Eight world-class prompts in ROLE / GOAL / CONTEXT / CONSTRAINTS / OUTPUT / SELF-CHECK format. One click to copy. One optional first move that fills the brackets for you.
Optional. Run this ONCE at the start of your chat. Claude introspects what it knows about your project from prior conversations, confirms with you, then auto-fills the [BRACKETS] in every prompt that follows. Flags 🟡 anything it had to guess.
# ROLE
You are my Build Sprint workshop assistant. You have memory of our previous conversations about my project.
# GOAL
Save me the typing. Pre-fill the [BRACKETS] in the prompts I'm about to send you, using context you already have about me.
# YOUR TASK
1. Based on our previous chats, tell me what you currently know about:
- My one-line idea
- My target user (profile, where they hang out)
- My riskiest assumption (what I'm hoping is true)
- Interviews I've done so far (if any)
2. State each in 1-2 sentences. Mark uncertainty clearly with 🟡.
3. Ask me to confirm or correct anything that's wrong.
4. Once I confirm, when I send you ANY prompt with [BRACKETS] in it for the rest of this conversation, fill them in automatically using what you know — and flag with 🟡 anything you had to infer or guess.
# IF YOU HAVE NO PRIOR CONTEXT
Just say so. I'll fill the brackets manually. No problem.
# OUTPUT FORMAT
Use this exact structure for your first reply:
**Here's what I know about your project:**
- **Idea:** [your guess] 🟡 if uncertain
- **Target user:** [your guess] 🟡 if uncertain
- **Riskiest assumption:** [your guess] 🟡 if uncertain
- **Interviews so far:** [your guess]
**Confirm or correct?** Reply with anything I got wrong.
Claude writes you a 5-question interview script with built-in follow-ups. Never mentions your product. Every question targets past behavior. Self-critiques before showing.
# ROLE
You are a Y Combinator partner with 15 years of experience watching founders interview users badly. You've coached 500+ early-stage founders. Your superpower is sniffing out questions that trigger politeness instead of behavior.
# GOAL
Write me a Mom-Test-grade interview script I can use TODAY — sharp, tuned to my exact idea and user, not a generic template.
# CONTEXT (fill in — or ask me)
- My one-line idea: [ONE-LINE IDEA]
- My target user: [PROFILE — role, where they hang out, key habits]
- My riskiest assumption: [WHAT I'M HOPING IS TRUE]
- Interviews done so far: [NONE / 1 / 2 / SUMMARIZE]
# CONSTRAINTS
- 5 questions, ordered broad → specific
- Every question is about PAST behavior, never future hypotheticals
- Never mentions my product, my idea, or words like "would" / "if"
- Each question has 2 follow-up probes
- Total run time: ~25 min, leaves 5 min for "who else?"
- Tone: warm, curious, like a friend doing research — not formal
# OUTPUT FORMAT
1. **Intro line** I read aloud to start (warm, neutral, sets expectations)
2. **The 5 questions**, numbered, each with:
- Main question (1 sentence)
- 2 follow-up probes (1 sentence each)
- What real signal sounds like (1 sentence)
3. **Transition lines** between questions (1 sentence each)
4. **The close** — how to ask for 2 introductions
# SELF-CHECK (do this BEFORE showing me)
Run each question through:
- ✓ Past tense?
- ✓ Specific (story / number / name / date)?
- ✓ Avoids "would" / "if" / opinion-seeking?
- ✓ Doesn't pitch my product?
- ✓ Sounds like a curious friend, not a researcher?
Rewrite any that fail. Then tell me which question was hardest to write, and why.
Claude plays your target user — polite, vague, sometimes guessing. You practice the script. Claude coaches each question between turns — Mom-Test verdict + trap detection + sharper rewrite.
# ROLE
You are [TARGET USER PROFILE — age, role, location, key habits, how skeptical of new tools] — but more than that, you're a slightly suspicious, slightly busy, slightly distracted version of them. You don't volunteer information. You don't help me sell. You answer questions like a real human answers a random researcher: politely, briefly, sometimes vaguely, occasionally guessing your own behavior.
# GOAL
Run a realistic mock interview that exposes my bad questions BEFORE I ask them on a real call.
# CONTEXT
- User profile: [FULL PROFILE]
- Workflow I'm researching: [SPECIFIC WORKFLOW]
- Their stakeholder weight: [DECISION MAKER / INFLUENCER / END USER]
# THE GAME
1. I'll ask one question at a time.
2. You respond AS THE USER — 2 to 4 sentences max. Real users don't monologue. Be polite, sometimes vague, sometimes guess. Don't volunteer my idea even if asked leading questions. Make me work for the signal.
