- Gemini Skills are portable instruction packages in SKILL.md format that allow you to automate recurring and customized tasks.
- The system allows the combination of multiple skills in the same chat to optimize complex results.
- Portability is key, allowing configurations to be migrated between different AI platforms using open standards.
If you've ever felt that repeatedly giving the same instructions to an AI is a chore, Gemini Skills are here to save the day. Essentially, they're reusable instruction packages that teach the assistant exactly how to perform a specific task, preventing you from having to type lengthy prompts every time you open a chat.
The most interesting thing of all is that Google has made a strategic shift by moving away from the old Gems to adopt an open standard based on SKILL.md files . This means that your configurations are no longer tied to a single platform, but you can move them between different AI assistants, making these skills incredibly versatile personal micro-tools.
What exactly are Gemini Skills?
To put it simply, a Skill is a set of rules with a clear name and, optionally, some reference files that provide context. You can activate them in two ways: by typing the / symbol in the chat to choose the one you need, or by letting the AI be smart and activate it automatically if it detects that your request matches the Skill's description.
One of its crown jewels is the ability to combine multiple Skills in a single conversation. Imagine you have one for writing style and another for your brand guidelines; Gemini can apply both simultaneously for a seamless result without you having to juggle the text.
Internal workings and data loading
Many people overlook how Gemini decides which Skill to use, but it's vital to avoid errors. The process occurs in three steps: first, the AI analyzes the name and description to see if there's a relevance; second, if there's a match, it reads the complete instructions; and third, it only opens attachments if they're truly necessary to respond.
Here's a detail where many people make a mistake: unlike Gems, which read Drive files in real time, Skills save a snapshot of uploaded files . If you update a document in your Drive, the Skill won't know; you'll have to re-upload the entire package with the updated files so the AI has the current information.
Requirements and costs: Who can use them?
Not everyone has immediate access. To play with Skills, you must be at least 18 years old and use a personal Google account, as educational or business accounts are currently unavailable. Additionally, you must have Keep activity enabled.
Regarding money, there's a lot of confusion on forums like Reddit, but the reality is that Skills are free . You don't need a paid subscription to create them or use them in your regular chats. What may vary depending on your plan is the overall Gemini usage limit or access to more advanced features like Gemini Spark.
How to create your own skill step by step
Google offers four ways to set up a Skill from the Settings menu or chat: you can ask Gemini to create it using a prompt, use pre-designed templates, start from scratch with a blank canvas, or, the most professional option, upload a SKILL.md file.
If you're looking for something serious, I recommend going straight for the .md file, as this gives you complete version control and allows you to easily transfer your work to Claude or ChatGPT. However, keep in mind that manually created Skills aren't saved in the mobile app, and those with attachments can only be edited via the web.
Limitations you should know about
It's not all sunshine and rainbows. There are limits you can't ignore: you can create as many Skills as you want, but only 100 active at a time . The maximum upload size is 100 MB, and some formats are prohibited: forget .docx or .xlsx; you must convert them to PDF, CSV, or Markdown before uploading.
Another critical point is scripts. If you import a Skill that relies on calls to external APIs, it most likely won't work, since Gemini scripts can't access the internet . They are limited to processing the Skill's internal information.
The transition from Gems to Skills
If you had Gems, don't worry, because Google will automatically convert them to Skills. But if you want to keep track, you can do it yourself by downloading the knowledge files , creating a folder with the Skill's name (in lowercase and with hyphens), and uploading the SKILL.md file along with its references.
A pro tip: When writing the description, don't use a generic phrase, but rather a trigger like "Use this when..." . Since Gemini uses the description to decide whether to activate the Skill, being specific helps prevent the AI from selecting the wrong tool.
Other implementations: Chrome and Enterprise Environments
There's a simplified version called Skills for Chrome that functions more like a shortcut of saved prompts in the sidebar. These aren't autonomous agents, but rather quick commands for scanning web pages or summarizing content with a single click, making them ideal for those who use AI as a constant navigation assistant.
In the Gemini Enterprise environment, things get more serious. Here, the app administrator controls who shares what . The process includes a review where the administrator can approve or reject sharing an agent with Google Identity groups, ensuring that sensitive information doesn't circulate unchecked throughout the office.
Advanced usage via CLI and Firebase
For those comfortable with the terminal, Gemini CLI allows you to discover Skills in the .agents/skills directory. This enables you to run skills from the shell , facilitating integration with development workflows. You can even install specific Skills, such as those from Firebase, so AI can help deploy web applications by managing authentication and Firestore databases.
This approach transforms AI from a simple chat application into a technical command center . The ability to install portable knowledge modules allows developers worldwide to share streamlined workflows, reducing token costs and increasing the accuracy of technical responses.
Integration into the education sector
In academia, Gemini LTI allows instructors to share notebooks and Gems with their students directly on platforms like Canvas or Schoology. Students have read-only access to the materials, while instructors manage the content from a sidebar, ensuring course materials are always just a click away.
The management of these educational tools is governed by Google Drive policies . If a teacher deletes a Gem, the student immediately loses access, allowing for complete control over the validity of the learning materials shared during the semester.
Ultimately, we are witnessing an evolution where AI is no longer just an empty text box, but rather an ecosystem of customized and portable tools. Whether through the simplicity of a / command in chat, the power of a SKILL.md file, or integration into corporate and educational environments, the key lies in knowing how to structure information and leverage interoperability to avoid dependence on a single software.