These 45 prompts are for people using Google Gemini inside Google's own products: the Gemini app, Gmail, Docs, Sheets, Slides, Canvas, Gemini Live, Gemini's image tools, and Gems and skills. Most are built around a step that Google's help pages describe, such as selecting Gmail, Calendar or Drive with @ so Gemini can read them, condensing a thread in Gmail's side panel, typing =AI() in a Sheets cell to tag or summarise a whole column, planning a Slides deck from Drive sources, editing an image in the Library, or sharing your camera with Live. Most notes name the Google help page the feature comes from.
They are grouped by where you use Gemini, in eight sections. Many ask Gemini to quote its source, to say 'not stated' instead of guessing, and to show you a proposal before it creates an event, task or file share. That matters because Google's help pages say Gemini can give inaccurate or outdated answers and cannot see everything in your account, for example images in Gmail or pictures and videos in Drive.
The pack does not include everyday starter prompts that work the same in any chatbot. Gemini's features, plans and limits change often, so the prompts send you to Google's pages instead of quoting limits. The one exception is the Gems-to-skills migration prompt, which repeats Google's published dates and asks you to check them first. Deep research has its own pack, so it is left out here.
Gemini App With Gmail, Calendar, Drive, Keep and Tasks
These prompts run in the Gemini app and use its connection to Gmail, Calendar, Drive, Tasks and Keep. Google's connection page says to type @ and select the service, to use keywords instead of links, and to mention the service in the first prompt of a chat. The prompts ask Gemini to show what it found and to say what it could not see, because Google lists things it cannot do, such as read images in Gmail or pictures in Drive, or count items. The connection needs Keep Activity on, and work or school accounts need an administrator to allow it.
Best for: Gathering trip or event details scattered across the inbox
Type @ and select Gmail, then search my inbox for confirmations related to {{TRIP_OR_EVENT}}: reservations, flights, bookings or parcels, from {{DATE_RANGE}}.
Return a table with the columns: Item | Provider | Date and time | Confirmation reference | Email subject and sender.
Rules:
- Include only items you can see in an email. Leave a cell blank rather than filling it from general knowledge.
- Do not give totals or say how many emails match. If asked for a count later, say you cannot reliably give one.
- After the table, list what a complete plan would normally need that you did NOT find, for example a return leg or a hotel.
- Put each email subject in quotes so I can search for it myself.
Fill in: TRIP_OR_EVENT DATE_RANGE
Google's Gmail help says Gemini can find information on reservations, flights and packages, and its connection page says it cannot count items in Gmail, so this asks for a table, not a number.
Best for: A five-minute briefing before any meeting
Type @ and select Google Calendar, and find my meeting called {{MEETING_TITLE}}. Then type @ and select Gmail and @ Google Drive, and look for recent emails and documents about {{MEETING_TOPIC}}.
Prepare me for it:
1. The meeting details you can see for that event. If you cannot find the event, say so.
2. A summary of the relevant emails and documents, each labelled with where it came from (Gmail or Drive) and its subject or file name.
3. Three questions I should be ready to answer, and why.
4. Two questions I should ask, based on gaps in what you found.
5. What you could not see. Google says Gemini cannot read images or comments inside Docs and Gmail, so name any file where that might matter.
Keep it under 200 words. Do not invent attendees, numbers or history.
Fill in: MEETING_TITLE MEETING_TOPIC
Needs the Workspace connection and Keep Activity on. Google's Gmail help says Gemini only gives information about events on your primary calendar.
Best for: Answering a question-heavy email from a policy or price file
Type @ and select Gmail, and find the email from {{SENDER}} about {{EMAIL_TOPIC}}. Then type @ and select Google Drive, and find the file called {{DRIVE_FILE_NAME}}.
1. List each question or request in the email, quoting it.
2. Answer each one from the Drive file only: the answer in one line, and the exact words from the file it comes from, in quotes. If the file does not cover it, write "not in file" and do not answer from general knowledge.
3. Then write a reply of under 150 words in a {{TONE}} tone that uses only those answers, with a bracketed gap such as [NOT IN FILE: price for 2027] for each one the file does not cover.
If you cannot find the email or the file, say which and stop. Say so if the file is a type Gemini may not be able to read, such as a picture or video.
Fill in: SENDER EMAIL_TOPIC DRIVE_FILE_NAME TONE
Uses the Gemini app's @ connection to read Gmail and Drive in one prompt, as Google's connection page describes; Google's Gmail help also says Gemini can use Drive files as sources. Use keywords, not links. Check each quote against the file before you send.
Best for: A daily or weekly plan from your own calendar, tasks and inbox
Type @ and select Google Calendar, then @ Google Tasks, then @ Gmail. Plan {{PLANNING_PERIOD}} for me (for example "today" or "the week starting 9 November"). My working hours are {{WORKING_HOURS}}.
1. My calendar for that period, by day and in time order. Flag any day with less than two free hours inside my working hours.
2. My open tasks. Show a due date only where you can see one.
3. Emails from the last {{LOOKBACK}} that mention a deadline or meeting in that period: sender, subject, and the date as the email writes it.
4. The three outcomes that matter most, based only on what you found, with the meetings, tasks or emails that support each.
5. Where my two most important tasks could fit in my free time.
6. What you could not see, including whether you checked any calendar other than my primary one.
Do not add or move anything. Do not invent meetings, tasks or deadlines, and do not state a count of emails. If a service returned nothing, say so. End with one question if a priority is unclear.
Fill in: PLANNING_PERIOD WORKING_HOURS LOOKBACK
Based on Google's connection-page examples such as 'What's on my calendar today?'. Needs the Workspace connection with Keep Activity on; work and school accounts need an administrator to allow it. Google's Gmail help says Gemini only gives information about events on your primary calendar.
Best for: Finding follow-ups worth checking, without trusting a claim of silence
Type @ and select Gmail. Look for email threads from {{TIME_WINDOW}} in which I asked {{PERSON_OR_TEAM}} a question or made a request, and the most recent message you can see is mine.
For each thread, give: subject, the date I wrote, what I asked (one line), and a follow-up of no more than 40 words in a polite, direct tone.
Rules:
- Treat these as candidates only. You cannot prove that no reply exists, so do not say "no reply was sent". Say "the latest message I can see is mine".
- Put any thread you are unsure about in a separate "Unsure" list.
- Do not tell me how many threads there are in total.
- Do not add commitments on my behalf.
End with two search terms I can type into Gmail myself to check the list.
Fill in: TIME_WINDOW PERSON_OR_TEAM
Gmail search is keyword-based, and Google says Gemini cannot count Gmail items. Check each candidate in your inbox before you chase anyone.
Best for: Looking up a term in a Drive document without opening it
Type @ and select Google Drive. Find the document about {{DOCUMENT_DESCRIPTION}}; use keywords from its title or subject rather than a link.
Then answer these questions from that document only (skip any I leave blank):
1. {{QUESTION_ONE}}
2. {{QUESTION_TWO}}
3. {{QUESTION_THREE}}
For each answer give:
- The answer in one line.
- The exact words in the document it comes from, in quotation marks.
- "Not found in this document" if the document does not say.
First tell me which file you used (name and, if you can see it, who wrote it or when). If more than one file fits, list them and ask which I mean before answering.
Google says Gemini cannot read pictures or videos in Drive, so tell me if the best match is one of those. Do not guess a clause or amount.
Fill in: DOCUMENT_DESCRIPTION QUESTION_ONE QUESTION_TWO QUESTION_THREE
Google's connection page gives Drive examples such as finding a lease agreement and checking the security deposit. It says Gemini reads documents, spreadsheets, presentations and PDFs in Drive, not pictures or videos.
