How to Create Skills in Gemini: A Step-by-Step Guide with Examples. Bonus: What’s Next for Gems

What Skills Are in Gemini

A skill is a named set of instructions that describes when to apply it. Google compares a skill to a label that is always within reach. The user defines once how to perform a task, and then triggers that workflow with a single command.

Working with skills is set up in three modes:

Manual trigger. In the input field, type / and select a skill from the list. Google promises to add triggering via @.

Automatic trigger. Gemini enables an active skill on its own if the user’s request matches its description. Auto-triggering can be disabled for each skill individually.

Combining. For complex tasks, multiple skills run at the same time. Google gives an example: a skill with your personal writing style and a skill with a brand guide together produce text “in your voice” and in the company’s branded format.

This logic aligns with the Agent Skills standard. According to the standard’s description, an AI agent initially sees only the name and description of each skill, and loads the full instruction text only when the task requires it. That’s why a skill’s description matters more than it seems: it determines whether auto-triggering will kick in.

In the teaser, Google shows three scenarios: prepping for a presentation, writing a personal-style email using context from Workspace, and brainstorming with a topic breakdown from three to five angles. All three map to tasks people repeat every week.

How Skills Differ From Gems

Gems also stored instructions, but they worked as standalone chatbots. To use a Gem, you had to open it in the sidebar and start a new conversation with it. A skill, by contrast, plugs into any chat and doesn’t require switching between assistants.

The main practical advantage is portability. You can download a skill as an archive and upload it into another compatible tool, such as Claude or Gemini CLI. Gems didn’t offer that option. 

When will Gems Disappear 

Google is phasing out Gems depending on the account type. The company promised to automatically migrate all Gems to skills when the time comes.

The timeline looks like this:

1. November 2026 — personal Google accounts
2. March 2027 — Google Workspace for business, enterprise, and nonprofits
3. June 2027 — Google Workspace for Education.

You don’t have to wait for the automatic migration. You can migrate a Gem manually right now — the steps are described below in a separate section. 

Where Skills are Available and Who can Use Them

According to Google data, skills are available on all Google AI plans. The rollout is global, but with a few conditions.

According to the help page, to use skills you need:

• a personal Google account — work and school accounts will get the feature later
• age 18+ — access for younger users is promised later
• Keep Activity action history enabled
• the web version at gemini.google.com, the mobile app, or the Gemini app for Mac

For Google Workspace customers, skills will become available “in the coming weeks.” Google did not name an exact date.

Five Ways to Create a Skill 

All skills are collected on a separate page. To open it in the web version, click “Settings” in the sidebar and select “Skills.”

Google’s help page describes five ways:

1. Create with Gemini. Gemini asks clarifying questions and writes the name, description, and instructions itself.
2. Template from the recommended section. A ready-made draft where you only need to tweak the text.
3. Create manually. A blank form with three fields: name, description, and instructions.
4. Upload. An SKILL.md file or a .zip archive containing it.
5. From chat. Just tell Gemini, “Create a skill based on these instructions,” and list your requirements.

When uploading a file, technical requirements apply. SKILL.md must be in the root folder, and the skill name inside it must be written in lowercase Latin letters with hyphens, for example my-new-skill. Supported text formats include .txt, .md, .py, .json, .yaml, .csv, .html, as well as PDFs and images. You cannot upload .docx and .xlsx files. The total size of the attached files is up to 100 MB.

The fastest route for a beginner is Create with Gemini. This is exactly what Google shows in the announcement video, so the hands-on part starts there.

Hands-On: A Skill for Presentation Prep

For the demo, the editorial team recreated the main use case from Google’s announcement — the Presentation Prep skill. In the company’s concept, it helps you outline your slides, write speaker notes, and prepare for tough questions. Then a second skill with a personal style is added to test combining skills.

The topic of the test presentation is a 10-minute talk, “How AI agents work,” for an unprepared audience.

Step 1. Launch Create with Gemini

On the Skills page, click Create with Gemini. A chat will open where Gemini will ask what task the skill should solve. It’s best to provide the description right away, and in detail.