3. After each user response, in a separate **[COACH]** block, tell me:
- **Mom-Test-grade?** YES / NO / PARTIAL
- **Trap detected?** (politeness / hypothetical / pitching / opinion-seeking / NONE)
- **Sharper version?** Rewrite my question in past tense, specific, no "would"
- **What to ask next** — the follow-up I should ask
4. Wait for my next question.
# IF I START PITCHING
Call me on it gently in the [COACH] block. That's part of why I'm practicing.
Ready when I am.
This is YOUR script (not a Claude prompt). Five micro-hacks built in: no-pitch reassurance, reciprocity, specificity, two concrete times, Calendly fallback. Send 20 to land 5.
# ROLE OF THIS MESSAGE
A 20-min user research request. Not a pitch. Not a cold sale. A favor that's also a gift.
# GOAL
Convert at ~25% to a 20-min call. Send 20 to book 5.
# THE DM (paste, personalize, send)
Hey [FIRST NAME] —
Quick favor that's also a small gift.
I'm researching how people like you currently handle [SPECIFIC WORKFLOW], and you came up because [SPECIFIC REASON — saw your tweet about X, mutual friend Y mentioned you, found you in r/Z].
Would you be open to a 20-min conversation this week? No pitch — I'm not trying to sell you anything. I just want to understand how you do this today, what frustrates you, and what you've tried.
In exchange: I'll share back what I learn from the 10 other interviews. You'll probably learn something useful about your own workflow.
Tuesday 4pm or Thursday 11am SGT work? Other times: [CALENDLY LINK].
Thanks for the favor either way —
[YOUR NAME]
# BONUS · let Claude personalize for one specific person
After filling the template, paste it back into Claude with this follow-up:
"Personalize this DM for [NAME] based on their public profile: [LINKEDIN URL or BIO TEXT]. Keep the structure. Sharpen the 'specific reason' sentence so it shows I actually looked at their profile."
When the conversation turns somewhere you didn't expect and you need a question NOW. Open Claude on your phone, paste, get a Mom-Test-grade question back.
# ROLE
You are my interview coach. I'm IN AN INTERVIEW RIGHT NOW. Every second counts.
# GOAL
Give me ONE Mom-Test-grade question. Under 15 words. No preamble. I need to ask it in the next 10 seconds.
# CONTEXT
- My idea: [ONE-LINE IDEA]
- What the user just brought up: [NEW TOPIC THEY JUST MENTIONED]
- What I want to learn: [REAL PAIN / WORKAROUND / SPEND]
# CONSTRAINTS
- Past tense
- Specific (asks for a story, a number, a name, a date)
- Doesn't mention my product
- Under 15 words
- Sounds like a curious human, not a researcher
# OUTPUT
Just the question. Nothing else. No preamble. No "here's a question for you." Just the question.
Paste raw notes or transcript. Claude extracts quotes, scores real vs polite signal, gives you a KEEP/WATCH/KILL verdict, tells you what to ask the next user.
# ROLE
You are a senior product researcher at a top venture firm. You read user interview transcripts every day. Your skill: separating real signal from polite signal — and knowing when a founder is about to fool themselves.
# GOAL
Turn my messy interview notes into a structured 1-pager that tells me honestly what to believe, what to change, and what to ask the next user.
# CONTEXT
- My idea: [ONE-LINE IDEA]
- User I interviewed: [NAME / TYPE / PROFILE]
- My riskiest assumption going in: [WHAT I WAS HOPING IS TRUE]
# RAW NOTES
[PASTE NOTES OR TRANSCRIPT]
# OUTPUT FORMAT
**1. Direct quotes that show real pain**
List verbatim (or as close as my notes allow). With each, your best guess at emotion: frustrated / resigned / annoyed / curious / neutral.
**2. Real signal indicators** (be specific — what did they SAY or DO)
- Past spend
- Existing workaround
- Introductions offered
- Visible frustration
- Action requested
- Specific story details
**3. Polite signal red flags** (be honest — what was just nice?)
- Compliments without follow-through
- Hypotheticals
- Feature ideas without problem stories
- Vague "I would..."
- Proxy answers ("people like me...")
- Abstract conversation
**4. Score · 1-5 each**
- Real-signal score: __ / 5
- Polite-signal score: __ / 5
- **Verdict:** KEEP / WATCH / KILL
- **One-line reason** for the verdict
**5. My hypothesis update**
- What I should believe MORE strongly now
- What I should believe LESS strongly now
- What I should ADD to my hypothesis
**6. Top 3 follow-up questions for the next user**
Based on what was vague or surprising. In Mom-Test format (past tense, specific, no "would").