Best for: Clearing a noisy sender or school/club thread in minutes
Type @ and select Gmail. Summarise my recent emails from {{SENDER_OR_ORGANISATION}} about {{TOPIC}}.
Sort them into two groups:
- Needs action from me: for each, the subject, what is being asked, and any date written in the email.
- For information only: one line each.
Rules:
- Use keywords to find the emails and name the sender as it appears in the email.
- If an email is mostly images, or the action is in an attachment you cannot read, put it in a third group called "Check manually".
- Do not state how many emails there are in total.
- Put dates only where the email states them.
End with the one email I should open first and why.
Fill in: SENDER_OR_ORGANISATION TOPIC
Based on Google's connection-page example 'Summarize my recent emails from...'. Google lists images in Gmail as something Gemini cannot access, so the prompt makes it say so.
Best for: Turning one busy email into checked events and tasks
Type @ and select Gmail and find the email from {{SENDER}} about {{EMAIL_TOPIC}}. Then type @ and select Google Calendar and @ Google Tasks.
Read the email and sort everything it asks of me into two lists.
A. Calendar events (anything at a set time, or a date I must attend): the title; the date and time exactly as the email gives them; the location or link if stated; and any clash you can see on my primary calendar. If the email gives a date but no time, or a time but no date, say what is missing. If it offers several options, list them and let me choose.
B. Tasks (things I must do, with or without a deadline): a title under 10 words; the due date only if the email states one, otherwise "no date stated"; and the exact phrase from the email it comes from.
Do not create anything yet. Show me both lists and wait. When I reply "add these", or tell you which to drop, create the events in Google Calendar and the tasks in Google Tasks, then tell me the titles you created so I can check them. If you cannot create one, say so and give it to me as text. If you cannot find the email, list the keywords you used.
Fill in: SENDER EMAIL_TOPIC
Google's connection page shows asking Gemini to check an email and add it to your calendar, and creating a reminder in Tasks. This version shows both lists first because Gemini can misread dates. Check Calendar and Tasks afterwards.
Example output · Claude (example written by The Prompt Index)
**Inputs used:** SENDER = Ms Okafor (Year 5); EMAIL_TOPIC = Science Museum trip. The email found: "Thank you for volunteering as a parent helper. The Year 5 trip to the Science Museum is on Friday 13 November. Coaches leave school at 8:45 and return at about 15:15. Please return the signed helper form and the £18 contribution by Monday 9 November. There is a parent helpers' briefing on Wednesday 11 November in the school hall, time to be confirmed. Packed lunch needed on the day." Your primary calendar on Friday 13 November shows "Dentist 15:00–15:30".
**A. Calendar events**
1. **Year 5 Science Museum trip (parent helper)**: Friday 13 November, coaches leave school at 8:45 and return at about 15:15. Location: departs from school; the Science Museum. Clash: "Dentist 15:00–15:30" overlaps the return time of about 15:15.
2. **Parent helpers' briefing**: Wednesday 11 November, school hall. Missing: the time ("time to be confirmed").
**B. Tasks**
1. **Return signed helper form**: due Monday 9 November. From: "Please return the signed helper form and the £18 contribution by Monday 9 November."
2. **Pay £18 trip contribution**: due Monday 9 November. From the same sentence.
3. **Pack a lunch for the trip**: no date stated. From: "Packed lunch needed on the day." ("The day" reads as the trip day, but the email does not give a date in that sentence.)
Nothing has been created yet. Reply "add these", or tell me which to drop. You may also want to move the dentist appointment.
Best for: Growing a Keep list without duplicates
Type @ and select Google Keep. I keep a running list about {{LIST_TOPIC}}, for {{PURPOSE}}.
1. Search my Keep notes for an existing note about {{LIST_TOPIC}}. Tell me its title and list the items already in it. If you find more than one, list the titles and ask which I mean. If you find none, say so.
2. Suggest {{ITEM_COUNT}} new items that are not already in that note. For each: the name, one line on why it fits my purpose, and one thing to check before relying on it (price, opening hours, availability, or similar).
3. Save the new items to Google Keep as a note titled "{{NOTE_TITLE}}", and tell me the title you saved it under so I can find it.
Rules:
- Do not state prices, hours or availability as facts; list them under "check before relying on it".
- Do not change or delete my existing note.
- If you could not save to Keep, say so and give me the list to paste.
Fill in: LIST_TOPIC PURPOSE ITEM_COUNT NOTE_TITLE
Google's connection page lists Keep among the services Gemini can use and shows saving a list to Keep. Type @ and select Keep, then check that the new note appears in Keep.
Best for: Catching a mismatch between what you were told and what is on file
Type @ and select Gmail, and find the email from {{SENDER}} about {{EMAIL_TOPIC}}. Then type @ and select Google Drive, and find the file about {{DRIVE_DOCUMENT_TOPIC}}.
Compare what the email says with what the Drive file says on {{COMPARISON_POINTS}}.
Output a table: Point | What the email says (quote) | What the file says (quote) | Match, conflict, or only in one.
Rules:
- Quote both sides; do not paraphrase a number or a date.
- If either source cannot be found, say which one, and stop instead of comparing against memory.
- Google says Gemini cannot read images in Gmail or pictures in Drive, so say if the point may be in an image.
- Finish with the conflicts ranked by how much they would matter if I acted on the wrong one, and the question I should ask the sender.
Fill in: SENDER EMAIL_TOPIC DRIVE_DOCUMENT_TOPIC COMPARISON_POINTS
Needs the Workspace connection; type @ and pick each service as Google's connection page says. Check any conflict against the original email and file.
Best for: Sorting scattered notes into an existing document's structure
Type @ and select Google Keep, and find my notes about {{PROJECT_TOPIC}}. Then type @ and select Google Drive, and find the document called {{DOC_SEARCH_TERM}}.
Turn my Keep notes into an outline that fits the document's structure, so I can paste it into that Doc.
Output:
1. The headings from the Drive document, in its order, as you can see them.
2. Under each heading, my Keep notes that belong there, quoted with the note title.
3. A list of Keep notes that fit no heading.
4. A list of headings that have no notes, so I can see the gaps.
Rules: do not add content that is not in my notes; if you cannot see the Drive document's headings, say so and propose headings, marked "proposed"; if a note is a picture or checklist you cannot read, put it under "Check manually".
Fill in: PROJECT_TOPIC DOC_SEARCH_TERM
Google's connection page lists Keep and Drive among the connectable services. Type @ and select each service; check the outline against your notes before you paste it.
Gmail: Side Panel and Help Me Write
For Gemini inside Gmail on a computer. The side-panel prompts follow Google's Gmail help: condensing a long thread, acting on Drive files you pick with @ after you review the suggested action, and creating an event on your primary calendar. The Help me write prompts name the recipient, action and tone, and use the Sources link to check what a draft drew on. These need an eligible Google Workspace or Google AI plan, and personal-account availability is limited, so check Google's pages.
Best for: Getting on top of a long thread before you reply
(In Gmail on a computer, open the thread, then click Ask Gemini at the top right. Type this in the prompt box at the bottom of the side panel.)
I am {{MY_ROLE}} on this thread and I need to act on it by {{MY_DEADLINE}}. Condense this email thread into concise bullet points for me.
Output, in this order:
1. Where things stand now, in two sentences.
2. Decisions already made: who made each one, with the sender's exact words in quotes.
3. Questions still open, each with the person expected to answer.
4. Action items for me only, each starting with a verb, with any date exactly as the thread writes it.
5. Anything that may sit in an attachment or image you could not read.
Use only this thread. If two messages disagree, show both and say which is more recent. Write "not stated" instead of guessing an owner or a date.
