Prompts that actually ship work.

12 prompts · all niches

• Strategy · Best with Claude 4.5

  The 'Surgical ICP' interview

  Stop guessing who your customer is. Get a 1-page ICP in 10 min.

  WhyClaude's careful long-context reasoning beats ChatGPT on synthesis-heavy strategic work. The role-play frame forces it to ask, not assume.

  You are a senior B2B strategist who has interviewed 500+ founders. I am about to describe my company in one paragraph.
  
  Your job: ask me 7 sharp questions, ONE at a time, to build a surgical ICP profile. After my 7th answer, output:
  
  1. The ICP in 3 bullets (role, company shape, trigger event)
  2. The single sharpest pain we solve (in their words, not ours)
  3. The 3 objections we'll hit in every sales call
  4. The 1-sentence positioning statement
  5. 5 channels ranked by where this ICP actually hangs out
  
  Rules: never ask multiple questions at once. Never accept a fuzzy answer — push back if I'm vague. Start now with question 1.
  
  COMPANY: {{ONE_PARAGRAPH_DESCRIPTION}}

  Copy promptExpandFill in: {{ONE_PARAGRAPH_DESCRIPTION}}

  #icp#positioning#interview

• Marketing · Best with ChatGPT 5

  Landing page → conversion rewrite

  Paste your URL, get a hero rewrite that actually converts.

  WhyChatGPT 5 with web browsing reads the live page and grounds the rewrite in real copy, not hallucinated assumptions.

  Read this landing page: {{URL}}
  
  Then act as a direct-response copywriter who has shipped 100+ winning hero sections (think Harry Dry, Joanna Wiebe).
  
  Output 3 rewritten hero sections in this format for each:
  
  H1 (under 12 words):
  Subhead (one sentence, name the pain):
  Primary CTA button (2-3 words):
  Why this beats their current copy (1 sentence):
  
  Variants must follow these angles:
  1. Pain-first (name the suffering)
  2. Outcome-first (paint the after-state)
  3. Contrarian (challenge an industry assumption)
  
  Forbidden: "revolutionize", "seamless", "AI-powered", "next-gen", em-dashes, three-word lists.

  Copy promptExpandFill in: {{URL}}

  #copywriting#conversion#hero

• Coding · Best with Claude 4.5

  Brutal PR review

  Paste a diff. Get the review your staff engineer would write.

  WhyClaude's code reasoning is the strongest in 2026. The 'staff engineer' frame raises the bar past surface nits.

  You are a Staff Engineer reviewing a pull request from a mid-level developer. Be direct and specific. No platitudes.
  
  Review this diff and output:
  
  1. BLOCKERS — bugs, security holes, broken contracts (file:line + 1-line fix)
  2. SHOULD-FIX — design smells, missed edge cases, naming (file:line + reason)
  3. NIT — style, but ignore anything a formatter would catch
  4. WHAT'S GOOD — one specific thing worth keeping, so they don't undo it
  
  Rules: no generic advice. Every point quotes a line. If the diff is fine, say "ship it" and stop.
  
  DIFF:
  ```
  {{PASTE_DIFF}}
  ```

  Copy promptExpandFill in: {{PASTE_DIFF}}

  #code-review#engineering

• Research · Best with Perplexity Pro

  Competitor teardown in 5 min

  One prompt → competitive brief you can drop in a deck.

  WhyPerplexity's cited research with Spaces lets you trace every claim back to a source — safe for client decks.

  Research {{COMPETITOR_NAME}} ({{COMPETITOR_URL}}) and produce a 1-page competitive brief.
  
  Sections:
  1. Positioning — one sentence in their words
  2. Pricing — every public tier with $/seat and what's included
  3. Top 3 use cases they emphasise (with source links)
  4. Distribution — where they get their traffic (SEO topics, paid, social, partnerships)
  5. Public reviews — what users praise (3 quotes) and complain about (3 quotes), with links
  6. Funding & headcount if public
  7. Their single weakest spot we can wedge into
  
  Cite every claim. Skip anything you can't source.

