These are 40 complete, paste-ready prompts for Claude — built for Claude Opus 4.8 and the top-tier Fable 5, with Sonnet 5 and Haiku 4.5 for speed. Every prompt follows the Role → Task → Context → Format (RTCF) pattern, uses [bracketed placeholders] you fill in, and wraps long input in XML tags — the one habit that reliably makes Claude sharper. Each has a one-line note on why it works.
Want the theory behind these? Read the guide to prompting Claude for coding, keep the Claude prompt cheat sheet open while you copy, and grab fill-in-the-blank templates when you want skeletons.
Work & productivity
Seven prompts that turn scattered inputs into plans, briefs, and decisions. Turn on extended thinking for the decision memo; the rest run fine without it.
1. Turn a Brain-Dump Into a Plan
Act as my chief of staff. Turn my brain-dump into an organized action plan.
Here is everything on my mind right now:
<braindump>
[paste your unsorted notes, tasks, and half-formed ideas]
</braindump>
Do this:
- Group the items into themes and label each theme.
- Tag every item as a task, an idea, or a question.
- List the 3 highest-leverage tasks to do first, and why.
- End with the information you'd need from me to go further.
Keep it scannable with short bullets.Why it works: Wrapping the mess in a <braindump> tag separates your raw notes from the instructions, so Claude organizes instead of obeying stray lines inside them.
2. Meeting Prep Brief
Prepare me for a meeting so I walk in ready.
The meeting is [topic] with [who is attending and their roles]. My goal is [what I want out of it].
Background:
<background>
[paste agenda, prior notes, or the email thread]
</background>
Give me:
- A 5-line situation summary.
- 3 outcomes to push for and 3 objections to expect, with a one-line response to each.
- 5 sharp questions I should ask.
- Anything I'm missing that could derail the meeting.Best for: Walking into any meeting with a point of view instead of reacting on the fly.
3. Prioritize My Week Realistically
Help me plan a realistic, high-impact week.
My goals this week: [goals].
My tasks and deadlines:
<tasks>
[paste tasks with any due dates]
</tasks>
I have roughly [X] focused hours per day.
Do this:
- Rank tasks by impact vs. effort and explain the ranking in one line each.
- Build a day-by-day plan that fits my available hours — do not overload any day.
- Tell me what to drop or delegate if it won't all fit.Why it works: Giving Claude your real available hours forces a plan you can execute, not a wish list.
4. Summarize a Long Document
Summarize a long document for a busy reader.
The reader is [role] who cares most about [their priority].
<document>
[paste or attach the report / contract / PDF / transcript]
</document>
Give me:
- A 3-sentence executive summary first.
- The 7 key points as bullets, in priority order.
- Every number, date, deadline, and commitment pulled out verbatim.
- The 2-3 questions this document leaves unanswered.Best for: Claude's large context window — paste an entire long report and still get one clean, prioritized summary.
5. Decision Memo From Options
Help me make a defensible decision and write it up. Think step by step before answering.
The decision is [decision]. My options:
<options>
[Option A]
[Option B]
[Option C]
</options>
What I care about: [budget, timeline, risk, team, etc.].
Do this:
- Score each option against my constraints in a table.
- Give a clear recommendation with your reasoning.
- Write a one-page decision memo I can send to my manager.
- List the top 2 risks of the recommended option and how to reduce them.Why it works: "Think step by step" triggers Claude's reasoning before it commits, so the recommendation follows the trade-offs instead of preceding them.
6. Summarize a Meeting Transcript Into Actions
Turn a meeting transcript into a clear follow-up.
<transcript>
[paste the Zoom / Teams / Otter transcript]
</transcript>
Give me:
- A 5-bullet summary of what was decided.
- Every action item as: owner — task — due date (mark "unassigned" or "no date" where missing).
- Any disagreements or unresolved questions.
- A short recap email I can send to attendees.Best for: Meetings you attended but didn't take notes in — the transcript captures, Claude structures. See the business & work prompts for more of these.
7. Be My Thinking Partner
Be my thinking partner on a decision, not a cheerleader.
Here's what I'm weighing:
<situation>
[describe the situation, what you're leaning toward, and why]
</situation>
Do this:
- Ask me up to 3 clarifying questions first if anything important is missing.
