cosmicstack-labs/idea-validator
A structured idea testing framework that runs every concept through a reality checklist before you invest time, money, or energy. Combines lean validation, customer discovery, and risk analysis into a single process for solopreneurs, creators, and innovators.
npx skills add https://github.com/cosmicstack-labs/mercury-agent-skills --skill idea-validator
Turns raw ideas into testable hypotheses. Instead of falling in love with an idea and building it in a vacuum, this skill runs it through a validation gauntlet — surfacing assumptions, identifying risks, and designing the smallest possible test to learn whether the idea has legs before you commit significant resources.
┌─────────────┐
│ RAW IDEA │
└──────┬──────┘
│
┌──────▼──────┐
│ CANVAS IT │ ← Problem, Solution, Audience, Channel
└──────┬──────┘
│
┌──────▼──────┐
│ CHECKLIST │ ← 20-point viability screen
└──────┬──────┘
│
┌──────▼──────┐
│ RISK MAP │ ← What could kill this?
└──────┬──────┘
│
┌──────▼──────┐
│ TEST DESIGN │ ← Smallest experiment that teaches most
└──────┬──────┘
│
┌──────▼──────┐
│ GO / PIVOT │
│ / KILL │
└─────────────┘
Rate each factor as ✅ Strong / ⚠️ Moderate / ❌ Weak
Score interpretation:
| Score | Action |
|-------|--------|
| 18-20 ✅ | Strong signal — proceed to test design |
| 12-17 ⚠️ | Mixed — address weak areas before investing |
| < 12 ❌ | High risk — kill or radically pivot |
| Risk | Description | Mitigation |
|------|-------------|------------|
| Solution Risk | You build the wrong thing | Talk to 10 customers before writing code |
| Market Risk | Nobody wants it | Get pre-orders or sign-ups before building |
| Execution Risk | You can't deliver | Audit your skills, time, and resources |
| Timing Risk | Too early or too late | Check search trends, competitor launches, industry shifts |
| Business Risk | Can't make money | Validate willingness to pay early — don't assume |
| Experiment | What It Tests | How To Run | Success Signal |
|-----------|---------------|------------|----------------|
| The Fake Door | Interest | Landing page with "Coming Soon" + email signup | 5-10% conversion from visitors → email |
| The $5 Interview | Problem | Talk to 10 target customers about their pain | 7/10 confirm it's a real problem |
| Social Post | Demand | Post about the problem on LinkedIn/Reddit/Twitter | Engagement, DMs, "I need this" comments |
| Google Trends | Market trajectory | Check trend direction for related keywords | Upward or stable curve |
| Experiment | What It Tests | How To Run | Success Signal |
|-----------|---------------|------------|----------------|
| Pre-sell / Landing Page | Willingness to pay | Landing page with "Buy Now" (actually collect payment or just gauge intent) | 1-3% conversion rate or 10+ pre-orders |
| Concierge MVP | Solution validity | Manually deliver the service for 3-5 beta customers | Customers get value and would pay |
| Waitlist | Demand volume | Collect email signups with a specific promise | 100+ signups from targeted outreach |
| Content Test | Content-driven demand | Write 3 articles about the problem. Track SEO/social traction | Organic traffic, subscribers, inquiries |
| Experiment | What It Tests | How To Run | Success Signal |
|-----------|---------------|------------|----------------|
| Prototype MVP | Product usability | Build the smallest functional version for 10 beta users | NPS > 30, weekly active usage > 3× |
| Paid Pilot | Willingness to pay at full price | Sell to 1-3 customers at full price | Customers pay and don't churn |
| Crowdfunding / Pre-order | Market validation at scale | Kickstarter, Indiegogo, or direct pre-orders | Hits 30% of funding target in first week |
| Phrase | Action |
|--------|--------|
| "Validate this idea..." | Full validation funnel: canvas → checklist → risk map → test |
| "Is this worth pursuing?" | Runs the 20-point checklist |
| "Test this idea..." | Designs the smallest meaningful experiment |
| "What could kill this idea?" | Risk map — identifies top 3 failure modes |
| "Should I build this?" | Pre-validation reality check |
| "Talk me out of this idea" | Devil's advocate — surfaces weakest assumptions |
| "How do I validate [X]?" | Specific experiment design for a specific risk |
| "Is this a good idea?" | Combined checklist + risk map + recommendation |
Get it into a single sentence:
> "I want to build [solution] for [audience] who struggle with [problem], so they can [outcome]."
If you can't complete this sentence, the idea isn't clear enough yet.
List everything you're assuming to be true. Group into:
| Assumption Type | Examples |
|-----------------|----------|
| Problem | "People actually have this problem." "They know they have it." |
| Solution | "My solution solves it." "They'll use it." |
| Market | "There are enough people." "I can reach them." |
| Business | "They'll pay." "The economics work." |
| You | "I can build this." "I'll stay motivated." |
Key insight: The assumption that would hurt most if wrong is your riskiest assumption. Test that one first.
Score all 20 items. Identify which areas score lowest — those are your focus areas for testing.
Follow the experiment tier table above. Rules:
| Result | Decision |
|--------|----------|
| Strong positive signal | Proceed to next tier of testing |
| Mixed / unclear | Revise assumption, test again with a different method |
| Clear negative signal | Kill or pivot. Do not ignore. Sunk cost is not a reason to continue |
Even for ideas you kill, document what you learned. Future you will benefit from knowing:
> Idea: "A paid newsletter teaching designers how to use AI tools"
>
> Clarified: "I want to build a paid weekly newsletter for mid-career designers who struggle with keeping up with AI tools, so they can stay relevant without spending hours researching."
>
> Riskiest Assumptions: 1) Designers are willing to pay for this. 2) I can consistently find valuable content.
>
> Experiment (Tier 1):
> 1. Post on LinkedIn/X: "Designers — what's your #1 question about AI tools?" (tests problem)
> 2. If engagement is strong → create a free 5-email mini-series with a "Subscribe for full version" CTA (tests willingness to pay)
> 3. Success signal: 1,000 free subscribers within 2 weeks
>
> If < 100 signups: Problem not urgent enough or audience not reachable. Kill or narrow niche further.
> If 100-500: Interesting but weak. Try paid ads to test willingness to pay more directly.
> If 500+: Proceed. Build the paid tier.
> Idea: "A tool that auto-generates social media posts from blog content"
>
> Risk Map:
> - Solution Risk (HIGH): Will the output be good enough to use without heavy editing?
> - Market Risk (MEDIUM): Do content creators feel this pain enough to pay?
> - Execution Risk (LOW): Buildable with existing AI APIs
>
> Experiment (Tier 1):
> 1. Build a manual concierge: Offer to take 5 bloggers' posts and manually create their social content for free
> 2. Ask: "If this were automated, what would you pay?"
> 3. Success signal: 3/5 say "I'd pay $20+/month"
>
> Learning: If users love the output but won't pay, pivot to free + ads. If they won't use the output even for free, solution risk is confirmed — kill or redesign.
Take cosmicstack-labs/idea-validator from the repository into ~/.claude/skills for personal
use, or into .claude/skills inside a project.
The agent identifies a skill by the name field in its header. Two skills with the
same name cannot sit side by side — one of them will be ignored.