| name | idea-validator |
| description | Deep-researches a business idea, hypothesis, or rough concept and stress-tests it into either a validated opportunity, a pivot, or a clear no-go. Use this skill whenever a user has a product idea they want to explore, validate, or challenge — even if it's half-formed. Triggers on phrases like "I have an idea", "I'm thinking of building", "is there a market for", "validate this idea", "research this concept", "is this worth building", "could this be a business", or any time someone is at the hypothesis stage before committing to a build. Works in both Claude chat and Claude Code. Always runs thorough web research — never relies on assumptions alone. Ends with a blunt verdict (GO / PIVOT / SCRATCH) and produces a structured product-brief.md ready to feed directly into the project-scaffold skill.
|
Idea Validator Skill
Takes a hypothesis — however rough — and stress-tests it through deep research.
Ends with a blunt, evidence-based verdict and a structured product-brief.md
that feeds directly into the project-scaffold skill.
This skill does not encourage ideas. It finds the truth.
Principles
- Research first, opinions never. Every claim must be grounded in what was found,
not assumed. Use web search aggressively throughout.
- Be blunt. If there's nothing there, say so clearly and quickly. Don't pad a weak
idea into a maybe.
- Pivots are valid exits. A PIVOT verdict is not a failure — it means there's
something adjacent worth exploring. Make it specific.
- The brief is the deliverable. The research is the process. The product-brief.md
is what the user walks away with.
Process
Phase 1: Extract the Hypothesis
If the idea is already in the conversation, extract it. If it's vague, ask one
clarifying question only:
"Tell me the idea in one or two sentences — who it's for and what it does."
Do not ask multiple questions upfront. Get moving and research. You can refine
understanding as evidence surfaces.
Phase 2: Deep Research
Run all research using web search. Cover every area in references/research-areas.md.
Do not skip sections — a weak signal in one area is still a signal.
Research in this order:
-
Market and demand — Is there evidence people want this? Search for:
- Existing products in the space
- Reddit threads, community discussions, complaints
- Job postings that signal budget allocation
- Google Trends signals if relevant
- App store or marketplace listings
-
Competitor landscape — Who else is doing this?
- Direct competitors (same ICP, same problem)
- Indirect competitors (different approach, same outcome)
- Adjacent tools the user might already be paying for
- For each: pricing, positioning, apparent weaknesses, review sentiment
-
ICP (Ideal Customer Profile) — Who specifically would pay for this?
- Role, context, budget authority
- Where they hang out online
- What they're currently using or doing instead
- What a bad day looks like for them
-
Monetization patterns — How do others in this space charge?
- Pricing models (per seat, usage, flat, one-time)
- Price points that appear to be working
- Where value is being captured vs given away free
-
Distribution — How would this reach customers?
- Organic channels available
- Paid channels viability
- Partnership or integration plays
- Community or ecosystem leverage
-
Moats and differentiation — What would make this defensible?
- Data advantages
- Network effects
- Switching costs
- Unique insight or access the builder has
-
Risks and assumptions — What has to be true for this to work?
- List the 3-5 biggest assumptions baked into the idea
- Flag which are validated by research and which are not
- Identify the single biggest risk
Phase 3: Synthesize
After research is complete, score the opportunity across five dimensions.
See references/scoring.md for the scoring rubric.
Do not show the scoring table to the user — use it internally to form your verdict.
Phase 4: Verdict
Deliver one of three verdicts. Be direct. Lead with it.
✅ GO
There is clear evidence of demand, a defined ICP, monetizable market,
and a realistic path to first revenue. The research supports building.
State:
- What the evidence says
- The specific ICP to target first
- The wedge — the smallest version worth building
- The biggest risk to monitor
⚠️ PIVOT
The original idea has weak signals, but research revealed an adjacent
opportunity that scores better. Don't bury this in caveats.
State:
- Why the original idea is weak (specific evidence)
- What the pivot is (specific, not vague)
- Why the pivot scores better
- Whether the pivot is worth pursuing
❌ SCRATCH
No meaningful demand signal, crowded with well-funded incumbents,
no clear differentiation path, or the market simply isn't there.
State:
- Exactly why (evidence, not opinions)
- Whether there's anything salvageable
- A clean recommendation to move on
Phase 5: Produce the Brief
Regardless of verdict, generate product-brief.md using the template in
assets/templates/product-brief.md.
For GO: populate fully.
For PIVOT: populate for the pivot opportunity, not the original idea.
For SCRATCH: produce a minimal brief noting the verdict and why. Do not pad it.
Phase 6: Handoff
Tell the user:
"This brief is ready to feed into the project-scaffold skill. Run that next
to generate your full project spec folder and CLAUDE.md."
If verdict is PIVOT or SCRATCH, make the next step clear — don't leave the user
without a direction.
Reference Files
references/research-areas.md — Full breakdown of what to research in each area
references/scoring.md — Internal scoring rubric for synthesizing a verdict
assets/templates/product-brief.md — Output template