- Then reflect back what you think my real goal is.
- Push back where my reasoning is weak or I'm avoiding something.
- Offer one option I probably haven't considered.
Be direct and honest; I'd rather be challenged than flattered.Why it works: Explicitly asking for pushback overrides Claude's default helpfulness, so you get a real second opinion instead of agreement.
Writing & communication
Six prompts for drafting, editing, and adjusting tone. Haiku 4.5 handles quick rewrites fast; use Opus 4.8 when nuance matters.
8. Draft an Email in My Voice
Draft a reply in my voice.
I need to reply to [sender] about [subject]. My goal for the reply is [goal]. Tone: [warm and professional / direct / apologetic].
The thread so far:
<thread>
[paste the email thread]
</thread>
Constraints:
- Under 150 words.
- Address every open question in the thread.
- Don't invent facts — if you need something I haven't given, leave a [bracketed placeholder].
- Give me two versions: one concise, one slightly warmer.Best for: Fast, in-context replies. For a full set, see the Claude writing prompts.
9. Rewrite in a Chosen Tone
Rewrite my text in a specific tone without changing the meaning.
<draft>
[paste text]
</draft>
Target tone: [confident and warm / formal / plain and direct].
Audience: [who reads this].
Constraints:
- Preserve every fact and the original intent.
- Cut filler, hedging, and throat-clearing.
- Give me the rewrite, then a one-line note on what you changed and why.Why it works: Locking tone and audience separately from the content stops Claude from drifting off-message while it polishes.
10. Sharpen My Writing
Act as a strict line editor. Make this sharper without making it sound like AI.
<text>
[paste your draft]
</text>
Do this:
- Tighten every sentence; cut words that earn nothing.
- Kill clichés, hedging, and corporate filler.
- Keep my voice and any deliberate quirks.
- Show the edited version, then list the 3 biggest changes and why.
Do not add new claims or facts.Best for: Turning a solid draft into a tight one while keeping it human.
11. Blog Outline From a Topic
Outline a blog post that's actually useful.
Topic: [topic]. Audience: [who it's for and their level]. Search intent I'm targeting: [what the reader wants].
Give me:
- A working title and 2 alternates.
- An H2/H3 outline with a one-line note on what each section covers.
- Where to add an example, a checklist, or a table.
- The 3 questions the piece must answer to be worth publishing.Why it works: Deciding structure and intent up front makes the actual writing go much faster.
12. First Draft From My Notes
Write a first draft from my rough notes.
Piece: [blog post / newsletter / LinkedIn post / section]. Audience: [who]. Desired length: [words].
My notes and the points I must hit:
<notes>
[paste your bullet points, quotes, and links]
</notes>
Constraints:
- Use only what's in my notes — don't invent facts or stats.
- Match this style: [plain and direct / warm / punchy]. No clichés, no "in today's world" openers.
- Open with a concrete hook, not a definition.
- Leave [bracketed placeholders] where I still need to add something.Best for: Beating the blank page — Claude assembles your material instead of making things up.
13. Shorten Without Losing Meaning
Cut this down to a target length.
<text>
[paste text]
</text>
Target length: [100 words / a single paragraph / a tweet].
Constraints:
- Keep every essential fact and the core argument.
- Remove repetition, hedging, and throat-clearing.
- Add nothing new.
- Show the trimmed version and the final word count.Why it works: A hard length target plus "add nothing new" forces real compression instead of a light edit.
Coding & technical
Seven prompts for real engineering. Turn on extended thinking for bugs, architecture, and logic. Ask for output as an Artifact to preview and iterate, or run these in Claude Code against your repo.
14. Review Code for Bugs
Act as a staff engineer doing a careful review. Think step by step.
Language/framework: [e.g. Python 3.12 / React]. What it should do: [behavior].
<code>
[paste the code]
</code>
Do this:
- List concrete bugs and edge cases, most severe first, each with the line and a fix.
- Flag security issues, race conditions, and unhandled errors.
- Suggest 3 tests that would catch these.
