# 25 Claude Prompts for Work You Can Copy and Paste (Emails, Excel, Reports)

> Twenty-five original, copy-paste Claude prompts for emails, Excel, reports and meetings, plus the Anthropic-documented prompting rules that make them work.

Source: https://www.epifive.com/blog/claude-prompts-for-work-copy-paste · Published 2026-11-16 · EpiFive

## Key takeaways

- A good work prompt names the task, the audience, the source material, the output format and the constraints.
- Anthropic recommends wrapping instructions, context and input in separate XML-style tags when a prompt mixes them.
- Put long documents above your question; Anthropic's tests show queries at the end can improve quality by up to 30 percent.
- Every prompt here is a template, so replace the bracketed parts with your own data and check the output before using it.
- Claude can be wrong, so verify numbers, formulas and facts before sending anything.

These 25 prompts cover the work most people hand to Claude: writing emails, cleaning up spreadsheets, summarizing documents and preparing meetings. Copy a prompt, replace the text in square brackets with your own material, and send it. They follow the rules in Anthropic's prompting guide: say what you want precisely, give context, show an example when format matters, and separate instructions from data.

## Who this is for and what you will have

This article is for people who use Claude at a desk job: office managers, analysts, team leads, IT admins and small business owners. You do not need to know anything about prompt engineering. By the end you will have a set of reusable prompts in five groups (email, Excel, reports, meetings, planning) and a short method for fixing any prompt that gives a weak answer.

None of these prompts were run in a pipeline for this article, so the sample outputs described are the shape of answer to expect, not recorded transcripts. Adjust each prompt to your data and read the result before you use it.

## The five rules behind every prompt

Anthropic's prompting best practices page, the reference for current Claude models, lists the same few techniques again and again. The prompts below apply them.

### Be clear and direct

The guide says Claude responds well to clear, explicit instructions and suggests thinking of Claude as "a brilliant but new employee who lacks context on your norms and workflows." Its key test: show your prompt to a colleague with minimal context and ask them to follow it. If they would be confused, Claude will be too. Be specific about the output format and constraints, and use numbered steps when order matters.

### Give the reason

The guide recommends giving context or motivation behind an instruction. Its own example replaces "NEVER use ellipses" with an explanation that the text will be read aloud by a text-to-speech engine that cannot pronounce them. In work terms: "Keep it under 120 words because it will be read on a phone" beats "Be short."

### Show examples when format matters

Examples are described as one of the most reliable ways to steer format, tone and structure. Anthropic suggests 3 to 5 examples that are relevant, diverse and wrapped in `<example>` tags. For a one-off task, one pasted sample of the style you want is still better than none.

### Separate instructions from data with tags

When a prompt mixes instructions, context and pasted input, Anthropic recommends wrapping each type in its own tag, such as `<instructions>`, `<context>` and `<input>`, and using consistent tag names. Most prompts below use this layout.

### Put long material first

For long documents, the guide says to place them near the top, above your query and instructions, and notes that queries at the end can improve response quality by up to 30 percent in tests. When you paste a long report, put it first and your request last.

## Email prompts

Email is the easiest place to start because the output is short and you can judge it at a glance.

### 1. Reply to a difficult email

```text
<context>I manage [team/department]. The email below is from [role, e.g. a customer] and is frustrated about [issue]. We can offer [what you can actually offer] but cannot [what you cannot do].</context>
<email>[paste the email]</email>
<instructions>Write a reply of about 120 words. Acknowledge the specific problem in the first sentence, state what we will do and by when, and do not make promises beyond what is in the context. Tone: calm and professional. Give me one version only.</instructions>
```

Expect a short, structured reply that names the problem first. Check every date and promise.

### 2. Shorten an email

```text
Rewrite the email below to half its length. It will be read on a phone, so put the decision or request in the first line. Keep all names, dates and numbers unchanged.
<email>[paste draft]</email>
```

### 3. Polite follow-up

```text
Write a follow-up to the message below. It was sent [number] days ago with no reply. Assume the reader is busy, not ignoring me. Two to three sentences, one clear question, no apology for "bothering" them.
<previous_message>[paste]</previous_message>
```

### 4. Three tones, one message

```text
Rewrite this message in three versions: formal, friendly and brief. Label each version. Keep the facts identical.
<message>[paste]</message>
```

### 5. Turn a thread into a summary

```text
<thread>[paste the whole email thread]</thread>
Summarize the thread in this format: (1) what was decided, (2) what is still open, (3) who owes what by when. If a date or owner is not stated, write "not stated" instead of guessing.
```

The "not stated" rule matters: it gives Claude a permitted way to say it does not know.

## Excel and spreadsheet prompts

Claude can explain and draft formulas, but it cannot see your workbook unless you paste or upload the data. Always test a formula on a few rows you can check by hand.

### 6. Write a formula from plain English

```text
I use [Excel for Microsoft 365 / Excel 2019]. Column A is [what], column B is [what], column C is [what]. Write one formula for cell D2 that [describe the result]. Explain it in two sentences and show what it returns for this sample row: [paste a row].
```

Naming your Excel version stops Claude from suggesting functions your version lacks.

