I built an AI agent that tells you to Kill your idea (before you waste months on it)

My first product got 5 free users and $0 after 6 months of building. Classic indie hacker mistake - I never validated the idea. Just assumed people wanted it. I didn't even want it :)

So when I started building a Distribution Framework for technical founders, I realized the system should start before you even write code. Not just "how do I get customers" but "should I build this at all?"

I built an idea validation agent for Claude Code. You type "Validate this product idea" and it does the research you'd normally spend days on:

- Searches Reddit for complaints and "I wish" posts about your problem
- Checks G2 and Capterra reviews for competitor weaknesses
- Looks for pricing signals (do people already pay for solutions?)
- Estimates market size and where potential customers hang out

Then it scores your idea on 8 criteria, weighted by what matters most:

- Pain evidence: 3x weight
- Willingness to pay: 3x weight
- Market size: 2x weight
- Market accessibility: 2x weight
- Competitive gap: 2x weight
- Build feasibility: 2x weight (can a solo dev ship MVP in 2 weeks on <$200/mo infra?)
- Revenue potential: 1x weight
- Personal fit: 1x weight

Total score out of 80. You get a verdict: Strong, Promising, Weak, or Kill.

The best part is the output doesn't just say "good idea" or "bad idea." It gives you concrete next steps:
- 3-5 specific communities to post research questions in
- A post angle that surfaces pain naturally (not "hey would you buy my thing")
- Who to DM for discovery calls
- A landing page headline to test

The whole desk research phase takes 1-2 hours instead of the 2-4 weeks you'd spend building an MVP only to find out nobody cares.

I've been using the two-phase approach: Phase 1 is this automated research. Phase 2 is 3-7 days of manual community validation - actually posting, talking to people, doing discovery calls. The Mom Test stuff.

The agent can't replace talking to real people. But it can tell you whether it's worth starting those conversations in the first place.

This is part of the Distribution Framework which now covers the full loop: validate > set up > find customers > outreach > track > iterate.

Would love to hear how you all validate ideas. Do you do any desk research before building, or do you just ship and see what happens?

PS: The agent is public as a Gist here.