How to Use Claude Skills in ChatGPT, Gemini and Copilot

You can use Claude skills in ChatGPT, Gemini and Copilot, because a skill is a text file and all four assistants now read the same one. The question comes up more than any other in this course, usually from someone whose office runs on Microsoft 365 or whose bookkeeper lives in ChatGPT: "I found a skill I want, but I don't use Claude." You don't have to. This lesson gives you the exact menus for each assistant as of October 2026, what changes when a different model reads the file, and the ten-minute test that tells you whether the install took.

Why one file works in four assistants

Every skill in the Vexlo toolbox is a folder with one file in it, SKILL.md. The top of the file holds a name and a one-line description of when to use it. The rest is the five sections you met in Chapter 1: Role, Setup, Capabilities, Rules, Output format. There's no code in it and nothing that only Claude can run.

That format turned out to be the one everyone adopted. Anthropic published it first. OpenAI's documentation describes a ChatGPT skill as a directory with a SKILL.md that must carry a name and a description. Google's help page says you can import skills from other platforms by uploading their SKILL.md files. Microsoft's Copilot Studio blog says its skills follow the same open format. Four companies, one file. The folder name has to match the skill's name in lowercase with hyphens, which is how the GitHub copies are already named.

So the install question is only ever "where does this assistant want the file." Here's the table, then the detail.

ChatGPT

OpenAI's release notes say skills are now generally available, and that Enterprise and Edu users find them under Plugins in the sidebar, on the Skills tab of the Plugin Directory. Business workspaces got the same note. On the desktop app there's a Skills entry in the sidebar that shows skills across your projects. Download the skill's folder from the Vexlo GitHub repo and add it there.

On a personal plan, look for that Skills entry first; the rollout has moved plan by plan all year and your account may have it by the time you read this. If it isn't there, the route that works on every plan is a ChatGPT Project: create one named after the skill, paste the whole SKILL.md into the project's instructions, and every chat inside that project starts with the skill loaded. Don't build a custom GPT for this. OpenAI plans to retire custom GPTs on December 11, 2026, and anything you build there has to move anyway.

Gemini

Google launched skills in Gemini on September 30, 2026, and they're replacing Gems. For personal accounts, Gems become read-only on November 17, 2026 and are removed on December 17, with existing Gems converted to skills automatically. If you built a Gem for a repeated job, this is the month to rebuild it as a skill on purpose rather than wait for the migration.

The upload path: on gemini.google.com, click Customize in the sidebar (click More first if you don't see it), then Skills, then Create. You can create one manually in the editor or upload a SKILL.md, which Google's help page says is how you import a skill made on another platform. Uploads work in the web app and the Mac app as of October 2026, not on the phone. A skill runs with a slash command, so the bookkeeping copilot becomes /bookkeeping-copilot, and Gemini can also trigger a skill on its own when a request matches. You can keep an unlimited number of skills with 100 active at a time, and each can carry up to 100 MB of reference files, which is where your chart of accounts or your price list goes.

Copilot

Microsoft 365 Copilot has two routes. The simple one: choose New agent, pick Skip to configure, and paste the skill's text into the Instructions field. The field holds 8,000 characters, and most Vexlo skills come in under that; if yours doesn't, cut the Example section and keep Role, Setup, Rules and Output format. Add your own documents as knowledge sources, and test it on the Try it tab before you share it. Microsoft's documentation notes that the knowledge sources available depend on your licence, so a Copilot Chat user without the full add-on will see fewer options than a licensed one.

The second route is newer: Microsoft's agent builder documentation says you can add skills to an agent by providing a skill package, and Copilot Studio reads a standalone SKILL.md or a zip. If your IT person manages Copilot Studio, hand them the folder. If you're on plain Copilot chat with no agent builder, the fallback from the install lesson still works: paste the file at the top of the chat and carry on.

What changes when a different model reads the file

The same instructions, read by a different model, behave a little differently. Expect three things.

The tone shifts. Claude tends to follow the Rules section closely; some ChatGPT models paraphrase the output format the first time and settle down after a correction; Gemini is more likely to add a summary you didn't ask for. None of this is a broken install. It's the same brief read by a different new hire.

The setup interview may be skipped. Some assistants start working before asking the Setup questions. If the first reply doesn't ask you about your business, type "run the setup interview first" and it will.

House rules need re-adding. If you customized a skill in Claude (Chapter 2's lesson on making it yours), those edits live in your copy of the file, so carry the edited file across, not the original from GitHub.

Try it now

Pick the assistant you use most and put one skill in it, the same one you installed in Claude in this chapter's first lesson. Then send the same first line in both:

If the two lists match, the file carried. Then give both assistants one real task from your week, the same words, and read the two outputs side by side. You'll see the tone difference described above, and you'll see which assistant you'd rather run this skill in. That's the answer to "which AI should I use," and it took ten minutes instead of a subscription decision.