### 7. Explain a formula you inherited

```text
Explain this Excel formula step by step as if I am new to it. Then list any ways it could return a wrong result (blank cells, text numbers, duplicates).
<formula>[paste]</formula>
```

### 8. Find problems in pasted data

```text
Below is a sample of my data in CSV form. List inconsistencies you can see: mixed date formats, trailing spaces, duplicate rows, misspelled categories, numbers stored as text. For each, give the column, two example values and the fix. Do not change the data yet.
<data>[paste 30-50 rows]</data>
```

### 9. Clean a column

```text
Here are 20 messy values from one column. Propose a standard form for each and give me a two-column table (original, cleaned). Then describe the rule you used so I can apply it with Find and Replace or Power Query.
<values>[paste]</values>
```

### 10. Plan a summary table

```text
I have a sheet with these columns: [list]. I need a monthly summary of [measure] by [category]. Recommend whether to use a PivotTable or formulas, and give the exact steps in Excel menus.
```

### 11. Choose a chart

```text
My audience is [who] and the point I need to make is [one sentence]. The data is [describe]. Recommend one chart type, say what goes on each axis, and give a title that states the conclusion.
```

## Report and document prompts

### 12. Summarize a long document

```text
<document>[paste the document]</document>
Summarize the document for [audience] in five bullet points, then list three questions a careful reader would ask. Quote the sentence that supports each bullet.
```

Asking for supporting quotes makes it easy to verify the summary against the source.

### 13. Executive summary

```text
Write an executive summary of about 150 words for [role]. Lead with the recommendation, then the two strongest reasons, then the main risk. Use only facts present in the document below.
<document>[paste]</document>
```

### 14. Report outline

```text
I need a [type of report] for [audience] about [topic]. They will use it to decide [decision]. Propose an outline with section titles and one sentence on what each section must prove. Keep it to eight sections or fewer.
```

### 15. Improve my draft

```text
Edit the draft below for clarity. Do not add new facts. Return (1) the edited text and (2) a list of the five most important changes you made and why.
<draft>[paste]</draft>
```

### 16. Plain-language rewrite

```text
Rewrite this paragraph so a non-specialist can understand it. Replace jargon with everyday words, keep every number, and keep it under [number] words.
<paragraph>[paste]</paragraph>
```

### 17. Review a document for gaps

```text
Act as a skeptical reviewer from [department]. Read the document and list claims without evidence, undefined terms, missing numbers and anything a reader could misread. Order by importance.
<document>[paste]</document>
```

## Meeting prompts

### 18. Meeting agenda

```text
Create a 30-minute agenda for a meeting about [topic] with [attendees]. The goal is [decision or outcome]. Give each item a time box, an owner placeholder and the question to be answered.
```

### 19. Minutes from rough notes

```text
<notes>[paste rough notes or a transcript]</notes>
Turn these into minutes: attendees mentioned, decisions, action items as a table (action, owner, due date), and open questions. Mark anything uncertain with "[confirm]". Do not invent owners or dates.
```

### 20. Prepare for a hard conversation

```text
I need to tell [role] that [message]. They are likely to react by [worry]. Draft an opening of three sentences, then list four likely objections with a short, honest response to each.
```

## Planning and thinking prompts

### 21. Compare options

```text
Compare [option A] and [option B] for [situation]. Use a table with these rows: cost, effort, risk, time to start, what could go wrong. Then give a recommendation and say what facts would change it. If you are unsure of a fact, say so.
```

### 22. Break a project into tasks

```text
Break this project into tasks I can assign: [describe]. Give a numbered list in order, with an estimate in hours, dependencies, and the first task I can start today. Assume a team of [size].
```

### 23. Ask me questions first

```text
I want to [goal]. Before you answer, ask me up to five questions that would most improve the result. Wait for my answers.
```

This is useful when you cannot tell what context Claude needs.

### 24. Prepare interview questions

```text
Write eight interview questions for a [role]. For each, say what a strong answer includes and one warning sign. Keep questions free of anything about age, family or other protected characteristics.
```

### 25. Critique my plan

```text
Here is my plan. Find the three weakest assumptions, explain how each could fail, and suggest one cheap way to test each.
<plan>[paste]</plan>
```

## Turn a good prompt into a reusable setup

If you use the same instructions daily, move them out of the chat. Anthropic's help center describes Projects as workspaces with their own chat histories and knowledge bases, where you can upload documents and set project instructions, for example a formal tone or an industry-expert role. Free accounts can create up to five Projects; the help page says extended knowledge capacity through retrieval-augmented generation requires a paid plan (Pro, Max, Ultra, Team or Enterprise), and Team and Enterprise plans can share projects with colleagues using "Can view" or "Can edit" permissions. For a comparison with Custom GPTs, see [ChatGPT Custom GPTs vs Claude Projects for team knowledge](https://www.epifive.com/blog/chatgpt-custom-gpts-vs-claude-projects-team-knowledge). If you plan to call Claude from code instead, see [Claude tool use in a Python agent loop](https://www.epifive.com/blog/claude-tool-use-python-agent-loop).