Prompt

Create a presentation prep skill. When I give it a topic, audience and time limit, it should: 1) propose a slide-by-slide outline with one key message per slide, 2) write short speaker notes for each slide, 3) list the 5 toughest questions the audience could ask and give a concise answer to each. Keep slides to max 12 for a 10-minute talk. Ask me for the audience and time limit if I don't provide them.

Gemini may ask a couple of clarifying questions, for example about the presentation language or the output format. After you answer, it will generate a draft of the skill.

Step 2. Check the Name, Description, and Instructions

Gemini will show a skill card with three fields. The name is automatically converted to the presentation-prep. format. The description is worth rereading especially carefully: Gemini uses it to decide when to invoke the skill without the / command. 

Google’s help page recommends writing when to use the skill, keeping instructions short, and adding an example of the desired result. If something doesn’t look right, you can edit the field manually or click Edit with Gemini. After checking, all that’s left is to click Create. 

Step 3. Invoke the Skill via / 

Now the skill is available in any chat. In the input field, type /, select presentation-prep, and finish typing your task.

Prompt

/presentation-prep Topic: how AI agents work. Audience: people with no technical background. Time: 10 minutes.

As a result, the user gets slides with a single request, without a long prompt. 

Step 4. Create a Second Skill From a Chat With a Sample File

The second use case is an email in a personal style. For this, it is convenient to create it from a chat and attach a file with samples of your own writing. A .txt or .md file with three to five excerpts from emails, posts, or articles will work. A .docx document will not work: skills do not accept that format. 

Prompt

Create a skill called my-style based on the attached writing samples. It should rewrite or draft any text in my voice: short sentences, no buzzwords, no exclamation marks, concrete examples instead of general claims. Use it whenever I ask to write or edit text for me.

Gemini will suggest saving the skill. After you confirm, the sample file will be stored inside the skill, and Gemini will refer to it every time you invoke it. 

Step 5. Combine two skills

Now you can apply both skills in a single prompt. The task is to rewrite the talk’s key points from step 3 so they sound natural.

Prompt

/presentation-prep /my-style Rewrite the speaker notes for slides 1-4 from the outline above so they sound like me talking, not like a textbook.

That’s the key difference from Gems. Before, you’d have to copy text from one Gem into another, but now both sets of instructions can work in a single response.

How to Manually Move a Gem Into a Skill

The automatic migration will happen without any user involvement. But if a Gem is important for your work, it’s better to move it ahead of time and verify the result.

Google describes the process as follows:

1. Open the Gem and download all knowledge files attached to it.
2. Create a new skill and copy the name, description, and instructions from the Gem into it.
3. If there are multiple knowledge files, put them into a single folder together with SKILL.md and upload it via Upload.

When migrating, keep one limitation in mind: GitHub files are not supported in skills. Text files, PDFs, and images migrate without issues.

Limitations 

Skills have only just launched, and some features you may be used to from Gems are not available yet. Google’s help page lists several limitations.

Not all tools. Canvas, Deep Research, video and music generation, and Guided Learning mode do not work inside skills. Gems that use these tools will lose some functionality after migration.

Limit. No more than 100 active skills.

No sharing. Google says sharing skills, adding files from Google Drive, and adding notebooks from Gemini Notebook will be available “in the coming weeks.”

Format limitations. You can’t upload .docx, .xlsx, or GitHub files, and the total size of attachments is limited to 100 MB.

Issues in the mobile app. According to Google’s documentation, manually created skills are not yet saved in the mobile apps, and skills with attached files can’t be edited there.

Personal accounts only, 18+. Workspace users and anyone under 18 will get access later.

Deleting a skill is irreversible. Before experimenting with an important skill, you should download it.

What Changes for Gemini Users 

With the shift to skills, Google is moving away from the “one bot per task” model and toward skills that can be plugged into any conversation and work together. Skills work in a similar way in Claude Cowork and in Gemini CLI.

Lazy Business: How to Use Plugins in Claude Cowork

04.03.2026

Read now

For personal account holders, the key deadline is November 2026. Until then, it’s worth checking your Gems, especially those that use Canvas or Deep Research, and manually migrating them if needed. For new workflows, it makes sense to create skills right away: they переносятся between services and aren’t tied to a single chatbot.

Skills in Gemini use the open SKILL.md format. The same skill can now be used in Gemini, Claude, ChatGPT, and dozens of agent tools without rewriting instructions for each service.