  Copy promptExpandFill in: {{COMPETITOR_NAME}} · {{COMPETITOR_URL}}

  #competitive#research#brief

• Sales · Best with Claude 4.5

  Cold email sequence (5 touches)

  A 5-email sequence built around ONE specific signal, not spray-and-pray.

  WhyClaude's tone-control is unmatched for sales copy. The trigger-event frame is what separates ignored from replied-to.

  Write a 5-touch cold email sequence to {{TITLE}} at {{COMPANY_TYPE}} companies.
  
  The trigger event: {{TRIGGER_EVENT}} (e.g. "just raised a Series A", "posted a hiring intent for X role", "switched from competitor Y").
  
  Constraints per email:
  - Under 75 words
  - Subject line under 5 words, lowercase
  - Reference the trigger in email 1
  - No "I hope this finds you well", no "circling back", no "just bumping this"
  - CTA in emails 1, 3, 5 only. Emails 2 and 4 are value-only (insight, resource, or pattern you've seen)
  - Last email is a polite breakup
  
  Voice: a senior peer, not a vendor. We sell {{WHAT_YOU_SELL}}.

  Copy promptExpandFill in: {{TITLE}} · {{COMPANY_TYPE}} · {{TRIGGER_EVENT}} · {{WHAT_YOU_SELL}}

  #cold-email#outbound#sequence

• Content · Best with Claude 4.5

  Blog post → X thread (that doesn't sound AI)

  Repurpose long-form into a thread without the giveaway tells.

  WhyClaude is the only model that reliably avoids the 'AI thread' fingerprints (em-dashes, three-list cadence, 'in essence').

  Turn this blog post into a 9-tweet X thread.
  
  Rules — break any and you fail:
  - Tweet 1 is the hook. Either a contrarian claim, a specific number, or a one-line story open. No "Here are 5 things".
  - No em-dashes anywhere. Use periods.
  - No three-item lists ("X, Y, and Z"). Pick the best one.
  - No "in essence", "let's dive in", "game-changer", "harness", "leverage", "delve", emojis.
  - Tweets 2-8 each make ONE claim and back it with a specific example, number or story.
  - Final tweet: one-line takeaway + soft CTA to the blog URL.
  - Average tweet length: 180 chars. Vary rhythm — short, short, long.
  
  POST:
  """
  {{PASTE_BLOG_POST}}
  """

  Copy promptExpandFill in: {{PASTE_BLOG_POST}}

  #x#repurposing#thread

• Ops · Best with ChatGPT 5

  The Friday founder review

  End your week with a brutally honest 1-page debrief.

  WhyVoice mode + memory makes ChatGPT the best 'thinking partner' for reflective ops work. Run weekly.

  Act as my Chief of Staff. Run my Friday review.
  
  Ask me, one at a time:
  1. What's the single most important thing that moved this week?
  2. What did I commit to that I didn't ship? Why?
  3. What surprised me — good or bad?
  4. What's the one decision I've been avoiding?
  5. What does next week look like if I'm honest — busy or focused?
  
  After my answers, output:
  - The 1 sentence that defines this week
  - 3 risks for next week, ranked
  - 1 commitment for Monday morning (specific, before-noon, single task)
  - A direct challenge to me about the avoided decision
  
  Push back if I'm vague. No therapy talk.

  Copy promptExpand

  #weekly-review#founder#ops

• Design · Best with ChatGPT 5

  Moodboard → creative brief

  Upload 5 reference images. Get a brief any designer can run with.

  WhyChatGPT 5's vision + writing combo turns a Pinterest wall into a structured brief — Claude can't see images as well in 2026.

  I've attached {{N}} reference images. Analyse them as a senior creative director and produce a 1-page brief.
  
  Output sections:
  1. The single common thread across all references (1 sentence)
  2. Visual language — palette (hex), typography style, photography treatment, motion vibe
  3. Emotional brief — what should the viewer feel in the first 2 seconds?
  4. What's NOT in the references — pitfalls a junior designer might add by mistake
  5. 3 hero-shot concepts we could shoot/render this week, each with a one-line description and the reference image # it draws from
  
  Tone: confident, specific, no design clichés ("clean", "modern", "elevated" are banned).