## How to adapt a prompt to your own work

A template only helps if the bracketed parts are filled in with specifics, so spend your effort there. Three habits make the biggest difference across all 25 prompts.

First, name the reader. "Write a summary" produces a different result from "write a summary for a finance director who has two minutes before a meeting." The reader decides vocabulary, length and what to leave out, and Claude cannot guess it. If you have a style you like, paste one real paragraph you wrote earlier and say "match this tone." That follows the guide's advice to steer tone and structure with examples.

Second, tell Claude what it may not do. Most bad outputs come from the model filling gaps: inventing a date, assigning an owner, or smoothing over a number it was not given. Phrases such as "use only the text below," "write 'not stated' if missing" and "do not add new facts" give it a legitimate alternative to guessing. They cost one line and remove the most common risk in work documents.

Third, iterate in the same chat. After the first answer, reply with a specific correction: "shorter, remove the second paragraph, keep the deadline in the first line." That is usually faster than rewriting the whole prompt. If the conversation drifts after many edits, start a new chat and paste the improved prompt, so earlier instructions do not conflict with the new ones.

## Which prompts suit which job

Some prompts are safe to use almost as they are, while others need a closer check. Prompts that only reshape text you supplied, such as numbers 2, 4, 15 and 16, carry low risk because every fact comes from your own draft. Prompts that summarize or extract, such as 5, 12, 13 and 19, depend on the source you paste, so skim the original against the output once. Prompts that ask Claude to propose something, such as 10, 14, 21 and 22, produce a starting point to react to, not a finished decision. Excel prompts 6 to 9 produce formulas and rules that must be tested on rows whose answer you already know before they touch a full dataset.

A practical routine for a busy week: keep five favorites in a note, one per group, and replace the bracketed parts each time. If your team shares them, agree on one place to store them and one person who prunes the list monthly, so prompts that stopped working do not linger.

## Fix a weak answer

When the answer is generic, change the prompt rather than asking again with the same text.

| Symptom | Likely cause | Fix |
| --- | --- | --- |
| Generic, padded answer | No audience or length | State who reads it and the word limit |
| Wrong format | Format not specified | Describe the format, or paste one sample |
| Invented facts | Source not provided | Paste the source and say "use only the text below" |
| Ignores part of the request | Too many tasks in one message | Number the steps, or split into two prompts |
| Misreads pasted data | Data mixed with instructions | Wrap data in tags such as `<data>` |
| Formula errors | Version or layout unknown | State the Excel version and column layout |

## Safety habits

- Verify numbers, dates, quotes and formulas yourself. A fluent answer is not a checked answer.
- Remove personal data and secrets when your policy requires it. Use placeholders such as Contoso Ltd or admin@contoso.com for practice.
- Keep a human review step before anything goes to a customer, regulator or court.
- Save the prompts that work in a note, a Project or a shared document so your team reuses them.

## What we checked

- Opened Anthropic's "Prompting best practices" page on platform.claude.com on 2026-11-16 and confirmed the sections "Be clear and direct", "Add context to improve performance", "Use examples effectively", "Structure prompts with XML tags" and "Long context prompting".
- Confirmed on that page the recommendation of 3 to 5 examples wrapped in example tags, and the note that queries at the end can improve response quality by up to 30 percent in tests.
- Opened the Claude help center article "What are Projects?" on 2026-11-16 and confirmed: free users can create up to five projects, RAG capacity needs a paid plan, and sharing uses "Can view" and "Can edit".
- Opened the Prompt engineering overview on 2026-11-16 and confirmed it points to the best practices page as the living reference.
- Counted the prompts: 5 email, 6 Excel, 6 report, 3 meeting and 5 planning, 25 in total.
- Not checked: none of the prompts was executed against Claude for this article, and no spreadsheet formulas were tested. Output descriptions are expectations, not recorded results. Plan features can change, so confirm them on the help center.

Last tested: 2026-11-16

## References

- Prompting best practices: https://platform.claude.com/docs/en/build-with-claude/prompt-engineering/claude-prompting-best-practices
- Prompt engineering overview: https://platform.claude.com/docs/en/build-with-claude/prompt-engineering/overview
- What are Projects? (Claude help center): https://support.claude.com/en/articles/9517075-what-are-projects

## Frequently asked questions

### Do I need a paid Claude plan to use these prompts?

No. The prompts are plain text and work in any Claude chat. Anthropic's help center says Projects are open to all users including free accounts (up to five), but the extra knowledge capacity from retrieval-augmented generation needs a paid plan.

### Why do my Claude answers feel generic?

Usually the prompt lacks context. Anthropic's documentation says to be specific about the output format and constraints, and to explain why you want something. Name the audience, the purpose, the length and the format, and paste the real source material.

### Should I paste confidential company data into these prompts?

Only if your company policy and your plan's data terms allow it. Replace names, account numbers and personal data with placeholders when you are unsure, and check your organization's AI-use policy first.

### How many examples should I include in a prompt?

Anthropic's prompting guide suggests 3 to 5 examples for best results, kept relevant and diverse, and wrapped in example tags so Claude can tell them apart from your instructions.