Fill in: MY_ROLE MY_DEADLINE
Google's Gmail help says Gemini in the side panel can condense long threads into concise bullet points, and gives 'Create a list of action items for me based on this email' as an example. It needs an eligible Google Workspace or Google AI plan; personal accounts can try it through Workspace Experiments.
Best for: Sending the right file with the right access, from inside Gmail
(In Gmail on a computer, open the email that asks for a file, then click Ask Gemini at the top right. In the prompt, type @ and pick the Drive file from the menu where the variable is.)
This email asks me to share a file. The file is {{DRIVE_FILE_NAME}}.
1. Tell me in one line what the email asks for and who should get it, quoting the request.
2. Propose these file actions, but do not do them yet: make a copy named {{COPY_NAME}}, then share the copy with {{RECIPIENT}} as {{ACCESS_LEVEL}}. Never share the original.
3. Point out anything in the email that suggests a different access level or a different file, so I can check before confirming.
4. Draft a reply of under 80 words saying the copy is shared and what access they have. Do not mention anything that is not in this email.
I will review the suggested file actions and confirm them myself.
Fill in: DRIVE_FILE_NAME COPY_NAME RECIPIENT ACCESS_LEVEL
Google's Gmail help says you can type @ to reference Drive files in the side panel, that Gemini can move, rename, share and copy Drive files (its example: 'Share the @Q3 Marketing Plan with @dougm as commenter'), and that you should review suggested file actions before confirming. Needs an eligible Google Workspace or Google AI plan.
Best for: Turning an agreed meeting into a calendar event without leaving Gmail
(In Gmail on a computer, open the thread where the meeting was agreed, click Ask Gemini at the top right, and type this in the side panel.)
This thread agrees to a meeting about {{MEETING_PURPOSE}}. Before creating anything:
1. Quote the message that sets the date or time window, and list the people on the thread who should attend.
2. Check my primary calendar in that window and suggest up to three free {{DURATION}} slots for me. Do not say whether the other attendees are free.
3. Propose the event: title, the slot I choose, attendees, and a two-line description naming the decision the meeting is for.
Wait for me to pick a slot and reply "create it". Then create the event and tell me its title and time so I can check it in Calendar. If the thread does not agree a time window, say so and draft a short reply asking for one instead.
Fill in: MEETING_PURPOSE DURATION
Google's Gmail help gives 'Create a 30 minute meeting with Kevin this week' as a side-panel example, and says Gemini only gives information about, and creates events on, your primary calendar. Needs an eligible Google Workspace or Google AI plan.
Best for: A grounded first draft in Gmail's own writing bar
Draft an email in Gmail's Help me write bar.
Recipient: {{RECIPIENT}}
Subject: {{SUBJECT}}
What I need them to do, and by when: {{REQUIRED_ACTION}}
Tone: {{TONE}}
Facts to use: {{SOURCE_FACTS}}
If I also reference an email or file by typing @ and picking it, you may use facts from that source too. Use no other facts.
Write it as a normal email: a subject line that matches my subject, a short opening that gets to the point, the request with its deadline, and a polite close. Keep it under 150 words and make it sound like a real person.
If a fact the email needs is not in my facts or the referenced source, put a bracketed gap such as [ADD: room number] instead of inventing it. Do not add a sign-off name.
Fill in: RECIPIENT SUBJECT REQUIRED_ACTION TONE SOURCE_FACTS
Use this in the Help me write bar in Gmail compose. Google's Gmail help says to name the recipient, subject and required action, add a tone word, and reference a specific email or file; click Sources to see what Gemini drew on. Personal-account availability is limited, so check Google's page.
Example output · Claude (example written by The Prompt Index)
**Inputs used:** RECIPIENT = Priya Nair; SUBJECT = Venue change for Thursday's workshop; REQUIRED_ACTION = confirm whether the new room works for her team, by Tuesday at 5 pm; TONE = friendly; SOURCE_FACTS = "Room 4B on level 2 replaces Room 2A. Capacity 20. Start time unchanged at 10:00."
**Subject:** Venue change for Thursday's workshop
Hi Priya,
Quick heads-up: the venue has moved Thursday's workshop from Room 2A to Room 4B on level 2. The room holds 20 people, and the start time is unchanged at 10:00.
Could you confirm by Tuesday at 5 pm whether the new room works for your team?
Thanks so much
Best for: A personalised draft whose borrowed details you can verify
Draft a reply to the email I have open. My aim: {{REPLY_AIM}}. Tone: {{TONE}}. Keep it under 150 words.
You may draw on my other emails and Drive files to personalise it, but only where they bear directly on this reply, such as a date, a figure or something we already agreed.
Rules:
- Put [CHECK] straight after any sentence that uses a detail from another email or file, so I can confirm it against Sources before sending.
- Do not bring in personal or confidential details from other threads unless the reply needs them.
- If something the reply needs is not available, write [ADD: what is missing] instead of guessing.
- Write the reply only, with no notes, headings or explanations, because everything you write goes into the email body.
Fill in: REPLY_AIM TONE
Type this in Gmail's Help me write bar while replying. Google's Gmail help says Help me write can use your other emails and Drive files and that you can click Sources to see which ones it used. Compare each [CHECK] sentence with Sources, then delete the markers. Personal-account availability is limited, so check Google's page.
Docs and Slides
For Gemini inside Google Docs and Slides. The prompts use features Google documents: summarising the comments on a draft, drafting from Drive files picked with @, applying a writing style taken from a Drive file, building one slide or a whole planned deck from your sources, and Help me visualize for slide images. Several are desktop only or need an eligible plan or Workspace Experiments, which the notes say.
Best for: Working through a heavily commented draft
(In Google Docs, open the draft that has comments, click Ask Gemini at the top right, and type this.)
I wrote this draft for {{AUDIENCE}} and must finalise it by {{DEADLINE}}. My goal for the draft: {{DRAFT_GOAL}}. Review and summarise the comments on it.
1. Group the comments by theme. For each theme give the commenters' names, a one-line summary, and the part of the draft the comments are attached to, quoted briefly.
2. Mark where two commenters ask for opposite changes, and quote both.
3. For each theme, recommend accept, push back or ask a question, with one line of reasoning tied to my goal.
4. List the comments that need a reply from me rather than a text change.
Do not edit the document or resolve any comment. If you cannot see the comments, tell me so instead of summarising the text.
Fill in: AUDIENCE DEADLINE DRAFT_GOAL
Google's Workspace Experiments page lists 'Review and summarize comments in a draft with the help of Gemini' among Docs features and says to click Ask Gemini at the top right of Docs. Google's connection page says the Gemini app cannot read comments in Docs, so run this inside Docs. Google lists it under Workspace Experiments, a trusted tester program for personal accounts aged 18 and over, and says those features roll out gradually and may not be available yet.
Best for: Combining two source documents without hidden choices
Type @ and select {{FIRST_FILE_NAME}}, then type @ and select {{SECOND_FILE_NAME}}, so both are attached to this prompt.
Using only those two files, write a {{DOCUMENT_TYPE}} for {{AUDIENCE}}.
Rules:
- When the two files disagree, do not choose. Put the disagreement under a heading "Conflicts to resolve" with a quote from each file.
- Mark every statement with the file it came from, as (File 1) or (File 2).
- Where neither file gives a fact the document needs, leave [ADD: ...].
- Keep it to {{LENGTH}}.
After the draft, list anything in either file that you chose to leave out and why.
