A skill that fights AI homogenization.
Stop your AI assistant from handing you the same generic name, the same
dark-gradient design, and the same concept it hands everyone else. distinct
forces it to check what already exists and steer toward something original.
Ask an AI to "name my finance app" or "design a clean landing page," and you'll often get the same answers everyone else gets — because the model converges on its statistical defaults. The symptoms are everywhere:
- Names:
calm+X, X+OS, X+AI,smart+X, two stacked dictionary words. - Design: dark background, neon/gradient glow, sage-or-indigo accent, the same hero-with-three-cards layout, a leaf or blob logo.
- Concepts: the obvious take on the obvious idea.
The result: people independently launch products that look and sound nearly identical, sometimes days apart. This isn't a leak — it's homogenization, a known weakness of AI output. And it's a real cost: a generic name collides with existing products and is instantly forgettable.
It's an Agent Skill. Once installed, whenever you're naming, designing, or conceiving something, it forces your assistant to:
- Check what already exists first — search the name, its domain variants, the category, and the top competitors. Report it before suggesting anything.
- Reject generic names — flag the overused patterns and generate distinctive candidates from unexpected angles instead (5+, each with its rationale).
- Break out of default design — offer 2–3 genuinely different visual directions tied to the product's character, not one default with swapped colors.
- State the difference — one line on what makes the result unlike the field.
This is an Agent Skill (SKILL.md format), compatible with Claude and other
assistants that support skills.
- Claude (skills-enabled): add the
distinctfolder /.skillfile to your skills. - Or just drop
SKILL.mdinto your project's skills directory.
Once installed, it triggers automatically when you start naming or designing something — no need to invoke it by hand.
distinct reduces lookalike results. It does not guarantee uniqueness — the
same real-world need can lead different people to similar ideas independently, and no
tool controls that. What it controls is the controllable part: not defaulting to
the obvious, checking the landscape, and choosing distinctive names and designs on
purpose.
MIT — use it, change it, ship it.