ko-prompt-kit: Production-ready Korean LLM prompts for Claude & GPT

#korean #ai #typescript #promptengineering

If you're building AI apps that need to output natural Korean, translating English prompts doesn't cut it. Korean has formal/informal speech levels (존댓말/반말), unique document conventions, and cultural context that English-native prompts completely ignore.

So I built ko-prompt-kit — 14 production-ready Korean prompt templates across 5 categories, with a TypeScript API and CLI.

Zero install, instant use

npx ko-prompt list
npx ko-prompt get coding/code-review

14 prompts across 5 categories

Category	Prompts
Business	Email reply, meeting minutes, report summary
Coding	Code review, commit message, bug analysis, JSDoc
Customer Service	Complaint reply, FAQ answer
Writing	Blog post, marketing copy
Analysis	Document summary, sentiment, competitive analysis

TypeScript API

import { getById, buildPrompt, search } from 'ko-prompt-kit';

const prompt = getById('coding/code-review');
const built = buildPrompt(prompt, {
  language: 'typescript',
  code: yourCode,
  focus: '보안',
});

// Use with Claude
const response = await anthropic.messages.create({
  model: 'claude-opus-4-7',
  system: built.system,
  messages: [{ role: 'user', content: built.user }],
});

What makes Korean prompts different

Each prompt is designed around Korean language specifics:

Speech level: formal (합쇼체) vs informal (해체) — selected per use case
Document structure: Korean business docs have specific conventions
Cultural context: complaint handling, business email norms

Search and filter

// Find formal prompts for business use
const formal = search({ category: 'business', speechLevel: 'formal' });

// Search by keyword (Korean or English)
const emailPrompts = search({ query: '이메일' });