Fill in: FIRST_FILE_NAME SECOND_FILE_NAME DOCUMENT_TYPE AUDIENCE LENGTH
Google's Docs help says to type @ and search and select from the menu. Treat each file-name variable as a search term; select the matching file from the menu so it attaches as a chip. Click Sources, if your screen shows it, to see which files Gemini used.
Best for: Writing in your own house style inside Docs
(In Google Docs: open the bottom bar, click Tools, choose Writing style, then Add doc from Drive, pick the document whose style you want, click Add, and confirm the style summary Gemini shows. Then use this prompt.)
Draft a {{DOCUMENT_TYPE}} about {{TOPIC}} for {{AUDIENCE}}, in the writing style I just added.
Facts to use, and no others: {{FACTS_TO_USE}}
Before drafting, tell me in three bullets the style traits you are applying (tone, sentence length, vocabulary) so I can check that they match the summary I confirmed. Then write the draft. If a fact you need is missing, put [ADD: ...] instead of making one up.
Finish with two sentences on where the draft departs from the style, and why.
Fill in: DOCUMENT_TYPE TOPIC AUDIENCE FACTS_TO_USE
Google's page on personalised documents says to use Tools > Writing style > Add doc from Drive, then confirm the style summary. It says the option needs an eligible Google Workspace or Google AI plan, or Workspace Experiments, and is on desktop only.
Best for: A sourced slide built from your own files
Create a slide about {{SLIDE_TOPIC}} for {{AUDIENCE}}.
Use these files as your only sources: type @ and select {{SOURCE_FILE_ONE}}, then @ and select {{SOURCE_FILE_TWO}}.
The slide needs:
- A headline that states the point, not a label.
- No more than four bullets, each under 12 words.
- Speaker notes I can read aloud in about 30 seconds.
Rules:
- Every number or claim must come from the files. If you cannot find support, leave it out and tell me.
- After the slide, list which file each claim came from so I can use Sources to check.
If the files do not contain enough to make a slide about {{SLIDE_TOPIC}}, say what is missing.
Fill in: SLIDE_TOPIC AUDIENCE SOURCE_FILE_ONE SOURCE_FILE_TWO
Google's Slides help shows prompts like 'Create a slide outlining the next steps from @meeting notes', and says Sources shows which files a response used. Treat each file variable as a search term and select the file from the @ menu. New slides use your current theme.
Best for: A slide image that matches what you described, with room for text
(In Google Slides, click the slide, then Insert > Help me visualize > Image. Type the paragraph below in the right panel, pick a style or aspect ratio if you want one, and click Create. Preview the image, then click Insert.)
A {{SHOT_DISTANCE}} photograph of {{SUBJECT}}, made of {{MATERIALS}}, in {{SETTING}}, with {{BACKGROUND}} in the background. Even, natural light. A plain composition with empty space on the {{EMPTY_SIDE}} side for slide text. No text, letters, logos or people.
(To change one thing later, click the image, choose Edit image, and type one literal change, for example "Make the background a plain light grey wall". Then choose Replace with image, or Insert as background.)
Fill in: SHOT_DISTANCE SUBJECT MATERIALS SETTING BACKGROUND EMPTY_SIDE
Google's Slides image help says to use Insert > Help me visualize > Image, include the subject, setting, distance to the subject, materials and background, and avoid figurative language; Edit image offers Replace with image or Insert as background. Google says generated images are for use only within Google Slides and Google Vids, and the feature needs an eligible Google Workspace or Google AI plan.
Best for: A sourced deck built from an outline you approve
(In Google Slides on a computer, open a blank presentation and click Ask Gemini. Use Add sources > Add from Drive to attach your reference files. Keep the Slides tab open while Gemini generates.)
Create a {{SLIDE_COUNT}}-slide presentation on {{TOPIC}} for {{AUDIENCE}}. The decision or action I want from them is: {{DESIRED_ACTION}}.
Use only the sources I added for facts. Before building, show me the plan Gemini proposes: for each slide a title and a one-line description, and which source each fact will come from. I will edit the plan and approve it myself.
Rules:
- Tell me which slides have no source support.
- Do not invent statistics or quotes.
- After the deck is built, I will ask you to refine one slide at a time.
If you ask follow-up questions, keep them to the ones that would change the deck.
Fill in: SLIDE_COUNT TOPIC AUDIENCE DESIRED_ACTION
Google's Slides help for generating a presentation describes Ask Gemini, Add sources > Add from Drive, a reviewable plan you approve, and refining slide by slide. It needs an eligible plan or Workspace Experiments, is desktop only and English only, and the Slides tab must stay open.
Sheets: Ask Gemini and the AI Function
Two ways to use Gemini in Sheets. The Ask Gemini panel summarises a sheet or fills a range on the tabs you pick from its dropdown. The AI function, typed as =AI() in a cell and filled down a column, tags, extracts, summarises or drafts text row by row, and can pull current facts from the web with a Sources link. Google says the function sees only the range you give it, generates only the first 350 selected cells at a time and has usage limits, so check its page. Both need an eligible Google Workspace or Google AI plan, or Workspace Experiments.
Best for: A sourced summary you insert into the sheet before it disappears
(In Google Sheets, click Ask Gemini at the top right. Use the tab dropdown in the prompt box to include the tab or tabs you want, then paste this.)
Summarise this spreadsheet for {{AUDIENCE}}, so they can decide {{DECISION}}.
Give me:
1. What the sheet is for, in two sentences.
2. The five most useful numbers or patterns, each with the cells or columns it comes from.
3. Anything odd you can see: blanks, duplicates, values that look out of range, mixed formats.
4. A short chart suggestion, such as a chart with {{X_AXIS}} on the x-axis and {{Y_AXIS}} on the y-axis, and what it would show.
Format the summary as short lines that fit one column of cells, so I can insert it into the sheet. For the chart suggestion, name the cells it would use.
Use only the included tabs. If you could not see something, say so.
Fill in: AUDIENCE DECISION X_AXIS Y_AXIS
Insert the summary before you reload or close the sheet: Google's Sheets help says conversation history is lost on reload, close or going offline, and that charts do not update when the data changes. Google gives 'Summarize this spreadsheet' as an example and says to choose tabs from the dropdown. Needs an eligible Google Workspace or Google AI plan.
Best for: A safe, checked range fill
(In Google Sheets, click Ask Gemini at the top right.)
In the tab dropdown in this prompt box I have selected only the tab {{TAB_NAME}}; work on that tab only. Fill a specified range of cells on it: {{RANGE_TO_FILL}}.
What to put in it: {{FILL_RULE}}
Before you change anything:
1. Repeat back the exact range and what you will put in each cell, with the first three and last three cells as examples.
2. Tell me if the range already contains data you would overwrite, and which cells.
3. Wait for me to say "fill it".
After filling, tell me the first and last cell you changed so I can check the edges.
Rules: do not touch any cell outside the range; if you cannot do it as asked, say what you can do instead.
Fill in: TAB_NAME RANGE_TO_FILL FILL_RULE
Google's Sheets help describes filling a specified range of cells with a single value or a repeating pattern. Give one value or one pattern (for example "Pending" or "Mon, Wed, Fri repeating"), and check the edge cells after it fills.
Best for: Tagging a column of comments or tickets, 350 cells per pass
(In Google Sheets, enter the formula below in the first empty cell next to your data, then fill it down. Select the filled cells and click Generate and Insert. Check a sample of rows before you rely on the column. Google says "only the first 350 selected cells with AI functions will be generated", so for a longer column, select and generate up to 350 cells at a time; check Google's page for the current limit.)
