markdown-segment-translator

Markdown segmented translation skill

⚠️ Mandatory requirements - 3-step translation process ⚠️

When the user asks you to translate a markdown file, youmust strictly follow the following 3 stepsto complete it, do not skip any step!

📋 Step 1: Document segmentation

Tool: Bash
Command format:

python3 workspace/skills/markdown-segment-translator/scripts/segment_tool.py segment --input [输入文件] --segment-size 1000 --max-segments 3

Purpose:

Split a large document into multiple small segments
Segment files are saved in output/segments/segment_1.md, segment_2.md...
Metadata is saved in output/metadata/segments_info.json

Example:

python3 workspace/skills/markdown-segment-translator/scripts/segment_tool.py segment --input workspace/2407.14333v5.md --segment-size 1000 --max-segments 3

📋 Step 2: Translate each segment

Tool: Read + Your own LLM capability + Write

Important: You must translate by yourself, do not call any external API!

Process:

For each segment file:

Read the segment: Use Read to read output/segments/segment_1.md
Translate by yourself: Use your own language ability to translate the content from English to Chinese
Save the translation: Use Write to save to output/translations/translated_segment_1.md
Repeat: Process segment_2.md, segment_3.md...

Translation requirements:

Keep all Markdown formatting (headings, lists, code blocks, tables, etc.)
Accurately translate academic terminology
Maintain professionalism and fluency
Do not translate code, formulas, URLs
Keep the paragraph structure

Translation prompt template:

请将以下Markdown内容从英文翻译为中文：
- 保持所有Markdown格式标记
- 准确翻译学术术语
- 不要翻译代码块、数学公式、URL链接
- 保持专业性

[段落内容]

📋 Step 3: Merge the translation results

Tool: Bash
Command:

python3 workspace/skills/markdown-segment-translator/scripts/segment_tool.py merge

Purpose:

Merge all translated segments into a complete file
The final file is located at: output/final/complete_translated_[原文件名].md

✅ Complete example

用户请求: 请翻译2407.14333v5.md文件

你的执行步骤:

【第1步 - 分段】
Bash: python3 workspace/skills/markdown-segment-translator/scripts/segment_tool.py segment --input workspace/2407.14333v5.md --segment-size 1000 --max-segments 3

输出: 创建 segment_1.md (1000行), segment_2.md (1000行), segment_3.md (700行)

【第2步 - 翻译】
循环处理每个段落:

Segment 1:
  Read: output/segments/segment_1.md
  [你自己翻译这段内容为中文]
  Write: output/translations/translated_segment_1.md (写入你的翻译)

Segment 2:
  Read: output/segments/segment_2.md
  [你自己翻译这段内容为中文]
  Write: output/translations/translated_segment_2.md

Segment 3:
  Read: output/segments/segment_3.md
  [你自己翻译这段内容为中文]
  Write: output/translations/translated_segment_3.md

【第3步 - 合并】
Bash: python3 workspace/skills/markdown-segment-translator/scripts/segment_tool.py merge

输出: output/final/complete_translated_2407.14333v5.md

【完成】
使用 Read 读取最终文件并向用户报告

❌ Strictly prohibited operations

❌ Do not use the old segment_translator.py script (that one calls the API by itself)
❌ Do not let a Python script call a translation API
❌ Do not skip the segmentation or merging steps
❌ Do not try to translate the entire document all at once
❌ Do not use Bash in Step 2 to call an external translation program

✅ Correct tool call sequence

第1步: Bash (分段工具)
第2步: Read → [你自己翻译] → Write (循环N次)
第3步: Bash (合并工具)
第4步: Read (读取最终结果)

🎯 Why design it this way?

Segmentation: Avoid token limits and ensure complete translation
Agent translation: Leverage your LLM capability without additional API calls
Segment-by-segment processing: Ensure quality and accuracy
Merging: Generate a complete and unified translated document

Skill function overview

This skill specifically solves the token limit problem in long document translation, ensuring 100% complete translation through intelligent segmentation and the Agent's autonomous translation mechanism.

Core advantages:

🔧 Purely tool-based segmentation/merging
🧠 Agent-led translation
📊 Completeness guarantee
🎨 Format preservation
📝 Terminology consistency