- Don't rewrite the whole file — point to the specific changes.Why it works: "Most severe first" plus step-by-step thinking makes Claude prioritize real defects over style nitpicks.
15. Debug an Error Message
Diagnose and fix an error. Think step by step before answering.
<error>
[paste the full error and stack trace]
</error>
<code>
[paste the relevant code]
</code>
Expected: [what should happen]. Actual: [what happens].
Do this:
- Give the most likely root cause first, then 2 alternates.
- Provide the exact fix and where it goes.
- Tell me how to confirm it's resolved.Best for: Getting unstuck — expected-vs-actual plus the full trace gives Claude enough to reason about the real cause.
16. Explain an Unfamiliar Codebase
Help me understand a codebase I just inherited.
I'm a [your level] developer new to this project.
<code>
[paste files, a module, or the whole small repo]
</code>
Give me:
- A high-level map: the main components and how they connect.
- One full request/data flow traced from entry point to output.
- The 5 files I should read first, and why.
- Anything that looks fragile, outdated, or unusual.Why it works: The large context window means you can drop in an entire small codebase and get a real map, not a guess.
17. Write a Function From a Spec
Write a well-tested function from a spec.
Language: [language]. The function should [inputs -> behavior -> outputs]. It must handle [edge cases].
Constraints:
- Clean, commented code following [style/convention].
- Include input validation and clear error handling.
- Add unit tests covering the happy path and every edge case I listed.
- Explain any non-obvious decision in one line.
Put the code in an Artifact so I can iterate on it.Best for: Getting code that matches the spec — the Artifact lets you refine it in place.
18. Build a Small App as an Artifact
Build a working single-file app as an Artifact.
What it does: [describe the tool, e.g. a tip calculator / a markdown-to-HTML converter / a habit tracker].
Users: [who] on [mobile / desktop].
Constraints:
- One self-contained HTML file with inline CSS and JS — no external dependencies.
- Clean, responsive UI; sensible defaults; handle empty and error states.
- Render it as an Artifact so I can preview it.
- After it works, list 3 features I could add next.Why it works: Artifacts render real, runnable apps beside the chat, so you see the result and iterate without leaving Claude.
19. Refactor for Readability
Refactor this code for readability without changing behavior.
<code>
[paste the code]
</code>
Do this:
- Keep the exact same inputs and outputs — behavior must not change.
- Improve naming, structure, and duplication; add comments only where they earn it.
- Show the refactored version, then a short list of what you changed and why.
- Flag anything that looked like a latent bug while you were in there.Best for: Cleaning up inherited code safely — the "behavior must not change" guardrail keeps the refactor honest.
20. Write a Regex From Examples
Build a regex I can trust.
I need to match [describe the pattern] in [language/flavor, e.g. JavaScript, PCRE].
<should_match>
[list examples that SHOULD match]
</should_match>
<should_not_match>
[list examples that should NOT match]
</should_not_match>
Do this:
- Give the final regex and a plain-English breakdown of each part.
- Show it passing every match example and rejecting every non-match example.
- Note any edge case it won't handle.Why it works: Positive and negative examples in separate tags give Claude a precise target to satisfy on the first try.
Research & learning
Six prompts for going deep. Turn on web search for anything current, and extended thinking for synthesis and critique.
21. Research a Topic With Sources
Research a topic and give me a sourced brief. Use web search for anything current.
Question: [specific question]. Purpose: [the decision or output this feeds]. What I already know: [context].
Give me:
- Current state, key players/options, trade-offs, and what's changed recently.
- A comparison table where it helps.
- Sources cited with links for every factual claim.
- A clear recommendation and the 3 open questions worth more digging.
Flag anything you're unsure about instead of guessing.Best for: Decisions that need breadth and sources — the "flag uncertainty" line curbs confident-but-wrong answers.
22. Compare Options in a Table
Compare my options using up-to-date information. Use web search and cite sources.
I'm choosing between [Option A], [Option B], and [Option C] for [purpose]. What matters most: [criteria, e.g. price, features, support].
Do this:
- Build a comparison table scored against my criteria.
- Use current pricing and features — don't rely on memory.
- Recommend one, with a one-line "choose X instead if..." caveat.
