How to name your project the way naming studios do (free Claude skill)

#ai #claude #opensource #branding

Naming is the decision you can't easily take back, and it's the one most of us rush. Think for ten minutes, grab the first free word, regret it in a month when it turns out forgettable or already taken.

The studios that name things for a living don't work that way. They run a careful process: lock the strategy first, generate widely, screen hard for trademark and language problems, then score what's left. The catch is they charge thousands for it.

I wanted that process without the price tag, so I built it into a Claude Code skill called Nomira. It's free, it's open source, and it works well enough that it named itself. Here's how to use it, and how it works under the hood.

Install

You'll need Claude Code. Then the one-liner:

npx skills add bazingga08/nomira

Open Claude Code, type /nomira, describe what you're naming.

Try it

/nomira

a budgeting app for freelancers who hate spreadsheets.
calm and trustworthy, not a flashy finance-bro vibe.

A couple of questions later, you get a ranked shortlist, each name with a reason, sound notes, and a trademark risk flag.

What it actually does

It runs the same stages a studio would:

Strategy first. It locks the one idea the name has to carry before generating anything. A vague brief is where most names go wrong.
Wide generation. Hundreds of candidates, judgment switched off on purpose, so the good weird ideas survive.
A quick cut, then a debate. Weak and copycat names get dropped, then a few expert viewpoints (strategy, sound, trademark) argue over what's left.
Screening and scoring. Survivors get checked for trademark and cross-language problems, then scored by a deterministic program so the numbers match the reasoning.

Two things that were hard to get right

Scoring has to be mechanical or it quietly lies. My first version stated good rules and then ignored them in the actual numbers. It once ranked a name first while admitting that name failed its own sound and trademark checks. The fix: make a flagged rule actually lower the score. If you ever build something that ranks with an LLM, watch for this exact trap.

The brief matters more than the generator. A vague or self-contradictory brief produces clever-but-hollow puns. A sharp brief produces names you'd actually ship, from the same engine. It speeds up taste; it doesn't replace it.

The proof it works

To test it, I pointed it at its own brief and let it run. The first round flopped. I sharpened the brief, ran it again, and "Nomira" (from the Latin nomen, "name") won on its own scoreboard. A naming tool that can pass its own bar is one I trust to name yours.

Honest limits

The trademark step is a pre-check, not a clearance, so do a real search before you commit. And the final call is yours: it gets you to a strong shortlist fast, but taste stays human.