50 Claude prompts that produce real work, not fluff

You are a senior tech journalist who writes for founders and developers.
Tone: direct, opinionated, zero corporate language.
Task: Write a 1,200-word article defending the idea that [TOPIC].
Audience: [FOR EXAMPLE TECHNICAL FOUNDERS AT SERIES A].
Format: one strong opening line, 4 H2 sections, punchy 2-sentence conclusion.
Do NOT include: passive voice, "in today's landscape", generic CTAs, unsourced claims.
Before you write, list the 3 strongest arguments you'll use and the single strongest counter-argument you'll address.

<task>Ghostwrite in my voice.</task>
<samples>[PASTE 3-5 OF YOUR BEST PIECES]</samples>
<analysis_step>
Before you write, analyze the samples and produce:
- 5 sentence-level patterns I use (rhythm, length, openings)
- 3 word tics or phrases I lean on
- 2 things I never do
- My implicit point of view in one sentence
</analysis_step>
<deliverable>Then write [POST/EMAIL/ARTICLE] about [TOPIC]
in [WORD COUNT], matching those patterns.</deliverable>

I'm giving you a draft. Your job:
1. Identify the 3 biggest structural weaknesses (not grammar).
2. Ask me 2 clarifying questions before revising.
3. After I answer, produce a revised version.
Don't revise until I've answered the questions.
[PASTE THE DRAFT]

You are a direct-response copywriter, not an SEO writer.
Topic: [TOPIC]. Audience: [WHO].
Generate 15 headlines across these angles, 3 per angle:
counterintuitive claim, specific number/result,
direct appeal to the audience, cost of inaction,
question they're already asking themselves.
For each headline, add one sentence on the pain it targets.
Do NOT use: clickbait, "ultimate guide", "you won't believe this".

Write a newsletter issue.
Topic: [TOPIC]. Audience: [WHO + WHAT THEY DO ALL DAY].
Length: 400-500 words.
Structure: hook (1 line), context (1 paragraph),
the real insight (2-3 paragraphs),
one concrete takeaway to act on.
Voice rules: short sentences, concrete examples,
no stacked metaphors, no "let's dive in".
End with a single sentence that earns next week's open.

Turn my messy client notes into a case study. [PASTE THE NOTES]
Structure:
- Client + their situation before (2 sentences)
- The problem they couldn't solve (specific, with a cost)
- What they tried that didn't work
- What changed (the real mechanism, not the pitch)
- Result with a specific number
- One quote, 25 words max, that sounds like a real human
Do NOT include: marketing adjectives, "game-changer", vague metrics.

Source: [PASTE THE LONG-FORM CONTENT]
Repurpose it into:
- 1 LinkedIn post (200 words, hook-first, a single insight)
- 1 X thread (7 tweets max, each stands on its own)
- 1 Instagram carousel outline (6 slides)
- 3 email subject lines to send the original to a list
Do NOT just shorten the article. Each format takes
its single best standalone idea, not a summary.

Here's my argument: [PASTE THE ARGUMENT]
Steelman it first, make it stronger than I did (3 sentences).
Then attack it on: logical holes, missing evidence,
stakeholders I'm ignoring, cases where it would fail.
Score the surviving argument from 1 to 10 and explain the score.

You are a senior security engineer.
Review for: SQL injection, XSS, insecure auth, exposed secrets,
CSRF, SSRF, insecure deserialization.
For each issue: severity (Critical/High/Medium/Low),
exact location, why it's dangerous, fixed snippet.
Output: numbered list, severity first, sorted Critical → Low.
If there are no issues in a category, say so.
[PASTE THE CODE]

The following code produces this error: [ERROR + STACK TRACE].
Diagnose the root cause step by step before writing a single fix.
List the 2 most likely causes and how to tell them apart.
Then give the corrected code and explain in 2 sentences
what was wrong and why your fix works.
[PASTE THE CODE]

Review this PR like a senior engineer who values shipping.
Comment on: correctness bugs (must fix),
security/perf (must fix), maintainability (should fix),
style (skip unless it hurts readability).
For each comment: file:line, severity, the exact fix.
End with: Ship / Ship after must-fixes / Send back.
[PASTE THE DIFF]

Refactor this code. Constraints in priority order:
1. Behavior must not change, list the tests that must still pass.
2. Reduce cyclomatic complexity below [N].
3. No new dependencies.
4. Keep the public API stable.
Before refactoring, describe in 3 bullets what makes
the current code hard to maintain.
[PASTE THE CODE]

Generate tests for this function. [PASTE THE CODE]
Cover: happy path, edge cases (empty, null, max, min),
failure modes, concurrency if relevant.
Each test name is a sentence describing the behavior.
Framework: [JEST/PYTEST/ETC].
Flag anything in the function that's hard to test:
that's a design signal, not a test problem.

