
Google just replaced one of the most useful Gemini features with something better, and most users have no idea it happened. Gems, the custom AI assistants you could build inside Gemini, are being retired. Their replacement is called Skills, and the upgrade matters more than the name change suggests.
I spent the past week testing Skills across real tasks: meeting notes, code review checklists, content editing, and travel planning. The result surprised me. Skills remove the friction that kept Gems from going mainstream. This guide explains what changed, how to build your first skill, and which seven skills deliver the most value right now.
Gemini Skills are reusable custom instructions you invoke inside any chat with a slash command. Google is replacing Gems with Skills starting November 17, 2026 for personal accounts, with Workspace business accounts following in March 2027 and education accounts in June 2027. Anyone with a personal Google account can create Skills for free, something that previously required a paid subscription for Gems.
The three biggest improvements: you can stack multiple skills in one conversation, Gemini can auto-detect when a skill fits your request, and you no longer need to open a separate side panel to use a custom configuration. If you used Gems before, Google will migrate yours automatically. If you never built one, now is the easiest time to start.
A skill is a bundle of instructions that tells Gemini how to handle a specific task. You define a name, a description of when it applies, and the step-by-step instructions Gemini should follow. Once saved, the skill lives in your account and waits for you to call it.
Gems did the same job from August 2024, but they worked like separate apps. You opened the Gems manager in the sidebar, picked one configuration, and started a new session inside it. Need a second configuration in the same conversation? You had to start over. That design kept Gems power-user territory.
Skills flip the model. Instead of visiting a special room, you bring the instructions into whatever conversation you are already having. Type a forward slash, pick the skill, and keep working. Google announced the change in a September 30 post on the official Gemini blog and confirmed that personal-account Gems will migrate automatically when they go away.
The practical difference shows up in daily use. Always-on AI agents pursue goals in the background, while skills stay on call inside your chat. They solve the smaller but more frequent problem: repeating the same detailed instructions every time you ask Gemini to do something routine.
The table below compares the two formats head to head, so you can see exactly what the migration means for your workflow.
| Feature | Gems | Skills |
|---|---|---|
| How you use it | Open Gems manager, start dedicated session | Type / in any chat, apply instantly |
| Multiple configs per chat | No | Yes |
| Auto-detection | No | Yes, Gemini can invoke skills itself |
| Cost | Required paid subscription | Free with personal Google account |
| Reference files | Supported | Supported |
| Sharing with others | Supported | Coming in the next few weeks |
| Google Drive files | Supported | Coming in the next few weeks |
| Active limit | No published cap | 100 skills |
| Retirement date | November 2026 (personal accounts) | Current format |
The free access change is the one that affects the most people. Gems sat behind a paywall, which limited adoption to subscribers. Skills open to every personal Google account, so the feature finally reaches the casual users who repeat the same prompts every week without ever configuring anything.
One detail worth knowing: Google says the slash trigger will eventually become an at sign. Do not build deep muscle memory around the slash, because the gesture is already scheduled to change.
Google offers two creation paths: a conversational one inside chat, and a manual one in Settings. Both produce the same result. Use chat when you know the workflow but not the wording; use manual when you already have precise instructions drafted.
Method 1: Create from chat. Open a standard Gemini chat and describe the workflow in plain language. For example: “Create a skill that turns my raw meeting notes into a summary with decisions and action items, each assigned an owner.” Gemini drafts a skill name, a description of when it applies, and the processing instructions. Read the draft in the conversation, ask for tweaks (“add a rule: never invent owner names”), then hit Save Skill. The skill registers to your account instantly and becomes callable with /.
Method 2: Create manually. Go to Settings, then Skills, then Create manually. Fill in three fields: name (short and specific, like “Meeting Notes Formatter”), description (when Gemini should use it, for example “when I paste raw notes from a call”), and instructions (the step-by-step rules). Click Create.
Google published best practices for writing effective skills. Three rules stand out: define the specific situations where Gemini should use the skill automatically, keep instructions focused on the unique guidelines for the task, and include one concrete example of the output you expect. I follow a fourth rule of my own: keep instructions under 200 words. Long instructions dilute; short instructions stick. Prompt chaining research backs this up: focused steps beat broad guidance.
You can also import skills from other platforms by uploading a SKILL.md file, which makes the format portable. If you wrote custom instructions for another AI tool, you may be able to reuse them directly.
Invocation is the simplest part. Inside any Gemini conversation, type a forward slash. A menu appears with your skills. Select one, and its instructions apply to everything you write next in that session. No tab switching, no new window.
Stacking is where skills pull ahead of Gems. You can activate more than one skill in a single chat. I often combine an editing skill (tone rules for our blog) with a formatting skill (output as a table). Under Gems, that meant choosing one configuration and losing the other. Under Skills, both run together.