=AI("You are sorting {{ITEM_TYPE}} for {{PURPOSE}}. Classify the text in the given cell into exactly one of these categories: {{CATEGORY_LIST}}. If more than one fits, choose the one the text spends the most words on. If none fits, use Other. Also give its sentiment as Positive, Neutral or Negative. If the cell is empty, output Blank. Output only the category and the sentiment, separated by a comma, with no other words.", {{FIRST_CELL}})
Fill in: ITEM_TYPE PURPOSE CATEGORY_LIST FIRST_CELL
Google's AI function help: type =AI() or =Gemini() in a cell, or use Insert > Function > AI; the form is AI("prompt", range). Google says it can categorise data, including sentiment analysis, and recommends stating the exact output format. It sees only the range you give it, cannot be undone (regenerate instead), and has per-selection and usage limits, so check Google's page. Needs an eligible Google Workspace or Google AI plan.
Example output · Claude (example written by The Prompt Index)
**Inputs used:** ITEM_TYPE = customer comments; PURPOSE = a café's weekly review; CATEGORY_LIST = Food, Service, Price, Cleanliness; FIRST_CELL = B2.
**Formula entered in C2 and filled down to C7:**
`=AI("You are sorting customer comments for a café's weekly review. Classify the text in the given cell into exactly one of these categories: Food, Service, Price, Cleanliness. If more than one fits, choose the one the text spends the most words on. If none fits, use Other. Also give its sentiment as Positive, Neutral or Negative. If the cell is empty, output Blank. Output only the category and the sentiment, separated by a comma, with no other words.", B2)`
| Row | Comment (column B) | Output (column C) |
|---|---|---|
| 2 | Lovely flat white and the cinnamon bun was still warm. | Food, Positive |
| 3 | £4.20 for a small tea is too much. | Price, Negative |
| 4 | Tables were sticky and the bin by the door was overflowing. | Cleanliness, Negative |
| 5 | (empty) | Blank |
| 6 | Parking nearby is hard to find. | Other, Negative |
| 7 | Staff were friendly, though the soup we ordered was only lukewarm. | Food, Neutral |
**Why row 7 is Food:** "Staff were friendly" is 3 words about service; "though the soup we ordered was only lukewarm" is 8 words about food, so the most-words rule picks Food. The praise and the complaint balance out, so the sentiment is Neutral.
**Check before relying on it:** rows 6 and 7 are the judgement calls, so read them yourself.
Best for: Getting dates, codes or amounts out of free-text cells
(Enter the formula in the first empty cell of a new column and fill it down, then select the cells and click Generate and Insert. The function only sees the cell you give it. Google says "only the first 350 selected cells with AI functions will be generated", so for a longer column, select and generate up to 350 cells at a time; check Google's page for the current limit.)
=AI("From the text in the given cell, extract the {{FIELD_TO_EXTRACT}}. Output it in exactly this format: {{OUTPUT_FORMAT}}, for example {{FORMAT_EXAMPLE}}. If the text contains more than one, output the first one only. If it contains none, output NONE. Do not guess, correct or complete a value that is not written in the text. Only output the value.", {{FIRST_CELL}})
Fill in: FIELD_TO_EXTRACT OUTPUT_FORMAT FORMAT_EXAMPLE FIRST_CELL
Google's AI function help: type =AI() or =Gemini() in a cell, or use Insert > Function > AI; the form is AI("prompt", range). Google's tips say to describe the exact task and format, for example 'Only output the date', and to put the data you want analysed in the range. Filter for NONE afterwards and check those rows by hand. Needs an eligible Google Workspace or Google AI plan.
Best for: Scanning a column of long notes, feedback or descriptions
(In a table, at the top right of the last column, click Insert AI column right. Enter the formula in the new column's first cell and fill it down. Select the cells and click Generate and Insert. Google says "only the first 350 selected cells with AI functions will be generated", so for a longer column, select and generate up to 350 cells at a time; check Google's page for the current limit.)
=AI("Summarise the {{TEXT_TYPE}} in the given cell for {{READER}} in one sentence of no more than {{MAX_WORDS}} words. Start with the main point or request, then add one key detail, such as a date, amount or name, only if the text states it. Use plain words and no quotation marks. If the cell is empty or too short to summarise, output the original text unchanged.", {{FIRST_CELL}})
Fill in: TEXT_TYPE READER MAX_WORDS FIRST_CELL
Google's AI function help: type =AI() or =Gemini() in a cell, or use Insert > Function > AI; the form is AI("prompt", range). Google says it can summarise information and extract key takeaways, and says to add an AI column by clicking Insert AI column right at the top right of the last column in a table. Use Refresh and insert after the source text changes. Needs an eligible Google Workspace or Google AI plan.
Best for: Product blurbs, listings or short descriptions at scale
(Put each row's facts in adjacent columns, enter the formula in the first empty cell of a new column, and fill it down. Select the cells and click Generate and Insert. Edit any row before you publish it. Google says "only the first 350 selected cells with AI functions will be generated", so for a longer column, select and generate up to 350 cells at a time; check Google's page for the current limit.)
=AI("Using only the facts in the given cells, which are {{COLUMN_MEANINGS}} in that order, write a {{TEXT_TYPE}} of no more than {{MAX_WORDS}} words for {{AUDIENCE}}. Do not add features, numbers or claims that are not in the cells. If a cell is empty, leave that fact out instead of inventing it. Output only the text.", {{FIRST_ROW_RANGE}})
Fill in: COLUMN_MEANINGS TEXT_TYPE MAX_WORDS AUDIENCE FIRST_ROW_RANGE
Google's AI function help: type =AI() or =Gemini() in a cell, or use Insert > Function > AI; the form is AI("prompt", range). Google's example is =AI("Generate slogan for event in 10 words or less", A2). It says the function cannot see your whole spreadsheet or other Drive files, so put the facts in the range (for example A2:C2). Needs an eligible Google Workspace or Google AI plan.
Best for: Filling a column with web facts you can trace
(Enter the formula in the first empty cell next to your list and fill it down. Select the cells and click Generate and Insert. Then a person should click into the CHECK rows and a sample of the others, hover over the result, and click Sources to judge whether the pages support the value. Google says "only the first 350 selected cells with AI functions will be generated", so for a longer column, select and generate up to 350 cells at a time; check Google's page for the current limit.)
=AI("Find the current {{FACT_NEEDED}} for the {{ITEM_TYPE}} named in the given cell, using up-to-date information from the web. Output only the value, in this format: {{OUTPUT_FORMAT}}. If you cannot find a value for this exact {{ITEM_TYPE}}, output CHECK instead of a value; do not give a value for a different {{ITEM_TYPE}} with a similar name.", {{FIRST_CELL}})
Fill in: FACT_NEEDED ITEM_TYPE OUTPUT_FORMAT FIRST_CELL
Google's AI function help: type =AI() or =Gemini() in a cell, or use Insert > Function > AI; the form is AI("prompt", range). Google says it can pull up-to-date answers from Google Search and that you can hover over a cell and click Sources to see the web sources used. CHECK only marks rows where no value came back; the function does not judge whether sources agree, so a person reviews Sources for a sample of rows and fills the CHECK rows by hand. Needs an eligible Google Workspace or Google AI plan.
Image Creation and Editing
Gemini can create images from a prompt, edit an image you upload or generate, and combine several uploaded images, and Gemini in Chrome can edit an image from a web page. These prompts follow Google's image help: start with a verb, describe the style, give a detailed visual description, and use the Library and Chat edit flow, Redo with Pro and Export to Docs. Google says some of this depends on the image model and on your plan and age, and the notes quote what it says.
Best for: A first image prompt that follows Google's own tips
(In the Gemini app, open Images in the sidebar or type in a chat. Fill in the template and send it as it is.)
