baoyu-translate Skill for Claude: Quick, Normal or Refined Translation
Has Claude translate an article, file or URL in quick, normal or refined mode, with your glossary, style and audience.
- Installs
- 25.8K
- on skills.sh
- Stars
- 27.2K
- on GitHub
- Files
- 10
- besides SKILL.md
- Updated
- Nov 2026
- last repo commit
baoyu-translate has Claude translate a file, a URL or pasted text. Quick mode translates straight away. Normal mode analyzes the text first, then translates. Refined mode adds a critical review, a revision and a polish for publication quality. Normal is the default, and Chinese (zh-CN) is the default target.
It saves every step into a folder next to your source, and the final text is always translation.md. Long articles of 4,000 words or more are split into chunks by a bundled script and translated in parallel by sub-agents. On first use it asks for your preferences and saves them to an EXTEND.md file before it translates anything.
Good for
- Writers and editors translating articles into natural Chinese, English or another language
- Teams that need one glossary applied across a long document
- Anyone who wants a fast draft now and a reviewed, polished version later
Install baoyu-translate
Claude Code
- 1
Copy this prompt.
Install the baoyu-translate skill in Claude Code. Run this command: npx skills add https://github.com/jimliu/baoyu-skills --skill baoyu-translate Then check that it loaded and tell me how to use it.
- 2
Paste it into Claude Code. It shows the change and asks before running it.
Requires Node.js.
From its listing on skills.sh. Source.
What's in the skill
When it loads
This skill should be used when the user asks to "translate", "翻译", "精翻", "translate article", "translate to Chinese", "translate to English", "改成中文", "改成英文", "convert to Chinese", "localize", "本地化", "refined translation", "精细翻译", "proofread translation", "快速翻译", "快翻", "这篇文章翻译一下", or provides a URL/file with translation intent. Supports three modes (quick/normal/refined) with custom glossary support.
Sections in SKILL.md
- User Input Tools
- Script Directory
- Preferences (EXTEND.md)
- First-Time Setup (BLOCKING)
- Defaults
- Modes
- Workflow
- Step 1: Load Preferences
- Step 2: Materialize Source & Create Output Directory
- Step 3: Assess Content Length
- Step 4: Translate & Refine
- Step 5: Output
- Extension Support
Key facts
- License
- MIT
- What installs
- 11 files including SKILL.md: 6 reference files and 4 script files (main.ts, chunk.ts, package.json, bun.lock)
- Installs on skills.sh
- 25,793 (Nov 2, 2026)
- GitHub stars
- 27,159 (Nov 2, 2026)
- Last commit
- 2026-11-01
- skills.sh audits
- Gen Agent Trust Hub pass, Socket pass, Snyk warn
What it includes
Translates a file, URL or inline text, as shown in the README
Options are --mode, --from, --to, --audience, --style and --glossary. Phrases like "translate to Chinese" or "精翻" also trigger the skill.
Try: “/translate article.md --mode refined --to zh-CN”
Write ·
/translate <source>Asks for language, mode, audience, style and save location, then saves EXTEND.md
This blocks translation until it is done. EXTEND.md can live in the project, in ~/.config/baoyu-skills/ or in ~/.baoyu-skills/.
Write ·
First-time setupTranslates in one pass with no analysis and no chunking
For short or informal text. For long text it warns that normal mode keeps terms more consistent.
Quick modeAnalyzes the text, assembles a prompt, then translates
After it finishes it offers to continue with review and polish if you reply "继续润色" or "refine".
Write ·
Normal modeAnalyze, draft, critique, revise and polish in six saved steps
Each step reads the file the last step wrote, from 01-analysis.md through 05-revision.md to translation.md.
Write ·
Refined mode9 style presets and 4 audience presets, plus custom text
Styles run from storytelling and formal to literal and humorous. Audiences are general, technical, academic and business.
Styles and audiencesMerges inline, file, built-in EN-ZH and --glossary terms
A built-in English to Chinese glossary loads when that pair applies. Your --glossary file overrides the others.
Read ·
GlossariesSplit long markdown at block boundaries, run with bun or npx -y bun
The splitter writes chunk files and keeps front matter in its own file. package.json lists markdown-it 14.1.1.
Runs code ·
scripts/main.ts and chunk.tsLists images that may still carry text in the source language
It only lists them and does not edit images unless you ask.
Image language reminder
From the official docs
Permissions
Low risk
It runs a bundled script through npx and bun, fetches the URL you give it, and writes translation files and an EXTEND.md settings file to disk. Check the output folder before you commit it.
What it can do
- Read your data
- Create and change things
- Delete things
- Read-only mode
Watch out for
It runs npx -y bun on your machine
SKILL.md resolves the runtime to bun, or to npx -y bun when bun is missing, and package.json lists markdown-it. Install bun yourself if you prefer not to run npx -y. github.com
It renames an existing output folder
If the output folder exists, it is renamed to a .backup-<timestamp> folder and not overwritten. github.com
Technical details
| Talks to | The URL you ask it to translate |
|---|---|
| Sign-in and keys | No sign-in or key is described in SKILL.md. |
| Scopes |
|
Questions
What does baoyu-translate do?
It translates articles, files, URLs and pasted text with Claude in one of three modes. It can follow a glossary, a style and an audience, and it saves its working files beside the result.
How do I install baoyu-translate?
Use the install box above. The README installs the whole 20+ skill repo with npx skills add jimliu/baoyu-skills but advises installing only what you need, because every installed skill adds context to each run.
Which mode should I pick?
Quick for short or informal text, normal for articles and blog posts, and refined for work you will publish. SKILL.md says normal gives better term consistency than quick on long text.
Why does it ask questions the first time?
SKILL.md makes first-time setup mandatory when no EXTEND.md exists. It asks for target language, mode, audience, style and where to save the settings, and then translates.
How does it handle long articles?
At 4,000 words or more in normal or refined mode it builds a term list, splits the text at markdown block boundaries into chunks of up to 5,000 words, and translates them in parallel with sub-agents if your client has them.
Does it need an API key?
No key is described in SKILL.md. It needs bun or npx so it can run the chunk script.
Is it safe?
It runs one bundled script through bun or npx -y bun, fetches URLs you give it, and writes files next to your source. It does not overwrite an earlier output folder, and it renames the old one with a timestamp.
Which languages does it support?
The default target is zh-CN, and setup lists zh-TW, English and Japanese. Other codes can be typed in. The built-in glossary covers English to Chinese only.
Information
updated Nov 2, 2026
- Repository
- jimliu/baoyu-skills
- Folder
skills/baoyu-translate
Sources
Docs, issue trackers and discussions this page is built from.
Show all 6Hide
- baoyu-translate SKILL.md · github.com
- baoyu-translate on skills.sh · skills.sh
- github.com · github.com
- references/workflow-mechanics.md · github.com
- references/config/first-time-setup.md · github.com
- github.com · github.com