K Compliance Agent — Development skill for Claude Code

K-Compliance Agent

Catches U.S. FDA compliance risks in Korean beauty product labels before a brand ships. Works in Korean or English.

Why this exists

The U.S. is now the largest market for Korean beauty, passing China for the first time. Korean cosmetics exports to the U.S. reached about $2.2B last year, up 15 percent, and much of that growth is small indie brands selling direct to consumers online. Those brands rarely have a regulatory team.

One label detail can stop a shipment:

• A claim that makes the FDA treat a cosmetic like a drug, such as "treats acne"
• A sunscreen ingredient that is routine in Korea and not approved in the U.S.
• Wording that is fine in Korean and becomes a drug claim in English

What it does

Paste in a product label or ingredient list. About twelve seconds later you get back:

• The U.S. compliance risks in that label, flagged
• The actual FDA rule behind each finding, cited
• A specific fix for each one

It works in Korean or English, so the same report is usable by a U.S. regulatory reviewer and by a Korean product team.

How it is built

Layer	Stack
Orchestration	n8n
Models	Claude and OpenAI
Front end	Next.js on Vercel

The models were the easy part. Getting the FDA rules right was the work.

Status

Portfolio project. Validated against three real products. No users, not commercially launched.

Repo contents

Being added. The n8n workflow export and the Next.js front end land here next.