Image: Google (official) — "Skills in Gemini" header from the Google Keyword blog (Sep 30, 2026)
Most people still use Gemini the way they use a search box: ask, read, close the tab. Gemini Tasks is Google's attempt to change that habit. Instead of answering a single question, Gemini takes on a goal, works through it in the background, and can come back on a schedule or when something happens in your inbox. Google introduced the feature as Gemini Spark at I/O in May 2026, and you will still see that name in marketing and in press coverage. Inside the app, though, Google has since renamed it Task, because, as the Help Center puts it, "tasks" was the word people kept using. Same feature, new label.
This guide walks through the parts that matter in practice: what you need before you start, how to create a first task, how skills and schedules fit together, and where you should stay hands-on.
Before you start: requirements
Tasks are not available to everyone yet. According to Google's Help Center, you need all of the following:
- You are 18 or older.
- You sign in with a personal Google Account. Work and school accounts are not supported for now.
- You have a Google AI Pro or Ultra subscription.
- Keep Activity is turned on.
- You have opted in to Personal Intelligence.
Two more limits are easy to miss. Tasks are not offered in the European Economic Area, Nigeria, Switzerland, or the United Kingdom. And they only work in the Gemini mobile app, the Gemini app on Mac, and the web app at gemini.google.com, so a Gemini panel inside some other product will not show the option.
Create your first task
Ignore older guides that send you to a separate Spark menu. In the current interface, Task sits right next to Chat:
- Open gemini.google.com on your computer (or the Gemini app on a Mac).
- Below the text box, click Task.
- Describe what you want done. You can also say when it should run, such as at a certain time or after an event.
- To use a skill you have already saved, type @ and pick it from the list.
- To give Gemini material to work from, click Upload & tools and choose Upload files, Add from Drive, or Add Notebook.
- If you want, change the model or thinking level from the model menu at the top.
- Click Submit, then check back. If Gemini needs you to do something in its browser, it asks you to take over.
If you are not sure where to begin, Google suggests typing "set up", "get started", or "interview me" in a task, and Gemini will help you build your first skills and tasks through conversation. Each task gets its own thread in the sidebar. Clicking the progress chip at the top of that thread opens a work panel with completed, current, and planned steps, which is the quickest way to see whether a job is actually moving.
Skills: write your instructions once
Google describes the three building blocks as what, when, and how. The task is the goal, the schedule is the trigger, and a skill is the reusable "how": your instructions plus any reference files. Skills are also replacing Gems. Google says Gems support for personal accounts ends starting in November, and existing Gems will be migrated to skills automatically.
To create one, open Customize in the sidebar (click More first if you don't see it), then select Skills at the top. From there you can:
- build a skill with Gemini's help,
- edit a prefilled template or start from a blank one, or
- upload a
SKILL.mdfile, or a folder or .zip withSKILL.mdin its main folder.
When uploading, keep the skill name inside SKILL.md lowercase with hyphens (for example weekly-report), use plain-text reference files, and note that scripts needing internet access are not supported. In a chat or task, type / (or @ with the newer shortcut) to call a skill. You can stack several skills on one job, and activated skills can be picked up automatically when Gemini thinks they fit.
My take: skills are where the real time savings come from. A vague task prompt gets you a generic result every time. A skill that spells out your format, tone, and file locations turns the same prompt into something you can rely on.
Schedules: put it on autopilot
A schedule tells Gemini when to run a task without you asking again. The Help Center lists three types:
- Time-based: once, hourly, daily, weekly, monthly, or yearly. Google's own example is a 7 AM digest that looks through your email, calendar, and Drive and tells you what to prioritize.
- Gmail monitors: run when an email matching a Gmail filter arrives, for example an action item from your manager that Gemini helps answer with a draft reply.
- Topic monitors: watch news, finance, sports, local events, and similar topics, and react when something new happens.
The simplest way to create one is to describe it in a Task. For a time-based schedule you can also click Scheduled in the sidebar, then Create a scheduled task, and fill in a name, a plan, and a run time. Use More → Run now to test it before trusting it. Keep in mind that run times are approximate, and Google itself says monitors are not suited to time-critical jobs like grabbing tickets the moment they go on sale.
Browsers, takeovers, and safety
Some tasks need a browser. On desktop, Gemini can drive your local Chrome (Chrome auto browse), which has access to the sites you are already signed into, but your computer and Chrome must stay awake. Otherwise Gemini uses a remote browser in the cloud, which keeps going after you close your laptop. On mobile it uses only the remote browser. Chrome asks permission the first time, and Gemini asks for confirmation before each browsing task.
Google's safety advice is worth following exactly:
- Never type passwords, payment details, or other sensitive data into the task thread. Take over the browser and enter them on the site yourself.
- Avoid scheduling sensitive tasks, especially ones that could run while you are offline and unable to stop them.
- You can stop browsing from the Gemini in Chrome side panel, or cancel the whole task with Stop in the text box.
Google also flags prompt injection: a web page, email, or file can hide instructions aimed at the agent. Treat anything that reads untrusted email or websites as something to supervise, not something to forget about.
A practical first afternoon
If I were starting today, I would keep it low-risk. First, create one skill describing how you like summaries written. Second, set a daily morning digest of email and calendar using that skill, and run it once with Run now. Third, try one Gmail monitor on a narrow filter where Gemini only drafts and never sends, similar to the photography-inquiry tracker example Engadget highlights from Google's Spark page. After a week, keep what saved time and delete the rest.
Honest limits
Tasks are explicitly experimental. You can run up to 15 tasks at once and keep up to 50 active schedules, and heavy use counts against your plan's compute-based limits, so a schedule may skip a run if you hit them. Regional and account restrictions rule out many readers, including anyone on a work account. And the agent can still make mistakes, so for now it works best as a careful assistant you check on, not an employee you leave alone.
Sources
- Google Help: Manage your tasks & workflows in Gemini Apps
- Google Help: Find & manage tasks
- Google Help: Create & manage skills for Gemini Apps
- Google Help: Create & manage schedules for tasks in Gemini
- Google Blog: Let skills in Gemini tackle your most repetitive tasks (Sep 30, 2026)
- Google Blog: The Gemini app becomes more agentic (May 19, 2026)
- Gemini Spark product overview
- Engadget: How to use Google Gemini to automate tasks (Oct 12, 2026)
댓글
댓글 쓰기