- Note anything that changed recently I should know.Why it works: Web search plus citations keeps buying comparisons current and checkable.
23. Learn a Concept With a Study Plan
Build me a study plan to learn something properly.
I want to learn [topic/skill]. Current level: [beginner / some experience]. I can spend [X hours/week] and want to be [target ability] in [timeframe].
Do this:
- Break it into weekly milestones with a concrete outcome each week.
- For each milestone, give one thing to read/watch and one thing to build or practice.
- Point out where people usually get stuck.
- Tell me how I'll know I've actually learned each part.Best for: A schedule you can follow — tied to your real weekly hours and a target ability.
24. Summarize a Research Paper Critically
Read a paper critically, not just briefly.
My background here is [level].
<paper>
[paste or attach the paper / whitepaper / PDF]
</paper>
Give me:
- The core claim in 2 sentences.
- The method and why it matters, in plain language.
- Key findings with the numbers that back them.
- The stated limitations and anything that seems overclaimed.
- 3 questions I could ask to test the paper's strength.Why it works: Asking for limitations and overclaims turns a summary into a critical read, not a shorter copy.
25. Quiz Me With Active Recall
Test my understanding with active recall.
Quiz me on [topic] at [difficulty]. Base the questions on:
<source>
[paste your notes or the material — or say "general knowledge"]
</source>
Rules:
- Ask one question at a time and wait for my answer.
- After each answer, tell me if I'm right, correct any mistake, and explain briefly.
- Mix recall, application, and one "explain why" question per round.
- After 10 questions, summarize what I know well and what to review.Best for: Studying for an exam or interview — active recall beats re-reading, and Claude adapts to your answers.
26. Steelman and Then Critique
Stress-test an argument or plan. Think step by step.
<argument>
[paste the claim, plan, or position]
</argument>
Do this:
- First, steelman it: state the strongest version of the case in 4-5 sentences.
- Then critique it: the weakest assumptions, missing evidence, and failure modes.
- Give the single strongest counterargument.
- End with what would change your assessment either way.Why it works: Forcing the steelman before the critique produces a fair, useful challenge instead of lazy nitpicking.
Business & marketing
Five prompts for founders and marketers. Turn on web search for anything about competitors or markets so the output reflects reality.
27. One-Page Business Plan
Draft a tight one-page business plan.
The business is [what it does] for [target customer], solving [problem]. Model: [how it makes money]. Stage: [idea / early / scaling].
Do this:
- Cover: problem, solution, target customer, revenue model, go-to-market, key risks.
- Keep each section to 2-3 sentences.
- Add 3 assumptions I must validate before investing more.
- Be honest about the weakest part of this plan.Why it works: Forcing honesty about the weakest part surfaces the risk you're most likely avoiding.
28. Marketing Campaign Concepts
Generate campaign concepts I can actually run.
Product: [product]. Audience: [who]. Goal: [awareness / signups / sales]. Budget vibe: [scrappy / mid / big]. Channels available: [list].
Do this:
- Give 5 distinct concepts, each with a hook, the channel, and the core message.
- For each, name the one metric that would prove it worked.
- Rank them by likely ROI for my budget.
- Flag which are quick to test this week.Best for: Getting past the blank page with concepts tied to your real channels and budget.
29. Landing Page Copy
Write conversion-focused landing page copy.
Product: [product]. Audience: [who and their main pain]. Primary action: [what they should do]. Proof I have: [testimonials, numbers, logos].
Do this:
- Write a hero headline, subhead, and 3 benefit-led sections (heading + 2 lines each).
- Add a clear CTA and one FAQ that handles the top objection.
- Lead with the customer's outcome, not our features.
- Give me 2 headline variants to test.Why it works: Naming the primary action and real proof keeps the copy focused on conversion, not vibes.
30. Customer Personas From Real Data
Build grounded personas from real inputs.
My product is [product].
<data>
[paste survey responses, reviews, support tickets, or a table of customer data]
</data>
Do this:
- Identify the 2-3 distinct personas actually present in this data.
- For each: goals, pains, objections, and the trigger that makes them buy.
- Quote 1-2 real phrases from the data per persona.
- Note where the sample is too thin to be confident.Best for: Personas grounded in your real customers — and quoted — instead of invented archetypes.
