fluent-korean
A Claude Code output-style that makes the model write clear, natural Korean instead of broken machine Korean
Low risk
We rate an entry low when it mostly gives the agent instructions and reference material.
Why this level
- It is only an output-style with instructions; it changes output style and runs nothing
Install
In your terminal, with SkillFoxx CLI
npx skillfoxx add plugins/fluent-koreanDetects the agents on your machine, checks the risk and pins the version.
Other ways to install
Run one by one in the Claude Code chat
/plugin marketplace add snflkd/fluent-korean
/plugin install fluent-korean@fluent-koreanInstall the plugin: /plugin marketplace add snflkd/fluent-korean, then /plugin install fluent-korean@fluent-korean. After install, pick one of the two output-styles in /config and start a new session.
Other ways from the author
/plugin marketplace add snflkd/fluent-korean
/plugin install fluent-korean@fluent-koreanAfter install, select the output-style in /config and start a new session.
This is third-party code. Review the repository files before installing.
What it does
The plugin defines an output-style that is added to the system prompt and steers the model to write clear Korean up front. It restores dropped particles and endings, removes telegraphic noun strings and rare vocabulary, and prioritizes clarity of meaning over elegant prose. It ships two variants: one keeps the coding instructions, the other drops them for cases where the model does not change code itself. The instruction text can also be pasted into other environments: the Claude web and desktop apps or other models. The README includes optional blocks to tune behavior, such as politeness level or a ban on rare words.
Who it is for. For Korean-speaking users of Claude Code and coding agents who need readable Korean output.
Good fit when
- The model answers in Korean in a broken way with lost meaning
- You want Korean text in work output to read as natural and unambiguous
- Agents pass Korean prompts around and language quality degrades step by step
Not a fit when
- You need literary style rather than clarity of meaning
- You need translationese cleanup, AI-tell removal or spelling fixes: the author points to other skills for that
Example request
Answer in Korean clearly and naturally without losing meaningLimitations
The tool aims at clarity of meaning, not elegant prose, and does not replace translationese cleanup or AI-tell removal. It slightly increases token usage because it restores dropped sentence parts. The effect weakens in long sessions and with many competing instructions. The author advises checking subagent behavior separately. MIT licensed.
How to disable. Switch the output-style back in the /config menu and start a new session or run /clear. If installed as a plugin, remove it via /plugin.
Security check
- It is only an output-style with instructions; it changes output style and runs nothing
README in short
The README explains that current models often write Korean poorly, especially coding agents tuned for brevity. The plugin provides an output-style that steers the model to write clear Korean up front and offers two variants: with coding instructions and without. Installation is via the Claude Code marketplace or by placing the file into the output-styles directory, after which the variant is selected in settings. The README includes tuning blocks: politeness, a ban on rare words, and applying the rules to all Korean output. The Korean part of the README is written by a human with a relevant background, under the MIT license.
FAQ
How do the two variants differ?
The fluent-korean variant keeps coding instructions and suits coding work. The fluent-korean-not-coding variant drops them, for cases where the model does not change code itself.
Can I use it outside Claude Code?
Yes. The instruction text from the output-style file can be pasted into the Claude web or desktop app, and into other models as a plain style instruction.
Related
A self-hosted knowledge base with block-level references and a built-in MCP server for connecting AI agents to your notes
A CLI for every Google Workspace API with JSON output and agent skills: Drive, Gmail, Calendar, Sheets and more
Local search over Markdown notes, docs and meeting transcripts: keywords, semantic search and reranking, with an MCP server
A task manager for AI-driven development: breaks a PRD into dependent tasks and guides the agent through them via MCP or CLI