PromptShifu
🌐 δΈ­ζ–‡

Why Gemini Prompts Need Their Own Library

Gemini is built to read images, video, audio, and long documents alongside text. The strongest Gemini prompts say exactly what is attached, what to look for, and how to cite locations such as timestamps, page numbers, or on-screen positions, and they tell the model to mark anything unreadable instead of guessing. Every template below follows that pattern.

Copy any prompt with one tap, or save it to PromptShifu and type /ps in Gemini to inject it. For general-purpose text prompts, see the ChatGPT & text library.

πŸ“š All Prompts 🎨 Midjourney πŸ€– ChatGPT 🧠 Claude πŸ‹ DeepSeek ✨ Gemini
GeminiMultimodalUXAccessibility

UI Screenshot UX Critic

Upload a screenshot and get a prioritized usability and accessibility review

View full prompt
I am attaching a screenshot of a [web page / mobile screen / dashboard].

Product context: [what the product does and who uses it]
Goal of this screen: [e.g. "get new users to start a free trial"]

Review the screenshot as a senior UX designer:
1. Describe what a first-time user notices in the first 5 seconds and whether that matches the goal.
2. Evaluate visual hierarchy, spacing, typography, and the clarity of the primary call to action.
3. Flag accessibility concerns you can see: low color contrast, small tap targets, text embedded in images, unclear labels.
4. Point out copy that is vague or inconsistent.
5. Refer to specific elements by their location on screen (for example "top-right button") so I can find them.

Return a table: Issue | Location | Severity (High/Medium/Low) | Suggested fix. End with the 3 changes that would have the biggest impact.
GeminiVideoMultimodalSummarization

Video Summary with Timestamps

Turn an uploaded video or YouTube link into chapters, key points, and quotes

View full prompt
Watch the attached video (or the YouTube video at this link: [URL]) and summarize it.

My purpose: [e.g. "decide whether the full talk is worth watching" or "take study notes"]

Produce:
1. A 3-sentence overview of the main argument or story.
2. A chapter list with timestamps in MM:SS format and a one-line description of each chapter.
3. The 5 most important takeaways, each with the timestamp where it is discussed.
4. Notable quotes, transcribed word for word with timestamps.
5. Anything shown on screen (slides, charts, code, demos) that is important but not said aloud.

If part of the video is unclear or inaudible, say so rather than guessing.
GeminiPDFLong ContextData Extraction

Long PDF Table & Data Extractor

Pull tables and figures from long PDFs into clean, verifiable structured data

View full prompt
I am attaching a PDF: [document name, e.g. "2025 annual report"].

Task: extract [e.g. "every table that reports quarterly revenue by region"].

Instructions:
1. Find every relevant table or figure in the document, including ones inside appendices.
2. Reproduce each table in Markdown with the original column headers and units.
3. Under each table, cite the page number and the table or figure title.
4. Do not compute, round, or convert values unless I ask; copy numbers exactly as printed.
5. If a value is unreadable or a table spans pages and is ambiguous, mark the cell as [unclear] and explain.

After the tables, list any figures in the document that contain relevant data only as a chart, with a description of what the chart shows.
GeminiMultimodalChartsAnalytics

Chart Image to Data Table

Read values from a chart image and rebuild the underlying dataset

View full prompt
I am attaching an image of a chart.

Please:
1. Identify the chart type, title, axis labels, units, and legend entries.
2. Estimate the value of every data point or bar and present them in a table with one row per category or time period.
3. State your estimation precision (for example, plus or minus 2 units) and flag any points that are hard to read.
4. Describe the main trend, the largest change, and any outliers in 3 bullet points.
5. Note anything potentially misleading in the chart design, such as a truncated y-axis or dual axes with different scales.

Finally, output the data table as CSV in a code block so I can paste it into Google Sheets.
GeminiMultimodalProductivityMeetings

Whiteboard Photo to Meeting Notes

Convert a photo of a whiteboard or sticky notes into structured notes and tasks

View full prompt
I am attaching a photo of a whiteboard (or sticky notes) from a working session about [topic].

Please:
1. Transcribe all readable text, preserving the grouping and any arrows or connections between items.
2. Mark any words you cannot read confidently as [illegible] instead of guessing.
3. Reorganize the content into: Key ideas, Decisions, Open questions, and Action items.
4. For action items, include an owner and due date if written on the board; otherwise leave them as TBD.
5. If the board contains a diagram or flow, describe it as a numbered sequence of steps.