31. Pricing and Positioning Pressure-Test
Pressure-test my pricing and positioning. Think step by step and use web search for comparables.
Product: [product]. Current price: [price/model]. Positioning line: [how I describe it]. Main competitors: [list].
Do this:
- Tell me who this price signals I'm for, and whether that matches my target.
- Compare against 3 real competitors' pricing and positioning (cite sources).
- Name the biggest risk in my current pricing and one experiment to test it.
- Suggest 2 alternative positioning angles I could own.Why it works: Grounding in real competitor pricing turns a gut-feel debate into a checkable one.
Everyday & personal
Nine prompts for life outside work and for working with data. Attach photos, PDFs, or spreadsheets — Claude reads them directly.
32. Analyze a CSV or Spreadsheet
Analyze a dataset and tell me what matters. Think step by step.
The question I care about is [question].
<data>
[paste the CSV or attach the spreadsheet; describe the columns]
</data>
Do this:
- Start with a quick data-quality check: row count, missing values, obvious anomalies.
- Answer my question with the specific numbers that support it.
- Show the 3 most important findings, ranked.
- Suggest one chart that would communicate the main finding.Best for: Fast insight from a raw export. For a full set, see Claude prompts for data analysis.
33. Find the Story in a Dataset
Find the narrative hidden in my data. Think step by step.
Audience for the story: [exec / team / customers].
<data>
[paste the data]
</data>
Do this:
- Surface the 3 most surprising or decision-relevant patterns.
- For each: the finding, the evidence, and the "so what".
- Note any correlation I'd be wrong to read as causation.
- End with the single headline finding, in one sentence.Why it works: Asking for the "so what" and the causation caveat turns numbers into something you can act on responsibly.
34. Clean and Structure Messy Data
Clean messy data into something usable.
<raw>
[paste the raw data]
</raw>
I want it as [target format, e.g. a table with columns X, Y, Z].
Do this:
- Normalize inconsistent formats (dates, casing, units, spelling).
- Flag rows that look duplicated or suspect instead of silently dropping them.
- Output the clean version as a table I can paste into a spreadsheet.
- List every transformation you applied so I can audit it.Best for: Auditable cleanup — the log of transformations means it isn't a black box.
35. Build a Chart as an Artifact
Turn my data into a clean chart as an Artifact.
<data>
[paste the numbers or a small CSV]
</data>
The point I want the chart to make: [message].
Do this:
- Pick the chart type that best shows that point and say why in one line.
- Render it as an interactive Artifact (single self-contained file).
- Label axes, add a clear title, and format numbers readably.
- Suggest one alternative view of the same data.Why it works: Claude renders working charts in an Artifact, so you see and tweak the visual instead of getting a description.
36. Plan a Trip Itinerary
Plan a realistic trip itinerary. Use web search for current hours, events, and closures.
Destination: [place]. Dates: [dates]. Travelers: [who]. Budget: [range]. We're into [interests] and want the trip to feel [relaxed / packed].
Do this:
- Give a day-by-day plan grouped by area to minimize travel time.
- Include 1-2 food suggestions per day and a rough daily budget.
- Add a rainy-day backup for each day.
- Flag anything that needs booking ahead.Best for: A plan grounded in current, real details rather than generic suggestions.
37. Explain a Bill or Contract
Help me understand a document before I sign or pay.
I'm worried about [your concern, e.g. hidden fees, cancellation terms].
<document>
[attach or paste the contract / bill / policy / lease]
</document>
Do this:
- Summarize what this commits me to, in plain language.
- List every fee, deadline, and obligation as a bullet.
- Flag anything unusual, one-sided, or worth questioning.
- Give me 3 questions to ask before I agree.
Note that this is general help, not legal advice.Why it works: Attaching the full document lets Claude catch the terms buried in fine print you'd skim past.
38. Draft a Difficult Personal Message
Help me say something hard, well.
I need to [message a friend / decline / apologize / set a boundary] with [who]. The situation: [context]. I want them to feel [respected / clear / cared for].
Constraints:
- Honest, kind, and not passive-aggressive.
- Own my part where fair; don't over-apologize.