Generate a {{STYLE}} image of {{SUBJECT}} in {{SETTING}}. In the foreground, {{FOREGROUND_DETAIL}}. Behind it, {{BACKGROUND_DETAIL}}. The light and colours are {{LIGHT_AND_COLOUR}}, and the viewer's eye goes first to the main subject. Keep the composition simple, with one clear subject. Do not include any text, letters, logos, watermarks or real people's faces.
After the image, add one line naming any part of this description you could not show. I will then fix the image one local edit at a time: open it from Library, click Chat under it, and ask for a single change, such as "Make the light warmer and keep everything else the same."
Fill in: STYLE SUBJECT SETTING FOREGROUND_DETAIL BACKGROUND_DETAIL LIGHT_AND_COLOUR
Google's image help says to start prompts with words like draw, generate and create, describe the image style, and include a detailed visual description. Google's page sets age and daily limits and says images that break its policies may be removed; check it for your account. Google's page also describes editing a generated image from the sidebar: Library, select the image, then Chat.
Best for: Fixing one thing in a photo without losing the rest
(In the Gemini app, start a new chat, upload the photo from your device, then send this.)
Edit this photo. Make one change only: {{ONE_CHANGE}}.
Keep these exactly as they are: {{WHAT_TO_KEEP}}. Do not change faces, body shapes, clothing, the camera angle, the framing or the overall lighting unless that is the change I asked for. Match the new part to the photo's existing light direction, colour and grain so it does not look pasted in.
Return the edited image. Under it, write one line saying what you changed and one line naming anything else that changed that I did not ask for.
Fill in: ONE_CHANGE WHAT_TO_KEEP
Google's image help says you can upload an image from your device and ask Gemini to make edits, and that editing images is not available to users under 18. Google's page also says Nano Banana 2 Lite "isn't designed for multiple image references or for multiple edits", so if repeated edits drift, check the page for the image model your account uses. Keep your original file.
Best for: Mixing references into a single picture
I have uploaded two images. The first is {{FIRST_IMAGE}}. The second is {{SECOND_IMAGE}}. Create one new image based on these uploaded images.
Use from the first image: {{ELEMENT_FROM_FIRST}}.
Use from the second image: {{ELEMENT_FROM_SECOND}}.
The new image should be: {{NEW_IMAGE_DESCRIPTION}}
Before you generate, restate in one line per image what you will take from it. If two of my instructions conflict (for example both images supply the background), tell me and ask which wins.
Rules:
- Do not add objects that are in neither image or my description.
- If something from an image cannot be reused (for example text that would be unreadable), say so.
Then give me one follow-up prompt to fix the most likely flaw in a combined image.
Fill in: FIRST_IMAGE SECOND_IMAGE ELEMENT_FROM_FIRST ELEMENT_FROM_SECOND NEW_IMAGE_DESCRIPTION
Google's image help says you can upload multiple images and ask Gemini to create a new image based on them, but that this depends on the model: Nano Banana 2 "can accept multiple reference images", while Nano Banana 2 Lite "isn't designed for multiple image references or for multiple edits". If the result ignores one image, check Google's page for the model your account uses. Make sure you have the right to use the images.
Best for: Fixing a near-miss image without losing what worked
(In gemini.google.com on a computer: open the sidebar, click Library, select the image, then click Chat under it. Use this prompt in that edit box. If you are on a paid plan and want a different result altogether, click More > Redo with Pro on the original image instead.)
I want to fix this image. Here is what is wrong, in order of importance: {{PROBLEMS_RANKED}}.
Fix only the first problem in this turn, as a local edit: change only {{AREA_TO_CHANGE}} and keep everything else exactly as it is.
Before editing, say in one sentence what you will change and what you will leave alone. After editing, tell me what you changed. I will say "next" when I want the second problem.
Rules:
- One change per turn.
- If a fix would change the style or composition of the whole image, say so and ask me before doing it.
- If a problem looks like a limit of image generation (for example small text), say so.
Fill in: PROBLEMS_RANKED AREA_TO_CHANGE
Google's image help: open the sidebar > Library, select an image and click Chat under it to edit; Redo with Pro is under More for paid subscribers. Google lists local edits ('making quick and specific changes to parts of an image') as a Nano Banana 2 feature, says Nano Banana 2 Lite "isn't designed for multiple image references or for multiple edits", and says "if you reach your daily quota of Nano Banana 2 images, you can't redo any additional images with Nano Banana Pro." Check the page for your plan and age limits.
Best for: Editing an image from a web page, with a rights reminder
(In Chrome on a computer, with Gemini in Chrome turned on: right-click the image and choose Create image with Gemini. If the option does not appear, the image may be protected or unsupported. Then use this prompt.)
Edit this image for {{USE}}: {{EDIT_REQUEST}}.
Before you edit:
1. Tell me what you see in the image in two sentences.
2. Ask me one question if my request could be done in more than one way.
3. Remind me to check that I have the right to edit and use this image.
Rules: keep {{ELEMENTS_TO_KEEP}} unchanged; do not add text, logos or real people unless I name them: {{ALLOWED_ADDITIONS}}; if you cannot do the edit, say why.
After the edit, give me a one-line description of what changed for my records.
Fill in: USE EDIT_REQUEST ELEMENTS_TO_KEEP ALLOWED_ADDITIONS
Google's Chrome image page says to right-click an image and choose Create image with Gemini; it needs Gemini in Chrome, the latest Chrome, a signed-in profile (not Incognito), is for users 18 or over to edit, and is not available with work or school accounts for now. Rights to the image are your responsibility.
Best for: A document with matching images, not loose pictures
I am making a one-page handout on {{HANDOUT_TOPIC}} for {{AUDIENCE}}.
Step 1. Give me the handout text first, in sections of no more than {{MAX_WORDS_PER_SECTION}} words, with a heading for each. Mark where an image would help, and no more than {{IMAGE_COUNT}} images.
Step 2. For each image, write a prompt that starts with a verb such as "Generate", describes the style, and gives a detailed visual description. Use the same style wording for every image so they match: {{STYLE_WORDING}}.
Step 3. When I say "images", generate the images one per message, in handout order, so that I can click More under each one and choose Export to Docs as it arrives. Number each image, and under it write a caption of no more than 15 words and the handout section it belongs next to.
Do not include text inside the images. Do not include real people.
Fill in: HANDOUT_TOPIC AUDIENCE MAX_WORDS_PER_SECTION IMAGE_COUNT STYLE_WORDING
Google's image help says to start prompts with verbs like generate or create, describe the style and give a detailed visual description, and that under a generated image you can click More then Export to Docs. Users under 18 on school accounts cannot export to Docs.
Gemini Live
Gemini Live is the voice mode in the Gemini mobile app. These prompts use what Google's Live pages describe: sharing your camera or screen, Connected Apps such as Tasks and Keep, interrupting by speaking, and switching to text in the same conversation. Google has separate Live help pages for Android and for iPhone and iPad, says Live is not in the web app, and says Connected Apps support is rolling out gradually.
Best for: Hands-free help with the thing in front of you
(In the Gemini mobile app, tap Live, then tap the Camera icon to show Gemini what you see. Google says the camera turns off if Live goes on hold and comes back when you resume, but if you leave the app or lock the screen it does not turn back on by itself. Then say this.)
I am showing you {{WHAT_I_AM_SHOWING}} and I want to {{GOAL}}.
First tell me what you can see in one sentence, so I know the camera is working. If it is blurry or out of frame, tell me how to move.
