secondary coverage.

Where Chrome Skills Goes Next

Undocumented rather than ruled out, and each worth re-checking rather than assuming:

• Multi-language rollout. English (United States) is still the documented requirement as of August 13, 2026. Google has not published a timeline for other languages.
• Mobile. Still desktop-only, despite Gemini being available in Android Chrome. This is the obvious gap.
• Sharing Skills. Skills remain personal, and the managed-account gate suggests Google is treating organizational deployment as an admin question first.
• Skill chaining. Running one Skill on another's output is not announced and not documented.

The right move for most readers: spend an hour browsing chrome://skills/browse, save the three patterns you find yourself wanting most often, and run them for a week. The workflows that stick reveal themselves quickly. Anything you find yourself typing repeatedly is a Skill waiting to be saved.

Sources

Every claim above was re-checked on August 13, 2026 against these pages:

• Google, "Turn your best AI prompts into one-click tools in Chrome", April 14, 2026: https://blog.google/products-and-platforms/products/chrome/skills-in-chrome/
• Google Chrome Help, "Customize your Gemini in Chrome experience": https://support.google.com/chrome/answer/16988996
• Chrome Enterprise and Education Help, "Prompting": https://support.google.com/chrome/a/answer/17031594

chrome://skills/browse and chrome://skills/yourSkills are browser-internal pages, so they cannot be fetched or archived and nothing about the library's contents is verifiable from outside your own Chrome profile.