Polish Your English-to-Chinese Translations with This Agent Skill

Where this fits

You're a developer shipping an English README to a Chinese audience. You ask your agent to translate it, but the output reads like a literal word-for-word copy — stiff, awkward, full of "translationese." You need idiomatic Chinese that flows naturally. This skill, en-zh-translation-polish, turns your agent into a translator who follows the principles of Ye Zinan: prioritize the reader, adjust domestication based on text type, and leverage Chinese parataxis over English hypotaxis. It's for anyone who wants English-to-Chinese output that sounds like a human wrote it, not a machine.

Why agents benefit

• Produces translationese-free Chinese — applies the "形合→意合" shift so the output reads like native writing, not a structural copy of English.
• Delivers a bilingual Markdown file by default — each English paragraph is a blockquote with the Chinese translation directly below, perfect for review or side-by-side comparison.
• Derives a pure-Chinese version scriptably — no re-typing; the agent runs a small Python script to strip blockquote lines, keeping both versions consistent.
• Adjusts domestication per text type — the skill's instructions guide the agent to scale between literal and free translation based on whether it's technical docs, marketing copy, or literature.
• Handles polish of existing drafts — if you already have a rough translation, the agent can refine it to remove translationese without starting from scratch.

Practical scenarios

Localizing Product Documentation

Your agent converts API docs or user guides from English to Chinese. Instead of a clunky first pass, it produces a polished Chinese version that respects technical terms while keeping sentences fluid — and you get the bilingual file for QA.

Polishing User-Submitted Translations

A contributor sends a draft translation that's overly literal. The agent applies the skill's polishing rules to recast sentences, fix word order, and improve readability, then outputs the cleaned version.

Creating Bilingual Reference Materials

For legal or educational content, you need an English-Chinese side-by-side. The skill's default output format gives you exactly that, with each paragraph paired, making it easy to publish or share.

Add it to your agent workflow

Install the skill with:

npx skills add HoraceLuBFA/en-zh-translation-polish --skill en-zh-translation-polish

Then prompt your agent: "Translate this English text to Chinese, following the skill's instructions." The agent will read the skill, apply the translation principles, and output the bilingual Markdown file. For a pure-Chinese version, ask it to derive one using the included script.

Compare before adopting

Before committing, compare this skill against alternatives. Check the quality of its output on your own sample texts — does it truly avoid translationese? Look at how recently the GitHub repo was updated and whether issues are addressed. Consider other translation skills or generic LLM prompts; this one adds discipline and a consistent workflow. Finally, ensure it fits your agent's existing tooling — it requires Read, Write, Edit, and Bash, which most agents have.

Why it is worth tracking

With 106 stars and a clear, opinionated approach to translation, this skill signals a niche but growing demand for quality localization in agent workflows. Its MIT license and active author make it safe to adopt. Evaluate it when you need English-to-Chinese output that goes beyond raw LLM translation — and track its updates if you rely on it for production content. For a curated view of this and similar skills, see OpenAgentSkill.