---

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

Level	Criteria	Examples
A (Must fix)	Meaning changed, factual error, or would cause rejection	Mistranslation, wrong terminology, reference number mismatch
B (Recommended)	Awkward but understandable, or non-standard usage	Verbose phrasing, minor grammar, style inconsistency
C (Source issue)	Error in Chinese original, faithfully translated	Wrong 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。