正體中文(Traditional Chinese),又稱繁體中文,為臺灣官方使用的標準漢字書寫形式。本技能以「正體中文」為正式名稱,並以臺灣華語慣用方式為規範。
Claude Code 的在地化翻譯技能,確保 AI 產出的中文內容符合臺灣華語母語者的用詞習慣,避免中國用語與簡體直譯。
- EN↔zh-TW 雙向翻譯,也支援 EN↔JA↔zh-TW 三語翻譯
- 用詞規範:內建臺灣 vs. 中國用語對照表(40+ 高頻詞彙速查)
- 領域污染檢查:按電腦資訊、日常生活、娛樂、政經、交通等領域,主動檢查中國用語殘留
- 兩岸詞彙資料庫:整合中華語文知識庫 4,800+ 筆兩岸差異用詞,可用 Grep 即時查詢
- 完整翻譯工作流程:理解→翻譯→校對→潤飾四步驟,附品質檢查清單
- 語言參考文件:正體中文、英文、日文各自的語言特性與翻譯策略
將此資料夾放入 Claude Code 的 skills 目錄即可:
# 方法一:直接 clone
git clone https://github.com/Qmo37/localization-tw.git ~/.claude/skills/localization-tw
# 方法二:手動下載後放入
cp -r localization-tw ~/.claude/skills/安裝後,Claude Code 在處理中文相關任務時會自動套用此技能。你可以:
- 翻譯英文文件為臺灣中文
- 產生符合臺灣用語的 UI 文字、註解、文件
- 檢查既有中文內容是否有中國用語殘留
❌ 請您通過以下鏈接獲取軟件的默認配置信息
✅ 請透過以下連結取得軟體的預設設定資訊
❌ 服務器返回的數據需要進行一個解析的操作
✅ 伺服器回傳的資料需要解析
內建抓取腳本可從中華語文知識庫取得完整兩岸差異用詞:
# 抓取全部 ~4,800 筆詞彙
python3 scripts/fetch-linguipedia.py
# 僅抓取「同實異名」類型(同一事物,兩岸用詞不同)
python3 scripts/fetch-linguipedia.py --type 同實異名
# 先看統計資料,不寫入檔案
python3 scripts/fetch-linguipedia.py --dry-run
# 儲存原始 JSON 以便離線使用
python3 scripts/fetch-linguipedia.py --save-cache cache.json
python3 scripts/fetch-linguipedia.py --from-cache cache.json抓取後的資料會存入 references/linguipedia-cross-strait.md,依子分類分節呈現。由於檔案較大,翻譯時建議用 Grep 搜尋特定詞彙,而非全文載入。
localization-tw/
├── SKILL.md # 主技能定義(工作流程、速查表、品質清單)
├── references/
│ ├── vocabulary.md # 臺灣用語對照表(名詞、動詞、句式、標點)
│ ├── linguipedia-cross-strait.md # 兩岸差異用詞(4,800+ 筆,腳本產生)
│ ├── chinese-traditional.md # 正體中文語言特性
│ ├── english.md # 英文語言特性
│ └── japanese.md # 日文語言特性
├── translation-challenges.md # 常見翻譯挑戰範例與解法
├── tools-resources.md # 辭典、語料庫、工具彙整
└── scripts/
└── fetch-linguipedia.py # 中華語文知識庫抓取腳本
技能會根據翻譯內容涉及的領域,主動提醒檢查高風險詞彙:
| 領域 | 污染程度 | 常見錯誤 → 正確用法 |
|---|---|---|
| 電腦資訊 | 🔴 嚴重 | 數據→資料、代碼→程式碼、優化→最佳化、接口→介面 |
| 日常生活/食物 | 🔴 嚴重 | 西紅柿→番茄、土豆→馬鈴薯、信息→資訊、視頻→影片 |
| 娛樂/遊戲 | 🟠 中高 | 主播→實況主、點擊→點選、下載安裝→下載並安裝 |
| 政經社會 | 🟡 中等 | 數字經濟→數位經濟、互聯網→網際網路 |
| 交通運輸 | 🟡 中等 | 出租車→計程車、地鐵→捷運、公交車→公車 |
本技能基於以下兩個開源技能合併、修改而成:
- translation-expertise by shino369 — 英日中(正體)三語翻譯方法論與最佳實踐
- taiwan-traditional-chinese-localization — 臺灣正體中文在地化規範
兩岸差異用詞資料來源:中華語文知識庫 Copyright © 中華文化總會(National Cultural Association of Taiwan, NCAT)版權所有。本技能引用其公開資料僅供學術與翻譯參考用途。
本技能程式碼以 MIT License 釋出。
兩岸詞彙資料庫內容之著作權屬中華文化總會所有,使用時請遵守其授權條款。
🌐 English Version
In Taiwan, the official writing system is called 正體中文 (Standard Chinese / Orthodox Chinese), commonly referred to internationally as Traditional Chinese (繁體中文). This skill uses "Standard Chinese" as the formal name.