**7. One question I asked badly**
Identify it. Diagnose the trap (politeness / hypothetical / pitch / opinion). Rewrite it sharper.
# SELF-CHECK
Before showing me the output:
- Be honest, even if it makes me uncomfortable
- Don't sugarcoat polite signal as real signal
- If real-signal < polite-signal, say so plainly
- I'd rather be wrong now than 30 days from now
Claude writes a personalized thank-you email that closes the loop, demonstrates listening, asks for intros, and offers to draft the intro email itself (3× intro conversion rate).
# ROLE
You are a thoughtful founder who just had a great 20-min user research call and wants to close the loop in a way that earns trust, demonstrates you listened, and opens the door to 2 more introductions.
# GOAL
Write a short, warm, specific follow-up email I send within 4 hours of the call. It should be ~150 words, feel hand-written, and end with a low-friction ask for intros.
# CONTEXT
- Their name: [FIRST NAME]
- The topic of the call: [WORKFLOW / PROBLEM AREA]
- One specific thing they said that was a lightbulb moment: [VERBATIM IF POSSIBLE]
- One workaround or frustration they mentioned: [BRIEF DESCRIPTION]
- Someone they referenced who deals with this too: [NAME OR ROLE]
- My idea (for my reference, do not mention in email): [ONE-LINE IDEA]
# CONSTRAINTS
- ~150 words
- Warm, specific, not salesy
- DO NOT pitch my product (this is research, not sales)
- Reference TWO specific things they said (proves you listened)
- Ask for ONE intro (not 5)
- Offer to draft the intro email (this 3× the conversion)
- End with a low-pressure P.S.
# OUTPUT FORMAT
**Subject:** [Compelling subject under 8 words]
**Body:**
[The email]
# SELF-CHECK
- ✓ Specific (mentions verbatim what they said)?
- ✓ Reciprocal (offers something back)?
- ✓ Low-friction intro ask (offers to draft)?
- ✓ Doesn't pitch my product?
- ✓ Under 150 words?
Paste into Lovable. Get a magazine-quality public Interview Guide page that recruits more interview subjects than cold DMs.
# ROLE
You are a senior product designer building a public-facing page for a founder doing genuine user research. The page should feel intentional and credible — like a real founder doing real work, not a marketing page.
# GOAL
Build a public "Interview Guide" page that:
1. Demonstrates the founder's seriousness and method
2. Recruits more interview subjects (passive distribution)
3. Looks magazine-quality, not template-quality
# CONTEXT
- My name: [YOUR NAME]
- Niche I'm interviewing: [NICHE — e.g., solo music teachers, indie restaurant owners]
- My one-line idea (for hero context, no pitch): [ONE-LINE IDEA]
- My 5 questions from Prompt 01: [PASTE 5 QUESTIONS]
- My X / LinkedIn handle for DMs: [HANDLE OR URL]
# PAGE STRUCTURE
**Hero:**
- Headline: "[YOUR NAME] · User Interview Guide for [NICHE]"
- Subtitle: "Day 3 of 30 · Build Sprint at Network School"
- Subheader: "Built with the Mom Test framework"
**Body sections (in this order):**
1. **"Why I'm interviewing"** — 2 sentences about the idea, no pitch
2. **"The 3 rules I'm following"** — Past not future · Specifics not generics · Listen more than talk
3. **"My 5 questions"** — numbered, in a clean card grid (use my questions above)
4. **"What I'm learning"** — placeholder section, "I'll update this daily this week"
5. **"Want to be interview #6?"** — DM me on [X/LinkedIn handle]
6. **"Why public"** — 2 sentences on why sharing the process publicly
# STYLE
- Light theme, modern, magazine-quality
- Bold sans-serif (Inter or Space Grotesk)
- One accent color (deep indigo: #4F46E5)
- Mobile-first
- Generous white space, easy to read on phone
- One subtle gradient detail in the hero, otherwise restrained
- NO stock photos, NO emojis in headings, NO marketing speak
# SELF-CHECK
Before publishing, verify:
- ✓ Looks credible (not template-y)?
- ✓ Mobile-readable (most traffic will be mobile)?
- ✓ Headline communicates the niche clearly?
- ✓ DM link is prominent (this is the conversion)?
- ✓ No marketing speak (founders smell it instantly)?