Skip to main content

CODITECT Idea-to-Questions Prompt Engine

Version: 1.0 | Classification: Prompt Library — AI-Assisted Question Generation Purpose: A series of prompts that transform any raw business idea into Mom Test-compliant interview questions through a predictable, interactive, and engaging workflow.

How This Works​

This document contains 8 sequential prompts designed to be used with an AI assistant (Claude, GPT, or CODITECT's validation agents). Each prompt takes the output of the previous one as context, progressively transforming a raw idea into conversation-ready interview scripts.

Workflow:

YOUR IDEA → [Prompt 1: Decompose] → [Prompt 2: Assumptions] → [Prompt 3: Scary Questions]
         → [Prompt 4: Question Generation] → [Prompt 5: Sequencing]
         → [Prompt 6: Dialog Scripting] → [Prompt 7: Anti-Pattern Audit]
         → [Prompt 8: Interviewer Prep Card]

Usage modes:

  • Manual: Copy-paste prompts one at a time into your AI assistant
  • Chained: Feed all 8 prompts as a system prompt, then provide your idea
  • Agent: CODITECT automation runs the full pipeline and outputs ready-to-use scripts

Prompt 1: Idea Decomposition​

Purpose: Break the raw idea into structured components that can be questioned.

## PROMPT 1: IDEA DECOMPOSITION

I have a business idea that I need to validate through customer interviews. 
I'll describe the idea below. Your job is to decompose it into its core components.

### MY IDEA:
[PASTE YOUR IDEA DESCRIPTION HERE — be as raw and unfiltered as you want. 
Stream of consciousness is fine. The messier, the more honest.]

### YOUR TASK:
Analyze my idea and produce the following structured decomposition:

1. **Problem Statement** (1-2 sentences): What problem does this solve, 
   in the customer's language (not mine)?

2. **Target Customer** (be specific): Who has this problem? 
   Describe them in terms of role, industry, company size, situation — 
   not demographics.

3. **Current Alternative**: What do these people do TODAY to solve this 
   problem (or live with it)?

4. **Proposed Solution** (neutral): What would this product/service do, 
   described without marketing language?

5. **Revenue Model**: How would money flow? Who pays, how much, how often?

6. **Key Assumptions** (list 5-10): What must be TRUE about the world for 
   this idea to work? Be brutally honest. Include assumptions about:
   - Problem existence and severity
   - Customer segment and reachability
   - Willingness to pay and price tolerance
   - Competitive landscape and switching costs
   - Technical feasibility and market timing

7. **Riskiest Assumption**: Which single assumption, if wrong, kills the 
   entire idea?

Format your response as a structured document I can review and refine 
before moving to the next step.

Prompt 2: Assumption Stress Test​

Purpose: Convert each assumption into a testable hypothesis with clear pass/fail criteria.

## PROMPT 2: ASSUMPTION STRESS TEST

Here is the decomposed idea from Step 1:
[PASTE PROMPT 1 OUTPUT HERE]

### YOUR TASK:
For each of the Key Assumptions identified, create a testable hypothesis with:

1. **Hypothesis**: Rewrite the assumption as a falsifiable statement.
   Format: "We believe [customer segment] has [problem] because [reason], 
   and they would [behavior] if [condition]."

2. **Kill Criteria**: What evidence would DISPROVE this hypothesis? 
   Be specific. "If 7 out of 10 interviewees say [X], this assumption is dead."

3. **Validation Signal**: What evidence would SUPPORT this hypothesis? 
   Use behavioral indicators, not opinions.
   Bad: "They say they like the idea"
   Good: "They describe spending 5+ hours/week on this problem and have 
   already tried 2+ solutions"

4. **Question Priority**: Rate each hypothesis as:
   - 🔴 MUST VALIDATE FIRST (idea dies without this)
   - 🟡 VALIDATE EARLY (important but not fatal if wrong)
   - 🟢 VALIDATE LATER (nice to know, not blocking)

5. **Evidence Needed**: How many conversations would give you confidence? 
   (Typically 5-10 for convergence on a clear signal)

Organize the output as a validation priority matrix, 
sorted by priority (🔴 first).

Prompt 3: The Three Scary Questions​

Purpose: Generate the questions you're most afraid to ask — the ones whose answers could kill your idea.

