If you've used generative AI tools for any amount of time, you've probably found yourself typing the same instructions over and over, explaining your organization's tone, the details that always need to be included, and the format you like, every single time you start a new conversation. An AI Skill is designed to take that repetition off your plate.
Understanding AI Skills
A Skill is a saved set of instructions, along with any examples or templates that go with them, that helps an AI tool do a specific task more consistently. Instead of explaining the task from scratch in each conversation, you write it once and save it, and a tool that supports Skills can pull it in whenever that kind of task comes up. A Skill provides guidance, and you still need to review the result.
What’s Included
Every Skill starts with a name and a short description, which is how the AI knows when the Skill applies. After that come the instructions themselves, written in plain language, the same way you'd explain the task to a colleague. Many Skills also include a good example of the finished product, a template to follow, or a reference file like a style guide or program schedule.
You don't need to write code to create or update an instruction-based Skill. Some Skills also include small programs, which may require technical help to build or review.
An Example
Say your membership team prepares a welcome email for every new family. A Skill could hold your preferred tone, the details every draft needs (hours, parking, how to book classes), and one sample email. From then on, a staff member can simply ask for a welcome email draft for a new family, and the AI can follow the Skill without needing the same long explanation.
Your team still checks the family's details, confirms that the information is current, and reviews the draft before using it. The Skill supports that work, and your people remain responsible for it.
Where Skills Work
Anthropic introduced Skills for Claude in October 2025 and published Agent Skills as an open standard in December 2025. In October 2026, Anthropic reported that more than 70 agent platforms, including ChatGPT, Cursor, and GitHub Copilot, had integrated the standard. Google also documents Skills support in Gemini CLI.
The shared format makes Skills easier to reuse across tools that support it, but a Skill may depend on files, connections, or capabilities that another tool does not have. Availability, setup, and activation also vary by product and plan. Check the specific app you're using rather than assuming every product from the same company works the same way.
How to Use One
- Get a Skill. You can build your own or add one someone else has made. If you're building one, you can describe the task to the AI in your own words and ask it to help you turn that description into a Skill.
- Add it to your AI tool. Follow your tool's instructions for creating, uploading, or enabling Skills. Some offer a Skills section or directory, while others use a different installation process. Your organization's administrator may manage access.
- Ask for the task the way you normally would. The AI can recognize when an enabled Skill fits your request and load its instructions. You can also mention the Skill by name if you want to be sure it is considered. Some tools ask for confirmation before activating it.
- Review the result and update the Skill as things change. When a program, policy, or detail changes, edit the Skill and any reference files so the guidance stays current. Check that the right version is available to the people using it.
Because a Skill can contain instructions and sometimes small programs that run on your behalf, only add Skills from sources you trust, read them before turning them on, and check your organization's AI policy for any guidelines on creating or sharing them.
Once a Skill is in place, the task that used to require a long explanation every time can begin with a single request, and the knowledge of how your team does that work is written down where the people with access, and the AI, can find it.
