Translate and localize docs with AI
Skills used in this guide (2)
把签证申请材料（图片）翻译成英文，并生成原文与译文对照的双语 PDF。
npx skills add affaan-m/ECC -g国际化与本地化模式。检测硬编码字符串、管理翻译、locale 文件、RTL 支持。
npx skills add sickn33/agentic-awesome-skills -g
Machine translation's scare is "fluent but distorted" — wrong terms, wrong tone, stiff local phrasing. The point: give a glossary and target-reader tone, have it "translate + keep proper nouns", then you verify key sentences — not trust the whole blindly.
Who this is for / not for
For:
- Translate materials, docs, product copy
- Maintain multilingual UI / i18n
- Want consistent terms, not one word many ways
Not for:
- Certified legal/contract translation (human review + pro translator)
- Confidential source (mask before cloud)
- "Zero glossary, 100% accurate" (terms are yours to set)
Before you start
- Glossary: proper nouns / product names / acronyms CN-EN.
- Reader & tone: formal / casual / marketing.
- Format: Markdown / doc, keep structure if possible.
Steps
1. Install the skill
visa-doc-translate is a document-level translation skill. Copy its folder into your client's skills directory (same as any Agent Skill):
| Client | Directory |
|---|---|
| Claude Code | ~/.claude/skills/ |
| Codex | ~/.codex/skills/ |
| Cursor | ~/.cursor/skills/ |
| Trae | ~/.trae/skills/ |
| CodeBuddy | ~/.codebuddy/skills/ |
| Tongyi Lingma | ~/.lingma/skills/ |
2. Glossary first, then translate
Translate the doc to Chinese, rules:
- glossary: API→接口(don't translate), token→令牌, Pipeline→流水线
- proper nouns (product/person) stay original
- tone: technical-doc, not casual
- keep Markdown headings and lists3. Product copy goes i18n
UI/product multilingual → i18n-localization, align by key, avoid many translations of one word:
Localize these UI strings to English, keep keys unchanged,
one concept one translation across the file, button copy ≤3 words.4. You verify key sentences
After translation, spot-check the glossary and 3 key sentences for "fluent but distorted".
Pitfalls
| Pit | Symptom | Fix |
|---|---|---|
| Wrong terms | API translated | Glossary first, keep proper nouns |
| Wrong tone | Tech reads marketing | State reader & tone |
| One word many | Same word varies | i18n by key, unified |
| Trust whole | Key sentence wrong | Spot-check glossary + 3 keys |
| Leak | Confidential to cloud | Mask sensitive first |
FAQ
Translate without glossary?
Yes but quality varies; glossary lifts accuracy clearly.
Avoid re-translating product?
i18n-localization by key; one concept translated once.
Keep format?
Yes — require Markdown / original structure; doc-level keeps format.
Contracts via AI?
Draft only; formal use pro translator + certified review.
Check quality?
Spot-check glossary + 3 key sentences for "fluent but distorted".
Related guides
Continue across these scenes:
- Write technical blogs and long-form with AI
- Turn a 100-page PDF into notes you can actually use
- Make podcast scripts and voiceover with AI
相关源码
- affaan-m/ECC · visa-doc-translate source
- sickn33/agentic-awesome-skills · i18n-localization source
- Agent Skill open format spec
一句话结论
结论：翻译先给术语表再翻、专有名词保留，最后你抽查关键句，别盲信"通顺但走样"的全文。
References
Related Scenes
Write PRDs and product specs with AI
The pain of PRD isn't "wrong format" — it's "goal and acceptance unclear, dev guesses". The point: give background, target user, problem; let AI output "feature list + acceptance + edges", not decide "why build" for you.
Write weekly reports and standups with AI
The pain of weekly reports isn't "can't write" — it's "turning a log into outcomes". The point: give AI the raw material (what you did, what's stuck), have it close into "progress / risk / next", not invent content for you.
Handle your inbox and support messages with AI
The red line for email automation is "draft yes, auto-send careful". "Triage + draft + you hit send" is safest; letting AI blast-send means you own the mistake, not it.
Turn meeting recordings into a task list with AI
The biggest waste in meetings isn't the meeting — it's nobody remembering "who, by when, what" afterward. The point of this flow is to compress scattered discussion into a task table with owners and due dates, not to write another minutes doc nobody reads.