A Claude Code skill for localization and translation, ensuring AI-generated Chinese content follows Taiwanese Mandarin conventions and avoids mainland Chinese (CN) terms.
- EN↔zh-TW bidirectional translation, with EN↔JA↔zh-TW trilingual support
- Terminology standards: Built-in Taiwan vs. China vocabulary reference (40+ high-frequency terms)
- Domain-aware pollution checking: Actively flags CN term contamination across IT, daily life, entertainment, politics, transportation domains
- Cross-strait vocabulary database: Integrated with Chinese Linguipedia — 4,800+ cross-strait terminology entries, searchable via Grep
- Complete translation workflow: Understand → Translate → Proofread → Polish, with a quality checklist
- Language reference files: Linguistic characteristics and translation strategies for Traditional Chinese, English, and Japanese
Place this folder in your Claude Code skills directory:
# Option 1: Clone directly
git clone https://github.com/Qmo37/localization-tw.git ~/.claude/skills/localization-tw
# Option 2: Copy manually
cp -r localization-tw ~/.claude/skills/Once installed, Claude Code automatically applies this skill when handling Chinese-related tasks. You can:
- Translate English documents into Taiwanese Chinese
- Generate UI text, comments, and documentation in Taiwan conventions
- Check existing Chinese content for mainland Chinese term contamination
❌ 請您通過以下鏈接獲取軟件的默認配置信息
(CN: uses mainland terms like 通過, 鏈接, 軟件, 默認, 信息)
✅ 請透過以下連結取得軟體的預設設定資訊
(TW: uses 透過, 連結, 軟體, 預設, 資訊)
❌ 服務器返回的數據需要進行一個解析的操作
(CN: 服務器, 數據, verbose phrasing)
✅ 伺服器回傳的資料需要解析
(TW: 伺服器, 資料, concise phrasing)
A built-in script fetches the complete cross-strait terminology database from Chinese Linguipedia:
# Fetch all ~4,800 entries
python3 scripts/fetch-linguipedia.py
# Fetch only "same thing, different name" entries
python3 scripts/fetch-linguipedia.py --type 同實異名
# Preview stats without writing
python3 scripts/fetch-linguipedia.py --dry-run
# Save/load raw JSON for offline use
python3 scripts/fetch-linguipedia.py --save-cache cache.json
python3 scripts/fetch-linguipedia.py --from-cache cache.jsonOutput is saved to references/linguipedia-cross-strait.md, organized by subcategory. Since the file is large, use Grep to search for specific terms during translation rather than loading the full file.
localization-tw/
├── SKILL.md # Main skill definition (workflow, quick ref, checklist)
├── references/
│ ├── vocabulary.md # TW vs CN vocabulary table (nouns, verbs, grammar, punctuation)
│ ├── linguipedia-cross-strait.md # Cross-strait terms (4,800+, script-generated)
│ ├── chinese-traditional.md # Traditional Chinese linguistic reference
│ ├── english.md # English linguistic reference
│ └── japanese.md # Japanese linguistic reference
├── translation-challenges.md # Common translation challenges with examples
├── tools-resources.md # Dictionaries, corpora, and tool references
└── scripts/
└── fetch-linguipedia.py # Chinese Linguipedia fetcher script
The skill actively checks for CN term contamination based on the translation domain:
| Domain | Severity | Common Mistakes → Correct TW |
|---|---|---|
| IT / Computing | 🔴 Critical | 數據→資料, 代碼→程式碼, 優化→最佳化, 接口→介面 |
| Daily Life / Food | 🔴 Critical | 西紅柿→番茄, 土豆→馬鈴薯, 信息→資訊, 視頻→影片 |
| Entertainment / Gaming | 🟠 High | 主播→實況主, 點擊→點選, 下載安裝→下載並安裝 |
| Politics / Society | 🟡 Moderate | 數字經濟→數位經濟, 互聯網→網際網路 |
| Transportation | 🟡 Moderate | 出租車→計程車, 地鐵→捷運, 公交車→公車 |
This skill was created by combining and extending two open-source skills:
- translation-expertise by shino369 — Expert EN-JA-ZH (Standard/Traditional) trilingual translation methodology and best practices
- taiwan-traditional-chinese-localization — Taiwan Standard Chinese localization standards
Cross-strait vocabulary data source: Chinese Linguipedia (中華語文知識庫) Copyright © National Cultural Association of Taiwan (中華文化總會, NCAT). Data referenced for academic and translation purposes only.
This skill's code is released under the MIT License.
The cross-strait vocabulary database content is copyrighted by the National Cultural Association of Taiwan (NCAT). Please comply with their terms of use.