L’interface est localisée ; les présentations et documents des Skills sont traduits automatiquement par IA dans votre langue. Le code, les commandes, chemins et termes techniques restent en version originale, et le document de l’auteur fait foi.
Explorer les Skills
Documents · SKILL.md

lov-translation-review

lov-translation-review

Review Chinese-to-English translations for accuracy, grammar, terminology, and consistency. Produces a structured review report with prioritized issues. Trigger when: user provides a Chinese document and its English translation for review/checking/proofreading, or mentions "翻译检查", "翻译审校", "translation review", "translation check", "proofread translation".

Independent GitHub snapshot · Automated classification · Original documentation governs

Documentation de la Skill

GitHub

name: lov-translation-review category: Content Processing tagline: "Chinese-to-English translation review. Compares source & translation across 6 dimensions." description: > Review Chinese-to-English translations for accuracy, grammar, terminology, and consistency. Produces a structured review report with prioritized issues. Trigger when: user provides a Chinese document and its English translation for review/checking/proofreading, or mentions "翻译检查", "翻译审校", "translation review", "translation check", "proofread translation". license: MIT compatibility: > No external dependencies. Supports .docx (via pandoc), .md, and .txt input files. Requires pandoc for docx conversion (brew install pandoc). depends_on:

  • lov-branding-consistency metadata: author: contributors version: "1.1.1" tags: translation, review, chinese, english, proofreading

译文校对 · Translation Review

Systematically compare a Chinese source document against its English translation, identify issues across 6 dimensions, and produce a prioritized review report.

When to Use

  • User provides a Chinese original + English translation pair for review
  • User asks to check/proofread an English translation from Chinese
  • User mentions "翻译检查", "审校", "translation review"

Workflow (MANDATORY)

You MUST follow these steps in order:

Step 1: Collect Input

Use AskUserQuestion to confirm before starting:

Ask the user:

  1. Document type — academic paper (SCI/EI), technical docs, business docs, general
  2. Domain — medicine, CS, law, finance, general (affects terminology checking)
  3. Focus areas — terminology accuracy, grammar, style, references, all (multi-select)

If the user has already provided context (e.g., "SCI 4区医学论文"), skip the obvious questions.

Step 2: Read Both Documents

Read the Chinese source and English translation. For .docx files:

pandoc "<chinese_file>.docx" -t plain --wrap=none
pandoc "<english_file>.docx" -t plain --wrap=none

For .md or .txt files, read directly.

If files are large, read in sections (offset/limit) to stay within context.

Step 3: Systematic Review

Compare the two documents across 6 dimensions, in this order:

D1: Mistranslation (meaning changed)

  • Words/phrases translated to a different meaning
  • Omissions (content in Chinese missing from English)
  • Additions (content in English not in Chinese)
  • Pay special attention to domain-specific terms that have been incorrectly generalized or specialized

D2: Terminology Accuracy

  • Domain-specific terms: check against standard English terminology
  • Abbreviation consistency: first occurrence should have full form + abbreviation
  • Proper nouns: names, institutions, gene/protein names, drug names
  • Units and measurements

D3: Grammar & Syntax

  • Subject-verb agreement
  • Tense consistency (especially in Methods vs Results)
  • Article usage (a/an/the) — common error in Chinese-to-English
  • Dangling modifiers, run-on sentences
  • Passive voice overuse

D4: Consistency

  • Same Chinese term translated differently in different places
  • Formatting inconsistencies (capitalization, hyphenation, number style)
  • Heading style consistency

D5: References & Citations (academic papers only)

  • Citation numbers match between Chinese and English versions
  • Reference format consistency (author names, journal abbreviations)
  • Missing author information in reference entries
  • In-text citation correctness

D6: Style & Register

  • Academic register appropriate for target journal
  • Overly verbose or literary phrasing
  • Chinese-influenced sentence structure (e.g., "through...method" patterns)
  • Paragraph length and structure

Step 4: Source Document Issues

While reviewing, also note issues in the Chinese source that the English faithfully translates but are themselves incorrect. Flag these separately.

Step 5: Generate Report

Output a structured review report with the following format:

# 翻译审校报告 / Translation Review Report

**Document**: [title]
**Reviewer**: 品牌方 (AI-assisted)
**Date**: [YYYY-MM-DD]
**Domain**: [domain]
**Document type**: [type]

---

## Summary

[2-3 sentence overall assessment]

---

## A-Level Issues (Must Fix)

### A1. [Issue title]
**Location**: [section/paragraph]
**Original (CN)**: "..."
**Translation (EN)**: "..."
**Problem**: [explanation]
**Suggested fix**: → "..."

---

## B-Level Issues (Recommended)

### B1. [Issue title]
...

---

## Source Document Issues

### C1. [Issue title]
...

---

## Priority Summary

| Priority | # | Issue | Impact |
|----------|---|-------|--------|
| Must fix | A1 | ... | ... |
| ... | | | |

---

## Conclusion

[Overall quality assessment + recommended next steps]

Step 6: Save Report

Save the report to the same directory as the input files:

Filename: 品牌方-translation-review-{YYYY-MM-DD}-v0.1.md

If a file with this name exists, increment the version.

Issue Severity Classification

LevelCriteriaExamples
A (Must fix)Meaning changed, factual error, or would cause rejectionMistranslation, wrong terminology, reference number mismatch
B (Recommended)Awkward but understandable, or non-standard usageVerbose phrasing, minor grammar, style inconsistency
C (Source issue)Error in Chinese original, faithfully translatedWrong terminology in Chinese, factual inaccuracy

Domain-Specific Checklists

Academic / SCI Papers

  • Abstract: structured, matches body content
  • Keywords: standard MeSH or domain terms
  • Reference numbers: sequential, match between CN/EN versions
  • Figure/table captions: translated and consistent
  • Statistical reporting: p-values, CI format

Medical / Biomedical

  • Disease names: ICD/MeSH standard
  • Gene/protein names: HUGO/UniProt standard (italics for genes)
  • Drug names: INN (International Nonproprietary Names)
  • Anatomical terms: Terminologia Anatomica standard

Technical Documentation

  • UI strings: consistent with product localization
  • Code snippets: untranslated, syntax preserved
  • API terms: match official documentation

Runtime context (shared)

运行前读取本 Skill 包的 skill.yaml,由宿主提供 skill-runtime/v1 上下文。字段解析顺序为:当前请求、项目上下文、个人 Preferences、品牌 Profile、通用默认值。

  • 只使用 Manifest 声明的字段;Profile 保存公开品牌事实,Preferences 保存个人工作偏好。
  • required: true 字段缺失时,按 Manifest 的问题配置向用户提出一个聚焦问题;用户明确同意后再保存回答。
  • 报错提供可复制的 context_id、字段路径与来源,诊断内容避开秘密、完整私人路径和原始配置。

通用反馈闭环

用户在 Skill 驱动任务中提出修改意见时,继续当前产物前必须执行:

  1. 先判断意见是 task-specific(仅本次)还是 reusable(可跨任务复用)。
  2. task-specific 只修改当前任务,不改 Skill。
  3. reusable 先确定作用域:领域规则先更新对应 canonical Skill;适用于所有 Skill 的规则先更新共享规范。
  4. 完成规则更新、版本、lint 与分发核验后,再把修改应用到当前任务。
  5. reusable 修改会使此前的“确认”“继续”“发吧”失效;完成当前产物修改和回读后必须停下,等待用户下一步指示,不自动进入发布、提交或其他外部写入。
Skills enregistrées