korean humanizer

Korean Humanizer helps revise stiff AI-generated Korean into clear, natural writing. It addresses translationese, formulaic openings, redundant phrasing, and expressions that sound awkward in context.

Its editing rules prioritize the original meaning, facts, degree of emphasis, and level of politeness. Unnecessary wording can be shortened without adding claims or changing the writer’s intent.

What it does

• Reduces formulaic wording by reviewing stock phrases, empty intensifiers, and repetitive expressions in context.
• Improves sentence flow by revising awkward passive constructions, excessive nominalizations, and literal translations.
• Prioritizes preservation of numbers, dates, proper nouns, links, quotations, conditions, and the original sentence-ending register.
• Supports writing preferences through brand voice profiles and inline lists of words to preserve, avoid, or prefer. These preferences remain subject to the preservation rules.

How it works

1. Review the input to identify its purpose, dominant sentence-ending register, and information that must be preserved. Ask for clarification only when ambiguity could affect the revision.
2. Consult the reference catalog of 100+ editing patterns across 12 categories, choosing changes that fit the context.
3. Rewrite the text into natural Korean. Remove redundant wording without enforcing a fixed character count or edit percentage.
4. Check and explain the revision against the original. Return the full revised text under Humanized and up to five change explanations under 주요 변경 (최대 5개).

Frameworks & tools

Korean Humanizer is a Markdown-based skill delivered through Agensi MCP. Connect a compatible agent to https://mcp.agensi.io/mcp, then ask it to load korean-humanizer and revise your text.

The agent retrieves the skill instructions and loads supporting references and examples through MCP as needed. No ZIP download or local skill-folder installation is required for this delivery method. Revisions and change explanations are written in Korean.

For environments where Agensi MCP is unavailable, PROMPT.md and PROMPT.short.md are available separately in the GitHub repository. These system-prompt versions are not included in the Agensi MCP package.

Why use this instead of a one-line prompt

A request such as “make this sound natural” leaves many editing decisions unspecified. This skill supplies explicit preservation rules, contextual examples, preference priorities, and a consistent output format.

It is designed to make those decisions clearer and easier to review. Results still depend on the model and the input.

Use cases

• Revising AI-generated LinkedIn posts, blog drafts, and newsletters.
• Making customer-facing replies clearer and more natural while preserving their intent, commitments, and politeness.
• Improving stiff translated Korean in marketing copy and YouTube scripts.

Known limitations

This skill is intended for Korean text. It is not intended for English editing, simple typo correction, internal technical notes, or legal and regulatory documents.

It does not verify factual accuracy or determine whether text was written by AI. Perfect preservation and identical results across models are not guaranteed; review important facts and commitments before publishing.

Available free under the MIT license. Visit the GitHub repository for the source, examples, and updates. If you find it useful, consider giving the project a star: https://github.com/dotoricode/korean-humanizer