Format the result as clean notes I can paste into Google Docs.
GeminiMultimodalFinanceOCR

Receipts to Expense Report

Read photos of receipts and build a categorized expense table

View full prompt
I am attaching photos of [number] receipts from a business trip to [destination].

For each receipt, extract:
- Merchant name
- Date (YYYY-MM-DD)
- Total amount and currency
- Tax amount if shown
- Category: Travel, Lodging, Meals, Supplies, or Other

Rules:
1. Only use values printed on the receipt. If something is missing or unreadable, write "not shown".
2. Do not convert currencies unless I provide exchange rates.
3. Flag duplicates, receipts with no date, and any amount over [policy limit].

Return a table sorted by date, then totals per category and per currency, and a short list of items that need my attention before I submit the report.
GeminiAudioMultimodalMeetings

Meeting Audio Transcriber & Summarizer

Transcribe an audio recording with speakers, then summarize decisions and tasks

View full prompt
I am attaching an audio recording of a meeting about [topic]. Participants: [names and roles, if known].

Please:
1. Produce a transcript with timestamps every few minutes and speaker labels. If you cannot tell who is speaking, use Speaker 1, Speaker 2, and so on.
2. Remove filler words but do not change the meaning of what was said.
3. Write a summary with: Purpose of the meeting, Key discussion points, Decisions made, Action items (owner, task, deadline), and Unresolved questions.
4. Quote verbatim, with a timestamp, any statement that commits someone to a deadline or a number.

If audio quality makes a section unreliable, mark it with the time range.
GeminiResearchDeep ResearchStrategy

Deep Research Brief Builder

Frame a research question so Gemini Deep Research returns a focused, sourced report

View full prompt
Research the following topic and produce a sourced report: [topic, e.g. "the state of solid-state battery commercialization"].

Scope:
- Time frame: [e.g. developments since 2023]
- Geography: [global / specific regions]
- Audience: [e.g. a product strategy team with no chemistry background]

The report must answer these questions:
1. [Question 1]
2. [Question 2]
3. [Question 3]

Requirements:
- Prefer primary sources: company filings, peer-reviewed papers, government and standards bodies.
- Cite a source for every factual claim and note the publication date.
- Separate established facts from forecasts and opinions.
- Highlight where credible sources disagree.

End with an executive summary of 5 bullets and a list of open questions that need more research.
GeminiFact CheckSearchResearch

Search-Grounded Fact Checker

Verify claims against current web sources and rate each one with citations

View full prompt
Fact-check the following text using current information from Google Search.

Text:
[Paste the article, post, or claims to check]

For each factual claim:
1. Quote the claim exactly.
2. Search for current, reliable sources and summarize what they say.
3. Rate the claim: Accurate, Mostly accurate, Misleading, Inaccurate, or Unverifiable.
4. Cite the sources with titles and links, and note their publication dates.
5. If the facts have changed recently, explain what changed and when.

Do not rate opinions or predictions; list them separately as "not checkable". End with a corrected version of the text that fixes any inaccurate statements.
GeminiEmailWorkspaceCommunication

Email Thread Reply Drafter

Summarize a long email thread and draft a clear reply in your voice

View full prompt
Here is an email thread I need to reply to:

[Paste the thread, or reference it in Gmail]

My role: [e.g. project lead]
What I want to achieve with my reply: [e.g. "push the launch by one week without losing the client's trust"]
Tone: [warm / direct / formal]

Please:
1. Summarize the thread in 3 bullets: what has been asked, what has been agreed, and what is still open.
2. List every question or request addressed to me that my reply must answer.
3. Draft a reply under 180 words that answers each point, proposes a clear next step, and ends with a specific ask.
4. Offer one alternative subject line if the current one is unclear.

Do not promise anything I have not stated.
GeminiGoogle SheetsWorkspaceSpreadsheets

Google Sheets Formula Builder

Describe what you need in plain English and get a working Sheets formula

View full prompt
Write a Google Sheets formula for this task: [e.g. "sum column D where column B is 'Paid' and the date in column A is in the current month"].

Sheet layout:
- Sheet name: [name]
- Columns: [A = Date, B = Status, C = Client, D = Amount, and so on]
- Header row: [row 1]
- Data range: [e.g. rows 2 to 5000, growing over time]

Requirements:
1. Prefer modern functions such as FILTER, QUERY, XLOOKUP, LET, and ARRAYFORMULA where they make the formula simpler.
2. Make it robust to blank rows and new data being added.
3. Explain how the formula works, piece by piece.
4. Give one simpler alternative if the main formula is complex.
5. List common errors (#N/A, #VALUE!, #REF!) this formula might show and how to fix each.
GeminiGoogle SlidesPresentationsWorkspace

Google Slides Deck Outliner

Turn a document or idea into a slide-by-slide deck outline with speaker notes

View full prompt
Create a presentation outline from the material below.