## PROMPT 3: THE THREE SCARY QUESTIONS

Here is the validation priority matrix from Step 2:
[PASTE PROMPT 2 OUTPUT HERE]

### YOUR TASK:
Generate my "Big 3" scary questions — the questions I MUST ask in every 
single conversation because the answers could kill my idea.

**Rules for Scary Questions:**
- They must make me uncomfortable to ask
- The "wrong" answer would force me to pivot or kill the idea
- They must pass the Mom Test (their life, past specifics, listen)
- They cannot mention my product, idea, or solution
- They should be open-ended and invite stories

**For each Scary Question, provide:**

1. **The Question** (exact wording I should use)
2. **What I'm Really Testing** (which assumption this targets)
3. **Killer Answer** (the response that means my idea is dead)
4. **Validation Answer** (the response that means I'm onto something)
5. **Probe Sequence** (2-3 follow-up questions to dig deeper regardless 
   of their initial response)
6. **Emotional Prep** (how to stay neutral when I hear an answer I 
   don't want to hear)

**Format:**

### Scary Question 1: [THE QUESTION]
- **Testing:** [assumption]
- **Kills idea if:** [specific response pattern]
- **Validates if:** [specific response pattern]
- **Follow-ups:**
  1. [probe 1]
  2. [probe 2]
  3. [probe 3]
- **Stay calm:** [coaching note for the interviewer]

Also generate 2 "bonus scary questions" that probe second-order risks 
(competitive, timing, regulatory, etc.).

Prompt 4: Full Question Library Generation​

Purpose: Generate the complete question set organized by the 7-layer framework, customized for this specific idea.

## PROMPT 4: FULL QUESTION LIBRARY

Here is the idea decomposition, validation matrix, and scary questions 
from Steps 1-3:
[PASTE ALL PREVIOUS OUTPUTS HERE]

### YOUR TASK:
Generate a complete Mom Test question library for this idea, organized by 
the 7 conversation layers. Every question must pass all three Mom Test rules.

**For each question, provide:**
- The exact question wording (conversational, natural language)
- Which assumption it tests
- What a "good" response sounds like (behavioral signal)
- What a "bad" response sounds like (noise, fluff, politeness)
- A suggested follow-up probe

**Layers to cover:**

### Layer 1: World Discovery (4-6 questions)
Open-ended questions about their world. No mention of your problem area 
until THEY bring it up.

### Layer 2: Problem Excavation (6-8 questions)
Once they've raised the problem area, dig into frequency, severity, 
cost, and emotional weight. All past-specific.

### Layer 3: Current Solution Mapping (4-6 questions)
What they use today, what they pay, what they love, what they hate. 
This is your competitive intelligence.

### Layer 4: Active Search Behavior (3-5 questions)
Are they actively seeking better solutions or passively tolerating?
This separates buyers from complainers.

### Layer 5: Decision Architecture (3-5 questions)
Who decides, who pays, how purchases happen. Essential for B2B.

### Layer 6: Commitment Extraction (3-4 questions)
Ask for concrete action today. Time, reputation, effort, money.

### Layer 7: Network Expansion (2-3 questions)
Generate your next conversations from this one.

**ALSO GENERATE:**
- 5 "Deflection Responses" — what to say when they give you a compliment 
  instead of data
- 5 "Fluff Anchors" — what to say when they give you a hypothetical 
  instead of a specific
- 3 "Abort Signals" — how to recognize when this person isn't your 
  customer and gracefully exit

**Total output: 30-45 questions with full annotation.**

Prompt 5: Conversation Sequencing​

Purpose: Arrange the questions into a natural conversation flow with timing, transitions, and branching logic.

## PROMPT 5: CONVERSATION SEQUENCING

Here is the full question library from Step 4:
[PASTE PROMPT 4 OUTPUT HERE]

### YOUR TASK:
Arrange these questions into a 15-20 minute conversation script with:

1. **Conversation Map** (visual flow)
   Create a decision tree showing:
   - Opening (2 min): How to start naturally
   - Discovery (5-8 min): Wide-open questions with branch points
   - Excavation (5-8 min): Deep-dive triggered by their responses
   - Commitment (2-3 min): Close with action
   - Expansion (1-2 min): Next conversations

2. **Branch Points**
   At 3-4 key moments, the conversation can go different directions 
   based on their responses. For each branch:
   - "If they mention [problem area]" → Go to [question set]
   - "If they DON'T mention it" → Ask [prompt question], then decide
   - "If they seem disengaged" → [Abort or redirect strategy]

3. **Transition Phrases**
   Natural language bridges between question layers:
   - "That's interesting — you mentioned [X]. Tell me more about that."
   - "I want to make sure I understand — so last time, you..."
   - "Going back to something you said earlier about [X]..."

