A Claude Code skill that runs a structured validation sprint on a product or SaaS idea before you write a single line of code.
It searches Hacker News, Indie Hackers, Product Hunt, YouTube, Reddit, G2, and Capterra for real signals that people share your target problem and are willing to pay to solve it. Then it delivers a clear GO / PIVOT / NO-GO verdict backed by actual quotes, links, and unit economics.
Five research phases, run in order:
- Community signal research — Searches forums and communities for people describing the problem in their own words
- Competitor discovery — Finds 2-4 tools already (partially) solving the same problem
- Competitor review mining — Pulls negative reviews from G2 and Capterra to find where existing tools fail
- Willingness-to-pay research — Looks for evidence people already pay (or would pay) to solve this
- Market economics — Assesses pricing ranges, competition density, and volume math (how many customers to reach a real monthly revenue number)
Every signal gets classified as PASS or FAIL using a strict taxonomy. Vague "I'd use that" comments don't count. Specific tool frustrations, workarounds, and price complaints do.
The final report includes a full signal log with citations, a competitor map, a pattern summary, and the verdict.
Copy SKILL.md into your Claude Code skills directory:
cp SKILL.md ~/.claude/skills/validate-idea.mdOr clone this repo directly into your skills folder:
git clone <this-repo> ~/.claude/skills/validate-ideaClaude Code picks up skills from ~/.claude/skills/ automatically. No restart needed.
This skill calls several MCP tools during research. Make sure these are configured in your Claude Code setup:
| Tool | Used for |
|---|---|
mcp__workspace__web_fetch |
Fetching pages from HN, Indie Hackers, Product Hunt, Reddit, G2, Capterra |
| Web search | Finding Reddit threads, competitor pricing pages, alternatives lists |
mcp__youtube__searchVideos |
Finding videos about the problem space |
mcp__youtube__getTranscripts |
Reading spoken content from relevant videos |
mcp__apify__call-actor |
Mining G2 and Capterra reviews via Apify actors |
If Apify is not configured, the skill falls back to scraping G2 and Capterra pages directly via web fetch. It will note in the output when it used a fallback.
Trigger the skill by describing your idea in natural language. Any of these phrasings work:
Validate this idea: [your idea]
Is there a market for [thing]?
I'm thinking of building [thing], should I?
Research this idea: [description]
Does anyone want [thing]?
Should I build [thing]?
The skill will ask you three questions before starting research if you haven't answered them already:
- The idea — what does it do and for whom
- The target audience — be specific (not "small businesses" but "solo consultants who invoice clients monthly")
- The problem it solves — in plain language
The more precise your answers, the better the search queries and the more useful the output.
Validate this idea: a lightweight invoicing tool for freelance developers
who hate the bloat in FreshBooks and just want to send a PDF and get paid
The skill will research the problem space, find what people complain about in existing tools, check if anyone is paying for alternatives, and return a structured report with a verdict.
The report follows a fixed structure:
- Audience definition — refined from your input based on what the research actually found
- Signal log — numbered list of every signal with source, type, quote, and link
- Competitor map — tools in the space, their pricing, and their main failure modes
- Pattern summary — what the research actually showed, in plain language
- Willingness-to-pay signals — concrete evidence people pay for things in this space
- Market economics — pricing floors and ceilings, competition density, volume math
- Verdict — GO, PIVOT, or NO-GO with a one-paragraph explanation
GO requires at least 8 PASS signals from at least 2 platforms, a clear pattern of shared frustration, WTP evidence, and market economics that work for a solo or small team.
PIVOT means some signal exists but something is off: wrong audience, frustration present but existing tools are good enough, or the unit economics only work at scale.
NO-GO means the signals are weak, WTP is absent, or the problem only affects a tiny fringe. The skill will say this directly and not soften it.