Then guide me one step at a time. After each step wait for me to say "done" before the next. If I say "the camera stopped", ask me to turn it back on, then recap the last step and carry on from there.
Before any step that could cause damage, injury or data loss, stop and ask me to confirm. If a part is unclear, ask me to bring it closer or read its label. If this needs a professional (electrics, gas, medical or similar), say so and why.
Fill in: WHAT_I_AM_SHOWING GOAL
Google's Gemini Live help (Android page, with a separate iPhone and iPad page) says to tap the Camera icon in a Live chat, that Live is not in the Gemini web app, and how the camera behaves on hold, leaving the app or locking. Ask permission before showing other people.
Best for: Getting unstuck in an app without describing the screen
(First turn on notifications for Gemini Live; Google says you need them to share your screen. In the Live screen, tap Turn on screen sharing, or open Gemini outside the app and tap Share screen with Live if it appears. Your full screen is shared. Then say this.)
I am trying to {{TASK_ON_SCREEN}} in {{APP_OR_SITE}}, and I am stuck on {{WHERE_STUCK}}.
1. Tell me what you can see on my screen in one sentence, so I know it is working.
2. Give me only the next tap, then wait. Ask what changed before giving the next one.
3. If I say "sharing stopped", ask me to start it again, then recap what we had done so far.
4. Before I enter anything private, warn me. Never ask me to read out passwords, card numbers or one-time codes.
5. If the app looks different from what you expected, say so rather than guessing a menu name.
If a step needs a setting I do not have, tell me and suggest where to check.
Fill in: TASK_ON_SCREEN APP_OR_SITE WHERE_STUCK
Google's Gemini Live help (Android page, with a separate iPhone and iPad page) says screen sharing needs notifications for Gemini Live on, shares your full screen, and stops when Live is on hold or the screen locks. Close private apps and notifications first.
Best for: Emptying your head into Tasks and Keep while your hands are busy
(Open the Gemini app and tap Live. Google says Connected Apps in Live are being released gradually, that personal accounts must first connect the app in Connected Apps settings, and that work or school accounts can use only Calendar, Keep and Tasks. Stay in Live until you have checked what was added, because Google says you can undo an action only while Live is open on your screen.)
I am going to talk through everything on my mind about {{TOPIC_OR_PROJECT}}. Listen until I say "that's it".
Then:
1. Read back the items you heard, numbered, in short sentences.
2. Sort them out loud: things to do (for Google Tasks), things to remember (one Google Keep note titled "{{NOTE_TITLE}}"), and anything at a fixed time (ask me before it goes in Google Calendar).
3. Wait for my yes before adding anything. Then add them and read back what you created.
If I interrupt you, stop and answer. If you could not add something, say so.
Fill in: TOPIC_OR_PROJECT NOTE_TITLE
Google's Gemini Live help says you can interrupt by speaking, that Connected Apps support in Live is rolling out gradually, that you can undo an action only while Live is open, and that after you tap End you can review the transcript. Use the transcript to check nothing was missed.
Best for: Turning a spoken ramble into a written decision note
(Open the Gemini app and tap Live. Talk first; when you are done, switch to text mode in the same conversation. Google says switching to text does not end the session.)
I want to think out loud about {{DECISION_OR_DRAFT}}. While we talk:
1. Ask me one short question at a time to draw out my reasons, worries and options. Do not give me your view yet.
2. If I go quiet or go round in circles, sum up what I have said in one sentence and ask what is missing.
When I switch to text and type "write it up", give me:
- My options as I described them, with the reasons I gave for and against each.
- The questions I could not answer.
- The next step I said I would take, in my words, or "none stated".
Use only what I said. Do not add options of your own unless I ask.
Fill in: DECISION_OR_DRAFT
Google's Gemini Live help says switching to text mode in the same conversation does not end the Live session, that you can resume a previous Live chat by opening it and tapping Live, and that Live is not in the Gemini web app. Gems cannot be used with Live.
Building in Canvas
Canvas is the Gemini workspace where you and the model edit a document, app or code together. These prompts use its documented controls: select-and-prompt editing, Show console, Show recent changes, Select & ask, and PDF export for documents with LaTeX. Export options and age limits are on Google's Canvas help page.
Best for: A Canvas draft you can edit section by section
(In the Gemini app, click Add Files, then choose Canvas, and send this.)
Create a first draft of {{DOCUMENT_TYPE}} for {{AUDIENCE}} about {{TOPIC}}, in Canvas.
Facts to use: {{FACTS}}. Use no others. Where the document needs a fact I have not given, write [ADD: what is needed].
Structure it for editing in Canvas: short sections under clear headings, each able to stand alone, so that I can select one section and prompt on it without touching the rest.
From now on, when I select text in Canvas and give an instruction, change only the selected text, keep every fact and [ADD] marker in it, and tell me in one line what you changed. If my instruction would need a change outside the selection, tell me instead of making it.
Fill in: DOCUMENT_TYPE AUDIENCE TOPIC FACTS
Google's Canvas help says you can select text and enter a prompt, use Suggest edits, Change length and Change tone, step through Previous Version and Next Version, and use Share & export > Export to Docs. Changes are auto-saved.
Best for: Building and fixing a small app inside Canvas
Open Canvas (Add Files > Canvas) and build a small app that {{APP_PURPOSE}}. It takes {{INPUTS}} and shows {{OUTPUT}}. Do not store or send any entered data anywhere.
Then guide me through testing it with Canvas's own tools:
1. Which test inputs to try, including one invalid input and one boundary case, and the result I should see for each.
2. If something breaks, I will click Show console and paste the errors. When I do, explain the likely cause in plain English before fixing it.
3. After each fix, I will use Show recent changes under Code. Summarise what changed in one line so I can compare.
4. For changes to one part, I will use Select & ask. Change only the selected section.
Before I share a link: tell me that anyone with a public app link can view and edit the app's data, and list what to remove from the app first. Do not add AI features unless I ask.
Fill in: APP_PURPOSE INPUTS OUTPUT
Google's Canvas help describes Code view, Show console, Show recent changes, Select & ask, Add Gemini features, and warns that anyone with a public app link can view and edit data associated with the app. Adding Gemini features needs an account aged 18 or over.
Best for: A printable maths sheet that keeps its equations
Open Canvas (Add Files > Canvas) and create a worksheet on {{MATHS_TOPIC}} for {{LEARNER_LEVEL}}. Use LaTeX for every equation.
Include {{QUESTION_COUNT}} questions in increasing difficulty and a separate answer key with the working shown. Check each answer yourself before including it, and list any question where you are not confident.
Formatting: a title, a short instruction line, numbered questions, and the answer key on its own page. Do not use images.
Then tell me how to export it: in Canvas, open Share & export, export the document with LaTeX as a PDF, and use the preview before downloading. Warn me that copying a response with LaTeX gives the unrendered code, so I should export instead of copying.
Finish with the three places I should check for errors before giving this to learners.
Fill in: MATHS_TOPIC LEARNER_LEVEL QUESTION_COUNT
Google's Canvas help says a doc with LaTeX can be exported as a PDF with a preview before download, and that copying a response with LaTeX gives the unrendered code. Check every answer yourself.
Gems and Skills
Google is moving Gems to skills. This section has three prompts: one to convert a Gem's instructions and check its files, one dated migration plan that tells you to confirm the dates on Google's page first, and one to create a new skill in chat and test that Gemini uses it at the right time. The dates appear only in the migration-plan prompt and in the FAQ below.
Best for: Moving a working Gem across before it disappears
I have a Gemini Gem and I want to move it to a skill. Here are its instructions:
GEM INSTRUCTIONS:
{{GEM_INSTRUCTIONS}}
Knowledge files attached to it: {{KNOWLEDGE_FILES}}
Do this:
1. Rewrite the instructions as a skill: a lowercase, hyphenated name, a description saying when the skill applies, then the steps.
