kappaemme-git/codex-first-customer-finder-skill

This project is a skill for Codex, an AI coding assistant, that helps a founder find their first potential customers. You give it a startup's website, a code repository, or just a description of the product, and it researches public sources to build a shortlist of people or companies who might be a good fit to buy or try the product. Instead of guessing, it looks for real public evidence such as someone complaining about a problem, describing a workaround, mentioning they are switching tools, or showing other signs of timing and demand. Each prospect it suggests is linked back to the original public source it came from, so you can verify the evidence yourself rather than trusting a black box score. It also drafts a respectful, source-based opening message for each prospect, but importantly it never sends anything automatically. All outreach stays manual and in your control. The tool can run in different modes depending on what you need. A quick mode returns a handful of strong prospects, a standard mode covers up to ten across multiple source types, and a deep mode goes further with up to twenty prospects and looks for repeated patterns in the pain points found. There are also specialized modes for finding early design partners who are likely to give feedback, for business to business research focused on company-level triggers, and for community mode which looks at public discussion signals. The final output is a polished, standalone HTML report that includes an early-customer verdict, a description of your ideal customer profile, the strongest prospects with their evidence and scores, draft outreach messages, and a seven day manual outreach plan. The README is clear that these are hypotheses based on public signals, not confirmed customers. You install it with a single command that places the skill into your Codex skills folder, then restart Codex to use it. It is released under the MIT license.