  Copy promptExpandFill in: {{N}}

  #design#brief#moodboard

• Strategy · Best with Claude 4.5

  Pricing page stress-test

  Find the 3 reasons buyers bounce from your pricing page.

  WhyClaude is the most willing to push back. ChatGPT tends to flatter your pricing — Claude will tell you it's confusing.

  Read my pricing page: {{URL}}
  
  Then play 3 personas in sequence, each in their own voice:
  
  PERSONA A — Skeptical CFO at a 200-person SaaS. Cares about per-seat math, predictability, hidden costs.
  PERSONA B — Solo founder who just started a 14-day trial. Cares about: "will I still afford this in month 3?"
  PERSONA C — Procurement at an enterprise. Cares about SSO, SOC 2, custom terms, invoicing.
  
  For EACH persona, output:
  - The 1 sentence on the page that made them hesitate (quote it)
  - The single question they couldn't answer from this page
  - Would they buy? Yes/No/Need-call + 1 sentence why
  
  Then close with the 1 change to the pricing page that would lift conversion the most. Be specific (which element, what to change it to).

  Copy promptExpandFill in: {{URL}}

  #pricing#conversion#audit

• Content · Best with ChatGPT 5

  30-second product video script

  From product URL → a Veo/Sora-ready 30s script with cuts.

  WhyMultimodal + agent mode means ChatGPT can read your page AND write to the format video models actually parse.

  Read {{PRODUCT_URL}}, then write a 30-second video script ready to feed into Veo 4 or Sora 2.
  
  Format: 6 shots × 5 seconds each. For each shot:
  
  SHOT [N] — [duration]
  VISUAL: [one sentence describing the frame — subject, camera move, lighting, lens]
  ON-SCREEN TEXT: [max 6 words OR "none"]
  VOICEOVER: [one sentence, max 12 words]
  SFX: [one sound cue]
  
  Opening shot must stop the scroll in 1 second. Closing shot must show the product + a 3-word CTA.
  
  Constraints: no "feel" words ("amazing", "incredible"). One concrete promise. One reason to believe it.

  Copy promptExpandFill in: {{PRODUCT_URL}}

  #video#script#veo#sora

• Research · Best with Claude 4.5

  Spreadsheet → insight (not just a chart)

  Drop a CSV. Get the 3 insights, not the 30 charts.

  WhyClaude's analysis tool + careful reasoning beats ChatGPT for not over-interpreting noise as signal.

  I'm attaching a CSV: {{DESCRIBE_DATA}}.
  
  Act as a senior data analyst. Do NOT just describe the data. Do this:
  
  1. State the 3 most important questions this data can answer (in plain English)
  2. Answer each one in 1 sentence + the single chart that proves it (build the chart)
  3. Flag any "insight" that's actually noise — small N, confounded variable, or sample bias
  4. End with: "If you only act on one finding, it's this:" + 1 specific recommendation
  
  Skip everything else. No table-of-contents. No "the data shows...". No 12-chart dashboards.

  Copy promptExpandFill in: {{DESCRIBE_DATA}}

  #data#analysis#csv

• Ops · Best with Claude 4.5

  Monthly investor update (5 min)

  Numbers + bullets in → polished investor update out.

  WhyClaude writes the cleanest, most trustworthy founder voice. ChatGPT tends to over-sell.

  Write our monthly investor update for {{MONTH}}.
  
  Tone: calm, specific, founder-voice. No hype. Investors should finish it thinking "they're in control."
  
  Use this structure, no deviations:
  
  KPI Snapshot (table): MRR, new logos, churn $, runway months, cash in bank. Show MoM delta.
  
  WINS (3 bullets, each 1 sentence, each names a specific event or customer)
  LOSSES / LEARNINGS (2 bullets, also specific — no "we're iterating")
  WHAT'S NEXT (3 bullets, each is a commitment with a date)
  ASKS (1-3 specific asks — intros, hires, advice. If none, write "No asks this month")
  
  Inputs:
  {{PASTE_RAW_NUMBERS_AND_NOTES}}

  Copy promptExpandFill in: {{MONTH}} · {{PASTE_RAW_NUMBERS_AND_NOTES}}

  #investor#update#founder