translation-pro
Handle professional translation — machine translation workflows, localization, quality assurance, and multilingual comms.
- License 许可
- MIT
- 最近更新
- 2026-11-07
- Stars
- ★ 0
- Forks
- 1
01 前言 Front Matter
- name
- translation-pro
- category
- enterprise-communication
- description
- Handle professional translation — machine translation workflows, localization, quality assurance, and multilingual comms.
03 SKILL.md 正文
Overview
Professional translation spans: quick machine translation for internal use, polished localization for customer-facing content, and real-time interpretation for meetings. This skill covers choosing the right approach per content type, machine-translation post-editing workflows, localization beyond words, and quality assurance.
Professional translation and localization adapt content for global audiences: beyond word conversion to cultural resonance. Mastery covers translation workflows, machine translation post-editing, localization engineering, and quality assurance across languages.
When to use
-
Translating product UI or documentation
-
Localizing marketing content
-
Setting up multilingual support
-
Running multilingual meetings
-
Building MT post-editing workflows
-
Ensuring translation quality
-
Localizing products for new markets
-
Translating marketing campaigns
-
Managing multilingual content operations
-
Implementing machine translation workflows
-
Translating user-generated content
Core concepts
Appropriateness tiers. Raw MT (internal understanding, gist), MT + light post-edit (internal docs, support macros), MT + full post-edit (help centers, UI), human translation (marketing, legal, brand content), transcreation (slogans, campaigns — rewritten for culture, not translated). Match investment to visibility and risk.
Localization vs. translation. Translation converts words; localization adapts: date/number/currency formats, cultural references, imagery, colors, legal requirements, and UX (text expansion — German runs ~30% longer than English; design for it). True localization feels native.
MT post-editing. Workflow: MT draft → human editor fixes errors and awkwardness → QA check. Provide editors with: glossaries (approved terms), style guides per language, and context (screenshots for UI strings — translators can't guess from isolated strings).
Glossaries and TMs. Term bases (approved translations for key terms — brand names, product terms) and translation memories (reuse of past translations) ensure consistency and cut costs. Maintain them as living assets; inconsistent terminology confuses users and looks unprofessional.
Quality assurance. Back-translation (for critical content), in-country review (native speakers checking naturalness), linguistic QA (spelling, grammar, consistency), and functional QA (UI fits, links work, no encoding issues). Sample-based QA for volume; full review for high-stakes content.
Real-time. Live captions/translation in meetings, interpretation for important calls (consecutive vs. simultaneous), and multilingual chat support (MT-assisted agents with human oversight). Set expectations: real-time MT is gist-level, not publication-level.
Translation vs. localization. Translation converts words; localization adapts meaning, imagery, formats, and cultural references. Marketing content needs transcreation (creative rewriting); UI needs functional localization; legal needs certified precision. Match approach to content type — one method does not fit all. Machine translation + post-editing (MTPE). Raw MT for gisting → light post-editing (accuracy) → full post-editing (publishable quality). MTPE cuts costs 30–50% for suitable content (technical, repetitive). Creative and brand content still needs human translation — MTPE produces adequate, not excellent. CAT tools and TMs. Translation memory (reuse past translations), terminology bases (consistent terms), and QA checks. TMs compound value — leverage grows with every project. Invest in terminology management; inconsistent terms confuse globally.
Practical workflow
- Classify content. Tier each content type (UI, docs, marketing, legal, support) to its translation approach. This drives budget and workflow.
- Build assets. Glossaries per language, style guides, translation memories, and context packages (screenshots, descriptions). Upfront investment pays back in consistency forever.
- Set up workflows. Source content → tier-appropriate translation → review → QA → publish. Integrate with content pipelines (docs, product releases) so translation isn't a last-minute scramble.
- Post-edit effectively. Train editors on: when to fix vs. rewrite, terminology adherence, and flagging MT systematic errors (feedback improves engines). Track edit distance — high edits mean the engine or source needs work.
- QA. Linguistic checks, in-country review for launches, functional testing of localized UI, and back-translation for legal/medical critical content.
- Maintain. Update TMs/glossaries with every project, review quality metrics (edit rates, in-country feedback, support tickets about confusion), and refresh stale translations when source content changes.
Source writing for translation: short sentences, no idioms/slang, consistent terminology, one idea per sentence, and context notes for ambiguous terms. Good source text halves translation effort.
Localization workflow: content freeze → extraction → translation/MTPE → linguistic QA → in-context review → functional testing → publish. Build localization into development (i18n) from the start — retrofitting is 10x harder. Quality framework: accuracy, fluency, terminology, style, and locale conventions. Sample-based QA (10–20%) for routine content; 100% for critical. Track quality scores per vendor; manage the supply chain actively.
Common pitfalls
- Raw MT for customer content. Publishing unedited machine output. Embarrassing errors erode trust — post-edit customer-facing text.
- No context for translators. Isolated UI strings translated blind. Provide screenshots and descriptions.
- Ignoring text expansion. German/French overflowing buttons designed for English. Design with 30–40% expansion room.
- Inconsistent terminology. Five translations for the same product term. Maintain and enforce glossaries.
- Translating idioms literally. "Break a leg" becomes medical advice. Flag idioms for transcreation.
- One-and-done. Translating once, never updating when source changes. Stale translations mislead.
- Skipping in-country review. Translations that are technically correct but culturally tone-deaf. Native review for launches.
- Translating without context. Strings without screenshots or descriptions. Context-free translation produces errors — provide visual context always.
- Ignoring text expansion. German runs 30% longer than English; UI breaks. Design flexible layouts; test in longest languages.
- One translator per language. No redundancy or review. Every language needs primary + reviewer minimum.
- Machine-only for brand content. Raw MT on marketing copy. Brand voice needs humans — MTPE minimum for customer-facing content.
- Ignoring right-to-left. Layouts breaking in Arabic/Hebrew. Design RTL from the start — retrofitting is expensive.
04 同仓库更多 Skills
web-scrape
Turn web pages into LLM-ready markdown via the Firecrawl API (renders JS, anti-bot handling). Use when browser_open/fetch fails on JS-heavy pages, or when you need clean markdown/structured extraction from a site. Requires FIRECRAWL_API_KEY.
design-taste-frontend
tasteskill v2, the anti-slop frontend skill. Art direction for websites, distilled from a study of 247 curated gallery sites (One Page Love, Awwwards Sites of the Day). Deals researched directions (idea, composition, type, palette, photographs, section blueprints) from a library, enforces a strict no-slop house style with a real-browser inspector, and runs a look-and-fix loop on screenshots. Use this whenever you build, design, redesign, restyle, polish or review a landing page, marketing site, portfolio, product page, one-pager, hero or any public-facing web page, in any stack (HTML/CSS, React, Next.js, Vue, Svelte, Astro, Tailwind) — also when the user says a page looks generic, templated, "AI-made", sloppy, or asks to make it look designed, premium, award-level or less like every other site. Not for dashboards or app screens.
test-driven-development
Use when implementing any feature or bugfix, before writing implementation code
systematic-debugging
Use when encountering any bug, test failure, or unexpected behavior, before proposing fixes
requesting-code-review
Use when completing tasks, implementing major features, or before merging to verify work meets requirements
brainstorming
You MUST use this before any creative work - creating features, building components, adding functionality, modifying behavior, or planning anything new, in software or out of it (a talk, a business, a renovation).