2. List every instruction that relies on a feature that may not work in skills. Google's transition page says some default Gem tools do not work with skills yet, including video creation, music creation, Canvas, Deep Research and Guided Learning. Mark each as "check this".
3. For each knowledge file, say whether its type is likely to be a problem. Google says skills do not support some types, including audio, video, GitHub files, archives, legacy .doc/.xls/.ppt files and Office formats such as .docx and .xlsx, while text files, PDFs and images are listed as supported. For a file that is a problem, say whether to convert it to plain text or Markdown, or leave it out.
4. Give me a five-item test list to run after migration.
Do not change what the Gem is meant to do.
Fill in: GEM_INSTRUCTIONS KNOWLEDGE_FILES
Google's help pages list the Gem tools and the knowledge-file types that skills do not support yet, and say to copy a Gem's name, description and instructions into a new skill and download its knowledge files one by one. Check those lists before you convert. The transition dates are in the migration-plan prompt below and in the FAQ.
Example output · Claude (example written by The Prompt Index)
**Inputs used:** GEM_INSTRUCTIONS = "You are Weekly Wrap. When I paste my notes, write a status update with Done, Next and Blockers. Keep it under 150 words. Use Canvas to put the final update in a document. If any note is unclear, ask me."; KNOWLEDGE_FILES = style-guide.docx, team-names.txt, kickoff-call.mp3.
**1. Skill version**
Name: `weekly-status-update`
Description: Use when I paste notes from my week and want a status update with Done, Next and Blockers.
Steps:
1. Read the pasted notes.
2. If any note is unclear, ask me before writing.
3. Write the update under three headings: Done, Next, Blockers.
4. Keep the update under 150 words.
5. Put the final update in a document.
**2. Check this**
- Step 5 came from "Use Canvas to put the final update in a document". Google's page says Canvas does not work with skills yet. Until it does, ask for the update in the chat and copy it into a document yourself.
**3. Knowledge files**
- `style-guide.docx`: a problem. Office formats such as .docx are not supported in skills. Convert it to plain text or Markdown.
- `team-names.txt`: fine, since text files are supported.
- `kickoff-call.mp3`: a problem. Audio is not supported. Transcribe it and keep the transcript as plain text, or leave it out.
**4. Tests after migration**
1. Paste notes with clear Done, Next and Blocked items; check the three headings appear.
2. Paste notes with one unclear item; check it asks before writing.
3. Paste long notes; check the update stays under 150 words.
4. Ask for a document; check what happens now that Canvas may not be available.
5. Paste notes that have no Blockers; check it does not invent one.
Best for: A dated checklist for moving your Gems, checked against Google
I have {{GEM_COUNT}} Gemini Gems on a {{ACCOUNT_TYPE}} account. Google's page "About the transition from Gems to skills" said on 28 October 2026 that on personal accounts Gems become read-only on 17 November 2026 and are removed on 17 December 2026; that Workspace business, enterprise and non-profit accounts move in March 2027 and education accounts in June 2027; and that Gems are transitioned to skills automatically when Gems go away. I have checked the current page and these are the dates it shows now: {{DATES_FROM_GOOGLE_PAGE}}.
My Gems, by name: {{GEM_LIST}}
Build me a dated plan:
1. Which dates apply to my account type, using the dates I pasted from the page.
2. A prioritised order for my Gems, with the reason for each.
3. Backup steps for each Gem: open it, click Edit, copy the name, description and instructions, and download each knowledge file.
4. For each Gem I keep, the tools or files it uses that Google says skills do not support yet.
5. Which Gems to retire instead of converting.
If the dates I pasted differ from the ones above, use mine and say so. If my account type is unclear, ask.
Fill in: GEM_COUNT ACCOUNT_TYPE DATES_FROM_GOOGLE_PAGE GEM_LIST
This is the only prompt in the pack that states the transition dates. They are from Google's 'About the transition from Gems to skills' page as read on 28 October 2026. Open that page first and paste what it says now; after the dates pass, this matters only for accounts that have not moved yet.
Best for: A reusable Gemini skill that triggers when it should
Create a skill based on these instructions, and save it to my Skills page.
Name: {{SKILL_NAME}} (lowercase words joined by hyphens)
When to use it: {{WHEN_TO_USE}}
What it should do, step by step: {{STEPS}}
What it must never do: {{NEVER_DO}}
Write the description so it says exactly when the skill applies, because Gemini uses activated skills automatically when they are relevant. Keep the instructions under 200 words.
After saving, give me three test messages to send in a new chat: one that should use the skill without my asking, one that should not, and one where I call it directly by typing / and its name. For each, say what a correct result looks like.
Fill in: SKILL_NAME WHEN_TO_USE STEPS NEVER_DO
Google's skills help says you can describe a skill in chat, for example 'Create a skill based on these instructions:', and Gemini saves it to your Skills page; activated skills are used automatically when relevant; names must be lowercase with hyphens; and you can call a skill by typing / (or @ with the updated shortcut). Skills need Keep Activity on and an account aged 18 or over, and work without a Google AI subscription. Review the saved skill on the Skills page.
Frequently asked questions
Do I need a paid plan to use these Gemini prompts?
It depends on the feature. Google's help pages say Gemini in Gmail, Docs, Sheets and Slides, and the AI function in Sheets, need an eligible Google Workspace or Google AI plan, with some features also offered to personal accounts through Workspace Experiments. Some image options need a Google AI plan, and Google's skills help says skills work without a Google AI subscription. Plans change, so check Google's current pages for your account.
How do I use Gemini inside a Google Sheets cell?
Google's help describes the AI function: type =AI() or =Gemini() in a cell, or use Insert > Function > AI, with a prompt and an optional range, then select the cells and click Generate and Insert. The function sees only the range you give it, and Google says only the first 350 selected cells with AI functions are generated at a time, so check its page for the current limit. You can also start from any cell by typing = and pressing Ctrl + Alt + g (⌘ + Ctrl + g on a Mac) to open the Gemini panel. The Sheets section has five AI-function prompts.
Why can't Gemini find my email or document?
Check these. You must be signed in to the same account you use with Google Workspace, Keep Activity must be on, and on a work or school account an administrator must allow the connection. Type @ and select the service in the prompt, mention it in the first prompt of a chat, and use keywords instead of pasting a link. Google also lists things Gemini cannot do, such as read images or comments in Docs and Gmail, read pictures or videos in Drive, or count items in Gmail or Drive.
Are Gems going away?
Google's page 'About the transition from Gems to skills' said on 28 October 2026 that Gems on personal accounts become read-only on 17 November 2026 and are removed on 17 December 2026, that Workspace business, enterprise and non-profit accounts follow in March 2027 and education accounts in June 2027, and that Gems are moved to skills automatically. Check that page for the current dates. The Gems and Skills section has a dated migration plan and a conversion prompt.
Can I use these prompts in ChatGPT or Claude?
They are written around Gemini features such as the Workspace connection, the Gmail and Sheets side panels, the AI function, Canvas, Gemini Live and Gemini's image tools, so they will not behave the same elsewhere. For Claude, see the Claude prompts pack.
Does Gemini make things up, and how do these prompts handle that?
Yes, it can. Google's help pages warn that Gemini can give inaccurate or outdated information. Many prompts here ask for exact quotes, ask Gemini to say 'not stated' or 'not found', or ask it to propose an action and wait for your yes. Verify anything important against the original email, file, cell or source yourself.