4. **Timing Guide**
   - Mark each section with approximate time allocation
   - Flag the 3 "must-ask" questions that cannot be skipped
   - Identify 5 "nice-to-ask" questions that can be dropped if time is short

5. **Energy Map**
   Note the emotional arc of the conversation:
   - Start warm and casual (low pressure)
   - Build engagement through storytelling prompts (medium)
   - Hit the hard questions when rapport is established (high)
   - Close with gratitude and concrete next steps (warm)

Prompt 6: Interactive Dialog Scripting​

Purpose: Create a realistic, engaging dialog script that interviewers can practice with — including expected responses, curveballs, and recovery techniques.

## PROMPT 6: INTERACTIVE DIALOG SCRIPTING

Here is the sequenced conversation from Step 5:
[PASTE PROMPT 5 OUTPUT HERE]

### YOUR TASK:
Create 3 complete dialog scripts — realistic conversations that show 
how this interview plays out with different customer archetypes.

**Script 1: THE IDEAL CUSTOMER**
A person who has the problem, knows they have it, has tried to solve it, 
and is actively seeking better options. This conversation should model 
what "good data" sounds like — rich behavioral facts, specific numbers, 
emotional engagement.

**Script 2: THE POLITE LIAR**
A person who doesn't really have the problem (or doesn't care) but is 
too nice to say so. They give compliments, hypothetical enthusiasm, and 
vague feature requests. The script should show the interviewer detecting 
and deflecting these false signals.

**Script 3: THE CURVEBALL**
A person who has a DIFFERENT problem than expected — one that's 
potentially more valuable. The script should show the interviewer 
following the signal rather than forcing the conversation back to their 
original hypothesis.

**For each script, include:**
- Character profile (role, situation, hidden motivations)
- Full dialog with stage directions (what the interviewer should be 
  thinking at each moment)
- Annotations marking:
  - 🟢 Good data (behavioral facts, specific numbers, past events)
  - 🔴 Bad data (compliments, hypotheticals, fluff)
  - 🔄 Deflection moments (interviewer redirecting from bad data to good)
  - 💡 Insight moments (interviewer learning something unexpected)
  - 🤝 Commitment moments (interviewer asking for action)
- Post-conversation debrief: What was learned? What Tier is each data 
  point? What changed about our beliefs?

**Tone:** Make these scripts feel REAL and HUMAN. Include:
- Awkward pauses
- Tangents that the interviewer navigates
- Humor and warmth
- Moments where the interviewer almost pitches but catches themselves
- Natural conversation rhythms (not robotic Q&A)

These scripts will be used for team practice sessions before real interviews.

Prompt 7: Anti-Pattern Audit​

Purpose: Review all generated questions for Mom Test violations and fix them.

## PROMPT 7: ANTI-PATTERN AUDIT

Here is the complete question library and dialog scripts from Steps 4-6:
[PASTE PREVIOUS OUTPUTS HERE]

### YOUR TASK:
Audit every question and dialog line for Mom Test violations. 
Be ruthless — even subtle violations matter.

**Check for these anti-patterns:**

1. **The Pitch Leak**: Any moment where the interviewer reveals their 
   idea, solution, or feature set
2. **The Hypothetical Trap**: Any question using "would", "could", 
   "might", "imagine if"
3. **The Compliment Hook**: Any question designed to elicit approval or 
   positive validation
4. **The Leading Question**: Any question where the "correct" answer is 
   obvious from the phrasing
5. **The Premature Zoom**: Diving into problem details before confirming 
   the problem matters
6. **The Feature Fish**: Asking customers to design your product 
   ("What features would you want?")
7. **The Price Fantasy**: Asking hypothetical pricing questions 
   ("Would you pay $X for Y?")
8. **The Binary Trap**: Questions that can be answered yes/no without 
   revealing anything useful
9. **The Opinion Collector**: Questions that seek beliefs rather 
   than behaviors
10. **The Emotional Manipulation**: Questions that make the customer 
    feel bad for not agreeing

**Output format:**

For each violation found:
| Location | Violation Type | Original Text | Fixed Version | Explanation |
|----------|---------------|---------------|---------------|-------------|

Also provide:
- **Violation count by type** (shows systemic weaknesses)
- **Overall compliance score** (% of questions that pass all 3 rules)
- **Top 3 patterns to watch** for this specific idea (the violations 
  you're most likely to make in live conversation)

Prompt 8: Interviewer Prep Card​

Purpose: Generate a pocket-sized reference card for the interviewer to use during conversations.