- Give me two versions: one softer, one more direct.
- Keep each under 120 words.Best for: Naming how you want the other person to feel steers the tone better than describing the situation alone.
39. Build a Habit That Sticks
Help me build a habit that actually sticks.
The habit I want: [habit]. My motivation: [why]. Past failures: [what's gone wrong]. My daily routine: [rough shape].
Do this:
- Anchor the habit to something I already do (habit stacking).
- Start absurdly small and scale over 4 weeks — give me the weekly targets.
- Design one "if I slip" recovery rule so one miss doesn't end it.
- Suggest a simple way to track it.Why it works: Anchoring, tiny starts, and a recovery rule are the parts that survive past week one.
40. Explain This to a Non-Expert
Explain a complex thing simply and accurately.
Explain [concept or text] to [audience, e.g. a smart 12-year-old / a busy executive / a new hire].
Constraints:
- Start with a one-sentence plain-English definition.
- Use one concrete analogy and one real-world example.
- Avoid jargon; if a technical term is unavoidable, define it inline.
- Keep it under 200 words.Best for: Naming a specific audience calibrates the reading level far better than "explain simply".
Once these feel natural, save the ones you reuse inside a Project so Claude keeps the context automatically, and keep the Claude prompt cheat sheet handy. For pre-built skeletons you fill in, grab the Claude prompt templates.
Frequently Asked Questions
Which Claude model should I use for these prompts?
Claude Opus 4.8 is the strong default for hard reasoning, coding, and long tasks. Claude Fable 5 — the top Mythos-class model — is best for the most complex, multi-step work with its always-on adaptive thinking and 1M-token context. Sonnet 5 is the agentic all-rounder for browsers, terminals, and everyday work, and Haiku 4.5 is the fast, cheap option for high-volume simple tasks. The prompt text is identical across models; you pick the model in the app or API.
Why do these prompts use XML tags like <doc> and <code>?
Claude was trained to pay close attention to XML tags, so wrapping pasted material in tags such as <doc>, <code>, <data>, or <transcript> tells Claude exactly where your input starts and ends. It keeps long context organized and makes Claude far less likely to confuse your instructions with the material you want it to work on. Tags are optional but reliably improve results on anything long or structured.
What is extended thinking and when should I turn it on?
Extended thinking lets Claude reason step by step in a visible scratchpad before it answers. Turn it on — or add "think step by step before answering" to the prompt — for math, code architecture, multi-document analysis, and any decision where a wrong answer is costly. For quick rewrites and simple lookups, leave it off for speed. Fable 5 runs adaptive thinking automatically and scales it to the task.
What are Claude Projects and Artifacts?
A Project is a saved workspace that holds shared knowledge (files, style, custom instructions) so Claude keeps context across every chat inside it — ideal for a codebase, a client, or a book. An Artifact is a live, editable panel beside the chat where Claude renders documents, code, and working apps you can preview and iterate on. Several prompts here ask for output as an Artifact.
Can Claude read my files, PDFs, and images?
Yes. Claude is multimodal — attach PDFs, spreadsheets, images, screenshots, and code files directly in the app or API, and it reads them in full. With the large context window you can paste an entire report, contract, or codebase in one message. Reference attachments by name in your prompt so Claude knows which one to use.
Can I use these prompts and Claude's output commercially?
The prompt text is free to use and adapt. Whether Claude's output is cleared for commercial use depends on your Anthropic plan and Claude's current usage policy, so check the terms tied to your account before shipping paid or branded work, and always review and fact-check output before relying on it.
How do I adapt these prompts to my own situation?
Replace every [bracketed placeholder] with your real details, keep the Role, Task, Context, and Format structure intact, and wrap any long input in XML tags. Turn on extended thinking for hard reasoning, save reusable setups as a Project, and ask for output in an Artifact when you want to edit it. Tell Claude what to do rather than what to avoid.
Do I need to write a system prompt to use these?
No. Every prompt here works pasted straight into a normal chat. If you use the API or want a role to persist across a Project, move the "Act as..." role line into the system prompt — Claude weights the system prompt heavily for tone and standards. In the app, the role at the top of the message works just as well for one-off tasks.