Document an architecture decision for future engineers.
Context: [WHAT TRIGGERED THIS].
Options considered: [LIST 2-4].
Output an ADR with: Title, Status, Context (what + why now),
Decision (one decisive paragraph),
Consequences (good and bad),
What we explicitly chose NOT to do and why.

I inherited this code. [PASTE THE CODE]
Explain in this order:
1. What it does (plain English, 3 sentences max).
2. Why it's structured this way (best guess at the intent).
3. The 3 things I should be most careful about when modifying it.
4. The 2 questions to ask the former owner.
Do NOT suggest a rewrite. Assume I ship in this file tomorrow.

This code is slow. [PASTE CODE + PROFILING DATA]
Before optimizing:
1. State the real bottleneck from the data.
If it's not enough, tell me what to measure first.
2. List 3 candidate optimizations, ranked by
(expected speedup x ease).
3. Flag the one you'd try first and why.
Then write the optimized version with the before/after in Big-O.

You are a senior data analyst.
Context: [DESCRIBE, what it measures, period, collector, why].
Identify the 3 dominant trends, flag the anomalies,
suggest 2 follow-up analyses.
Format: trend summary (2 sentences each, finding then evidence);
anomaly table (value | why unusual | what to investigate);
follow-ups (the question each follow-up answers).
If the data is too small or biased, say so.
[PASTE THE DATA]

Schema: [PASTE OR DESCRIBE]
Question: [WHAT YOU WANT TO KNOW]
Before writing: restate the question in your own words,
flag the assumptions about the shape of the data.
Then write the query, commented at every join
and every non-obvious WHERE.
Dialect: [POSTGRES/BIGQUERY/ETC].
End with one way the result could be misleading.

You are a senior experimentation analyst.
Inputs: [METRIC, VARIANT A, VARIANT B,
SAMPLE SIZE PER ARM, DURATION].
Tell me:
1. Statistical significance and confidence level.
2. Practical significance given the context: [DESCRIBE].
3. The 3 most likely ways this result is wrong.
4. Ship / kill / keep running, and what would change your mind.
Say "the data doesn't support a decision" if it's true.

Synthesize these customer responses. [PASTE THE RESPONSES]
Output:
- 5 themes ranked by frequency, with a quote (20 words max).
- 3 surprises, things you didn't expect.
- 2 things customers are NOT saying that matter.
- One sentence: what they'd actually pay to solve.
Do NOT smooth over contradictions. Name the segments
where customers disagree.

Shape of the data: [DESCRIBE THE COLUMNS WITH 2-3 EXAMPLE ROWS]
I need a formula that does: [WHAT YOU WANT].
Output:
1. The formula (Google Sheets / Excel, specify: [TOOL]).
2. A plain-English breakdown of each part.
3. The 2 most likely ways it breaks.
4. A simpler alternative if one exists.

You are a senior PM at a B2B SaaS company.
Write a PRD for: [FEATURE NAME].
Include: problem statement (2 sentences, user pain in their words);
3 user stories with trigger/action/success state;
3 success metrics with numeric targets and how to measure them;
2 non-goals; 2 known technical constraints;
1 risk that could kill the feature after launch.
Audience: Engineering. No marketing language. No "delight".

Teardown of [COMPETITOR]. Source: [WEBSITE / REVIEWS].
Output:
- Their positioning in their own words (1 sentence).
- Their real positioning based on what they ship (1 sentence).
- Top 3 strengths.
- Top 3 weaknesses customers complain about.
- 2 wedges we could exploit that play to our strength.
- 1 thing they do better that we should copy.
Compare on jobs-to-be-done, not feature lists.

Prioritize [LIST OF FEATURES/INITIATIVES].
For each, score Reach (1-10), Impact on [METRIC] (1-10),
Confidence (1-10), Effort (engineer-weeks).
Compute RICE = (R x I x C) / E. Sort descending.
Then write 1 sentence per item on what would move its score
(for example "if confidence rises after user interviews, it jumps to #1").

We're shipping [FEATURE]. Define the kill criteria.
- 3 leading indicators it's working (week 1-2).
- 3 lagging indicators (month 1-3).
- The single number below which we kill it.
- The review date and who decides.
- What "kill" means (deprecate / remove / merge into X).
Be honest. Most features should have a defined floor.