## PROMPT 8: INTERVIEWER PREP CARD

Here is the complete, audited output from Steps 1-7:
[PASTE ALL PREVIOUS OUTPUTS HERE]

### YOUR TASK:
Create a single-page "Interviewer Prep Card" that the interviewer can 
glance at during or before conversations. It must fit on one printed page 
(or one phone screen) and contain:

**Section 1: BEFORE (30 seconds)**
- Our segment: [one sentence]
- Our Big 3 Scary Questions: [3 bullet points]
- Today's specific learning goal: [one sentence]

**Section 2: DURING (reference while talking)**
- Opening line: [exact words to start]
- 5 must-ask questions: [numbered, brief]
- Deflection toolkit:
  - If they compliment → [response]
  - If they give fluff → [response]
  - If they pitch back → [response]
  - If they go silent → [response]
- Commitment ask: [exact closing words]

**Section 3: AFTER (2 minutes post-conversation)**
- Quick-capture template:
  - Top 3 quotes (exact words)
  - Behavioral facts learned (Tier 1 only)
  - Commitment received (ladder level)
  - Belief update: What changed?
  - Next action: Who to talk to next?

**Section 4: SIGNALS LEGEND**
Quick-reference symbols for note-taking:
- 🎯 = Confirmed behavioral fact (Tier 1)
- ⚡ = Emotional signal worth probing
- 💰 = Pricing/budget data point
- 🚫 = Fluff/noise — discard
- 🤝 = Commitment signal
- 🧟 = Zombie indicator (no advancement)
- ❓ = Follow-up needed
- 💡 = Unexpected insight — explore

Format this as a clean, printable markdown document with clear visual 
hierarchy. It should be usable by someone who has never read The Mom Test 
but has had a 5-minute briefing.

Meta-Prompt: The One-Shot Pipeline​

For teams that want to run the entire pipeline in a single interaction:

## THE ONE-SHOT PIPELINE

I need to validate a business idea through customer interviews using the 
Mom Test methodology. I'll describe my idea below. Please run the 
complete CODITECT validation pipeline:

1. Decompose my idea into assumptions
2. Generate testable hypotheses with kill criteria
3. Create my 3 Scary Questions
4. Build a 30-45 question library across 7 conversation layers
5. Sequence into a 15-20 minute conversation flow
6. Script 3 realistic practice dialogs (ideal customer, polite liar, curveball)
7. Audit everything for Mom Test violations
8. Produce my one-page interviewer prep card

### MY IDEA:
[DESCRIBE YOUR IDEA HERE]

### MY TARGET CUSTOMER:
[WHO YOU THINK YOUR CUSTOMER IS — be specific about role, industry, 
company size, situation]

### WHAT I'M MOST UNCERTAIN ABOUT:
[THE THING THAT KEEPS YOU UP AT NIGHT ABOUT THIS IDEA]

### CONSTRAINTS:
- Industry: [your industry]
- Stage: [pre-idea / pre-product / pre-revenue / growth]
- Interview format: [in-person / video / phone / casual]
- Time per conversation: [10 / 15 / 20 / 30 minutes]

Please produce all 8 outputs as a single, structured document with 
clear section headers.

Prompt Chaining Notes​

For AI assistants: When running these prompts sequentially, maintain context across the chain. Each prompt builds on the previous. The full chain produces approximately 5,000–8,000 words of output.

For CODITECT agents: This prompt chain can be automated as a validation workflow:

  • Input: Idea description + target segment + constraints
  • Processing: Sequential prompt execution with intermediate validation
  • Output: Complete interview kit (questions, scripts, prep card)
  • Checkpoint: Human review before first conversation

Quality gate: Before using generated questions in real conversations, have at least 2 team members role-play through Script 2 (The Polite Liar). If the interviewer can't detect and deflect false signals in practice, they're not ready for the real thing.

CODITECT Prompt Engine v1.0 — Transforms any idea into validated interview scripts. Compatible with Claude, GPT-4, or any instruction-following language model.