Source material: [Paste text, or attach a document or PDF]
Audience: [e.g. executive leadership]
Goal: [e.g. "approve budget for phase 2"]
Length: [number] slides, about [minutes] minutes

For each slide, provide:
1. A slide title written as a takeaway sentence, not a topic label.
2. 3 bullet points of no more than 10 words each.
3. A suggested visual: chart type, diagram, image, or table, with what it should show.
4. Speaker notes of 2 to 4 sentences.

Structure the deck as: hook, problem, evidence, proposal, risks, and a clear ask on the final slide. Flag any claim in the source material that needs a citation before presenting.
GeminiMultimodalFront-EndHTML

Screenshot to Responsive HTML

Recreate a UI from an image as accessible, responsive HTML and CSS

View full prompt
I am attaching a screenshot or mockup of a user interface.

Recreate it as a single, self-contained HTML file.

Requirements:
1. First describe the layout you see: sections, grid or flex structure, components, and approximate color palette.
2. Use semantic HTML (header, nav, main, section, footer, button, label).
3. Use [plain CSS / Tailwind CSS via CDN] and make the layout responsive from 360px to desktop width.
4. Meet accessibility basics: alt text for images, labelled form fields, visible focus states, and sufficient color contrast.
5. Use placeholder images and text where the screenshot contains real photos or personal data.
6. Keep JavaScript to the minimum needed for interactions visible in the design, such as a menu toggle.

Return the full file in one code block, then list any parts of the design you had to guess.
GeminiMultimodalE-commerceCopywriting

Product Photo to Listing Copy

Write an e-commerce listing from a product photo, grounded in what is visible

View full prompt
I am attaching photos of a product I want to sell on [Shopify / Amazon / Etsy / my own store].

Known details: [materials, dimensions, price, brand, anything not visible in the photo]
Target customer: [who buys this and why]

Please:
1. Describe what you can see: shape, color, materials, finish, and any visible features or text. Mark anything uncertain.
2. Write a product title under 80 characters with the most important search keywords.
3. Write 5 benefit-led bullet points.
4. Write a 120-word description with sensory language.
5. Suggest 10 search keywords and alt text for each photo.

Only state features that are visible in the photos or listed in the known details. Do not invent certifications, measurements, or materials.
GeminiLong ContextCodingOnboarding

Long-Context Codebase Explainer

Load a whole repository and get an architecture map plus answers with file references

View full prompt
I am providing the source code of a repository: [attach files, a zip, or paste the code with a file path header before each file].

I am new to this codebase. My first task is: [e.g. "add rate limiting to the public API"].

Please:
1. Summarize what the project does and its main technologies.
2. Map the architecture: entry points, key modules, data flow, and external services. Reference real file paths.
3. Explain the conventions I should follow: folder structure, naming, error handling, testing, and configuration.
4. For my task, list the files I will most likely need to change and explain why, citing the specific functions involved.
5. Point out risky areas: code without tests, complex functions, or TODO and FIXME comments related to my task.

Only describe code that is actually present. If something is missing from what I provided, say what you would need to see.

← Back to the full PromptShifu Prompt Gallery

Gemini Prompting FAQ

What makes a good multimodal prompt for Gemini?

Tell Gemini what you attached, what you want from it, and how to reference evidence: timestamps for video and audio, page numbers for PDFs, and on-screen locations for screenshots. Ask it to mark unreadable or uncertain parts explicitly so it does not fill gaps with guesses.

Can Gemini work with long PDFs and whole codebases?

Yes. Gemini models support a long context window, so you can attach long reports or many source files in one prompt. Ask for citations to pages or file paths so every answer can be checked against the original material.

Which templates use Google Search or Deep Research?

The Search-Grounded Fact Checker asks Gemini to verify claims against current web sources, and the Deep Research Brief Builder frames a scoped question for Gemini Deep Research. Both require a cited source and publication date for each factual claim.

Can I inject these Gemini prompts with PromptShifu?

Yes. Copy any template with one tap, or save it to the PromptShifu extension and type /ps in the Gemini composer to inject it instantly. Prompts stay in local browser storage with no account and no prompt logging.