Context: We sell [PRODUCT] to [BUYER] who uses it for [JOB].
Current pricing: [PRICING]. Constraints: [LIST].
Walk through:
1. The value metric to price on (vs 2 alternatives).
2. 3 plausible packages with names, prices, target buyer.
3. Anchoring logic for each price point.
4. The single thing to test first.
5. The biggest pricing mistake startups make here.

We're about to launch [INITIATIVE].
Imagine 12 months later, it failed badly.
Trace the failure back:
- One-paragraph post-mortem.
- 5 things that went wrong, in order of impact.
- For each, the warning sign we ignored.
- For each, the cheap thing to do today to avoid it.
Be specific. "We didn't talk to users" is too vague.

Help me write the "why now" narrative for [PRODUCT/COMPANY].
Audience: [INVESTORS / NEW HIRES / ENTERPRISE BUYERS, pick one].
Structure:
- The shift in the world (what changed over the last 2-5 years).
- Why the old solutions can't keep up.
- Why we're the right team for this moment (concrete, no "passion").
- A specific bet we're making that competitors aren't.
Do NOT use: "in today's landscape", "the future of",
"AI-powered" without a specific mechanism.

Cold email to [ROLE] at [COMPANY TYPE].
Goal: [SPECIFIC OUTCOME, for example a 15-min call].
Tone: direct, peer-to-peer, no sales language. Length: under 100 words.
Do NOT include: flattery, "I hope this email finds you well",
feature lists, generic CTA, "quick question".
Include: a specific observation that proves I did my research,
one sentence on why now.
Subject: under 6 words, intriguing not clickbait.

Call with [PROSPECT, name, role, company, what they do].
Goal: [DISCOVERY / DEMO / CLOSE].
Build me:
- 5 questions for the first 10 minutes that go past the surface.
- 3 likely objections and how I'd handle each.
- 2 signals that mean "qualified", 2 that mean "walk away".
- A single sentence I must ask before the call ends.
Bias toward questions that show I understand their business.

Objection: [PASTE VERBATIM].
Don't write the response right away.
First:
1. What is the prospect really worried about? List 3 real concerns.
2. Which is most likely given the context: [CONTEXT].
Then write the response, under 60 words, that addresses the real
concern, not the surface objection.
End with a question that keeps the conversation going.

You are a CRO specialist. [PASTE THE COPY OR THE URL]
Audience: [WHO]. Goal: [PRIMARY ACTION].
Critique:
1. Above the fold: does the visitor know what this is in 5 seconds?
2. The 3 biggest friction points before the CTA.
3. The 3 weakest sentences and why.
4. The single change with the biggest expected uplift.
Don't suggest "add testimonials" unless you name what
the testimonial must specifically prove.

Sharpen the positioning for [PRODUCT].
Inputs: what it does [X]; who buys [WHO];
what they did before [Y]; what we do better [Z];
what we don't do [N].
Output:
1. Positioning: "For [WHO], who [PROBLEM], [PRODUCT] is
the [CATEGORY] that [DIFFERENTIATOR], unlike [ALT]."
2. Three taglines that imply this without saying it.
3. The single line for the homepage, above the fold.

Audit my funnel: [STEPS WITH CONVERSION RATES IF KNOWN].
Tell me:
- The biggest leak (and how the numbers tell you).
- 3 hypotheses about why it's leaking.
- The cheap test to identify which one is right.
- One thing I'm probably obsessing over that doesn't matter.
Be opinionated. Generic CRO advice wastes my time.

Brief a creative team for a paid campaign.
Product: [WHAT]. Audience: [WHO + AWARENESS LEVEL].
Goal: [APP INSTALLS / MQLS / ETC].
Output:
- 3 distinct angles (problem-aware, solution-aware, comparison).
- For each: hook line, 3-line script, CTA, key visual.
- One rejected angle and why.
Each angle must be testable in isolation.

Explain [CONCEPT] to someone who knows [PREREQUISITES]
but has never come across [CONCEPT].
Use a concrete real-world analogy.
Then give an example where the analogy breaks.
Stay under 200 words.
End with the most common misconception about this topic.

I want to learn [SKILL] to the level where I can [OUTCOME].
Starting point: [WHAT I KNOW]. Time available: [HOURS/WEEK].
Build me:
- A 4-phase roadmap, each ending with something I can show.
- The most overrated resource people recommend (and why).
- The most underrated practice habit.
- The first thing to do in the next 2 hours.
Skip "fundamentals matter". Be specific.