Auto-invocation handles the third case: you never call the skill at all. A description field that clearly states the trigger situation lets Gemini detect the pattern and apply the skill on its own. This works best when descriptions name concrete signals. “When I paste text longer than 500 words asking for feedback” triggers reliably. “For writing tasks” does not, because it is too vague.
Skills also extend beyond chat. On Google AI Pro and Ultra plans, skills power Gemini Spark tasks and schedules, so the same instructions you build for conversation can drive automated workflows. Chat skills themselves need no paid plan.
Start with tasks you repeat. A skill earns its keep when it replaces instructions you type more than twice a week. These seven cover the most common patterns I see across freelancers, students, and small teams.
1. Meeting notes formatter. Paste raw notes; get back decisions, action items with owners, and deadlines in a consistent layout. Pair it with AI meeting-notes tools if you want the transcription automated too.
2. Email tone checker. Define your professional voice once. The skill flags sentences that sound curt, vague, or passive before you send anything important.
3. Code review checklist. Instructions that make Gemini check for specific issues: missing tests, insecure patterns, unclear naming. Developers who already use structured prompt testing will recognize the pattern.
4. Study explainer. Paste a dense paragraph from a textbook; get back a plain-language explanation plus three self-test questions. Reference files let you attach your syllabus so examples stay in scope.
5. Social caption writer. Brand voice rules, hashtag limits, and call-to-action placement encoded once. Every caption after that follows the same standard without re-explaining it.
6. Travel day planner. Instructions that turn “three days in Tokyo with a kid” into a timed itinerary with transit notes and meal breaks built in.
7. Budget review assistant. Paste monthly spending; get back category totals, the three largest changes versus last month, and one question about any line item that looks off. No-code automation workflows can feed it fresh data automatically.
Build the one you will use tomorrow, not the one that sounds impressive. A single well-used skill beats ten dormant ones.
If you built Gems in the past two years, Google says you do not need to do anything. Migration happens automatically when Gems go away for your account type. Personal accounts convert starting November 17, 2026. Workspace business, enterprise, and nonprofit accounts follow in March 2027. Education accounts follow in June 2027. Your Gems keep working normally until the cutoff.
If you prefer to move early, follow this manual process:
1. Back up knowledge files. Open each Gem, download any files listed under its Knowledge section, and save them locally. Google advises this step because the help center does not specify which file types transfer automatically.
2. Create the skill. Go to Settings, then Skills, then Create manually. Copy the Gem name, description, and instructions into the matching fields.
3. Reattach files. Place the downloaded files in a folder whose name matches the skill name in lowercase hyphenated format, alongside the skill file, then upload the folder together.
4. Test before the cutoff. Invoke the new skill with / in a fresh chat and run one real task through it. Fix anything that behaves differently from the Gem before you delete the original.
Two caveats from Google’s own documentation: GitHub files are not currently supported inside skills, and Opal, Google’s mini-app experiment, also shuts down in November 2026 without migrating into skills. If you relied on either, plan a replacement rather than waiting. The full migration timeline and the rollout details for Workspace accounts spell out which account types move when.
Skills are new, and the rough edges are documented. Go in with eyes open.
Some Gemini tools do not work with skills yet. Google lists Canvas, Deep Research, video creation, music creation, and guided learning as unsupported for now. If your workflow depends on those, skills cannot carry it yet.
Sharing and Drive integration are still on the way. Under Gems you could share configurations and pull files from Google Drive. For skills, both features are promised in the coming weeks but are not live today.
Personal accounts only, for now. Skills currently require a personal Google account. Workspace rollout began in early October, with the Gemini app following, so business users may need to wait a few more days depending on their admin settings. Requirements also include being over 18 and keeping the Keep Activity setting on, which matters if you turned it off for privacy reasons.
The 100-skill cap. You can create as many skills as you like, but only 100 can be active at once. In practice this affects almost nobody, yet power users migrating large Gem libraries should audit before importing everything.
None of these block the core use case. A step-by-step creation guide walks through both creation paths if you want screenshots, and the November 17 migration explainer covers the automatic conversion in detail. The difference between Gems and skills comes down to one idea: skills meet you inside the chat instead of asking you to visit them.
Skills are the rare platform change that makes a feature simpler instead of just newer. The slash invocation removes the setup tax, stacking removes the one-configuration limit, and free access removes the paywall that kept most people out. I recommend starting with a single skill for the task you repeat most often, testing it on two real requests this week, and refining the instructions based on what comes back.
The Gems era taught an important lesson: custom instructions only pay off when using them is easier than retyping them. Skills finally pass that test. Build your first one today, and the next time you catch yourself pasting the same long prompt, you will have a better option one keystroke away.