Korean-language Claude prompt engineering guide covers 20 patterns
Sangmin Lee's guide on Dev.to outlines 20 Korean-language prompt engineering patterns for Claude, centered on three core principles: assign a clear role, specify the desired output format, and provide one to three examples.
Score breakdown
The guide consolidates 20 concrete, copy-paste-ready Korean-language prompt patterns for Claude, giving Korean-speaking developers a structured reference that addresses Korean-specific formatting conventions alongside universal prompt engineering techniques.
- 01Three core rules: assign a clear role, specify output format, and provide 1–3 examples.
- 02English prompt engineering techniques transfer directly to Korean prompts.
- 03Korean-specific tip: use line breaks and bullet points instead of long run-on sentences for multi-constraint instructions.
Sangmin Lee's guide, originally published at claudeguide.io/korean-prompt-engineering, presents 20 verified Korean-language prompt patterns for Claude. The three core principles are: (1) assign a clear role, (2) specify the desired output format, and (3) provide one to three examples. The guide argues that standard English prompt engineering techniques work equally well in Korean, but that Korean-specific formatting habits — particularly breaking multi-constraint instructions into bullet-pointed lines rather than long compound sentences — help Claude parse intent more accurately.
Role assignment is demonstrated with examples such as "You are a senior backend developer who worked at Naver and Kakao" to elicit domain-specific terminology and perspective.
Several pattern categories are covered in depth. Role assignment is demonstrated with examples such as "You are a senior backend developer who worked at Naver and Kakao" to elicit domain-specific terminology and perspective. Format specification covers JSON output, markdown tables, and numbered lists with explicit character-count constraints. The few-shot examples section shows commit message generation with three input/output pairs, noting that three examples produce significantly more consistent results than one. Chain-of-thought prompting is illustrated with a database selection problem (PostgreSQL vs. MongoDB vs. MySQL) broken into explicit reasoning steps. Additional patterns include negative constraints to exclude unwanted elements, tone variants (formal corporate announcement vs. casual startup chat vs. CS apology vs. technical documentation), speech-level control (formal `-습니다` vs. informal `-다` endings), length constraints for formats like Twitter threads and elevator pitches, scored comparison tables, and system-prompt templates with placeholder variables for repeatable customer-support workflows.
Key facts
- 01Three core rules: assign a clear role, specify output format, and provide 1–3 examples.
- 02English prompt engineering techniques transfer directly to Korean prompts.
- 03Korean-specific tip: use line breaks and bullet points instead of long run-on sentences for multi-constraint instructions.
- 04Role assignment example: 'You are a senior backend developer who worked at Naver and Kakao' to get domain-specific responses.
- 05Three few-shot examples produce significantly more consistent results than one, per the guide.
- 06Patterns include JSON structured output, chain-of-thought reasoning steps, negative constraints, tone variants, and reusable template variables.
- 07The guide also promotes a paid product, 'Power Prompts 300', a collection of 300 prompt templates priced at $29.
Topics
Summary and scoring are generated automatically from the original article. We always link back to the publisher and never republish images or paywalled content. Last processed Jun 19, 2026 · 10:25 UTC. How this works →