I just finished [BOOK / EPISODE]. [PASTE NOTES OR DESCRIBE]
Distill it into:
- The 3 ideas worth remembering 2 years from now.
- The idea that contradicts what most people in [MY FIELD] believe.
- The specific thing to change in how I work this week.
- The weakness in the author's thinking.
No chapter-by-chapter summary.
Give me what survives forgetting.

Compare [CONCEPT A] vs [CONCEPT B] vs [CONCEPT C].
Table rows: one-sentence definition, when to use,
when NOT to use, concrete example,
the trap people fall into when using it.
Then: which one wins for [MY USE CASE: DESCRIBE],
and the condition under which the answer changes (2 sentences).

I want to learn [TOPIC] by being questioned, not lectured.
Rules:
- One question at a time.
- After my answer, tell me what I got right, wrong, and why.
- Don't move on until I've shown I understand.
- If I'm stuck, give a hint, not the answer.
- Increase the difficulty when I start getting it right.
Start with a question that reveals whether I understand the basics
or just have the vocabulary.

Build me a decision filter for [TYPE OF DECISION].
Output:
- 5 yes/no questions in the order I should ask them.
- The rule: a single "no" anywhere = decline.
- For each question, why it matters and the failure mode it prevents.
- The question I'll be tempted to skip but must not.
Short enough to run in 60 seconds without opening a doc.

Build me a time-blocked day plan.
Constraints:
- Energy pattern: [FOR EXAMPLE GOOD 7-11AM, LOW 1-3PM].
- Non-negotiables: [SCHOOL RUN, GYM, CALLS].
- Deep work needed: [HOURS] on [WHAT].
- Admin/comms: [HOW MUCH].
Include: buffers, a protected deep-work window,
a clear shutdown time, one real break.
Flag anything likely to break the plan during the week.

Build a batching system to plan, create, edit
and schedule a full month of [PLATFORM] posts in ONE day.
Constraints: posts/week [N]; format [FOR EXAMPLE CAROUSEL];
content pillars [LIST].
Output:
- Hour-by-hour schedule for the batching day.
- What to prep the day before.
- Where most creators lose their day, and how to avoid it.
- A minimum viable version if I only have 4 hours.

State of my inbox: [TOTAL UNREAD, TYPES, WHAT'S PILING UP].
Build me a 20-minute triage process:
- The order to handle the buckets in.
- 4 actions any email can become
(reply now / schedule / archive / delete).
- The 2-sentence template for "no, but politely".
- The rule for what escalates to my calendar.
End with a habit that keeps this backlog from rebuilding.

Run me through a 30-minute weekly review.
Ask one question at a time:
1. What got finished, and what does that imply?
2. What slipped: priority, capacity, or clarity?
3. What did I learn about how I work this week?
4. What's the one thing for next week that makes everything else less important?
5. What am I spending energy on that I should let go of?
After my answers, summarize 3 commitments for next week
and 1 thing I explicitly choose NOT to do.

We want [GOAL].
First, brainstorm 10 ways to guarantee failure on this goal.
Then, for each failure mode, flip it into a success strategy.
Flag the 3 flipped strategies that are most counterintuitive
but have real upside.
For those 3, name the person or team who should own it.

Decision: [DESCRIBE]
Walk through each stakeholder, specific people, not abstract roles:
- [NAME / ROLE 1]: what they want, what they fear, what they push for.
- [NAME / ROLE 2]: same.
- [NAME / ROLE 3]: same.
Then: where do interests align? Where do they clash hardest?
Whose objection is most likely to kill this?
The single move that addresses it without losing their support.

I'm about to commit to [DECISION + 1 PARAGRAPH OF REASONING].
Argue against it as hard as you can.
- The single strongest argument that I'm wrong.
- 3 assumptions I'm making that could be false.
- The historical case where this decision failed.
- The cheap experiment I could run before committing.
Don't soften it. If the decision survives this, it's stronger.

Problem: [DESCRIBE]. Standard solutions: [LIST].
Get down to first principles:
1. What's actually true about this, independent of industry norms?
2. Which constraints are real vs just convention?
3. If I had no knowledge of how others solve this,
what would the obvious approach be?
4. What does this rebuilt-from-scratch view suggest?
Be honest about which "first principles" are just contrarian takes.

Choose between [OPTION A] and [OPTION B].
Step 1: Steelman A. Build the strongest possible case.
Step 2: Steelman B. Same.
Step 3: For each, the single condition under which it loses badly.
Step 4: Given my context, [CONTEXT], which steelman is
most relevant to my real situation?
Step 5: Your recommendation, with the single thing
that could change it.
Don't soften step 5. Pick one option.