Chrome turns reusable Gemini prompts into one-click Skills

Chrome’s AI layer is shifting from one-off answers toward repeatable workflows. In an April 14 product post, Google introduced Skills in Gemini in Chrome, a way to save prompts and rerun them later with a single selection. The rollout starts on desktop Chrome for Mac, Windows, and ChromeOS users whose Chrome language is set to English-US.

The mechanism is closer to a small browser-native agent workflow than a bookmark. A user can save a prompt as a Skill directly from chat history. Later, inside Gemini in Chrome, they can type forward slash (/) or click the plus (+) button, choose the saved Skill, and run it against the page they are viewing plus any other tabs they select. Saved Skills can be edited, and users can create new ones as their workflows change.

Google’s examples are intentionally everyday, but they show the direction. A Health and Wellness Skill can calculate protein macros for a recipe. A Shopping Skill can build side-by-side spec comparisons across multiple tabs. A Productivity Skill can scan lengthy documents for important information. The point is not that any single prompt is complex; it is that Chrome can now remember the prompt, attach it to browser context, and make it repeatable across sites.

Google is also launching a library of ready-to-use Skills for common workflows. Users might add a Skill that breaks down product ingredients, or one that compares gift options by cross-referencing a budget with the recipient’s interests. Those library Skills are editable, so a user can treat them as templates rather than fixed automation buttons.

The permission model is the detail to watch. Google says Skills use the same safeguards applied to prompts in Gemini in Chrome, and that a Skill will ask for confirmation before certain actions, such as adding an event to a calendar or sending an email. Google also points to automated red-teaming and Chrome’s auto-update protections. This rollout is a sign that AI browsers are moving beyond answer panes: the more important layer may be reusable task units that carry page and tab context without hiding the prompt from the user.