Doc Translate

Office文档翻译工具集。用于翻译和本地化办公文档（PPT/Word/PDF/Excel），保持原文档模板和格式不变。当用户说"翻译PPT"、"翻译Word文档"、"翻译PDF"、"翻译Excel"、"翻译文档"、"文档本地化"、"英文文档翻译成中文"时使用本技能。支持PPT/Word/PDF/Excel四种格...

Security Scan

The skill's code and instructions match its stated purpose (translating DOCX/PPTX/XLSX/PDF) and do not request unrelated credentials, but there are a few operational inconsistencies and missing dependency declarations you should be aware of before installing.

Purpose & Capability

The name/description (document translation/localization) matches the provided scripts: pptx_translate.py, docx_translate.py, xlsx_translate.py, pdf_translate.py implement ZIP/XML extraction, mapping-based translations, and PDF -> DOCX -> PDF conversion. The presence of mapping tables and format-specific logic is appropriate for the stated purpose. One minor mismatch: SKILL.md documents Feishu (Lark) download/upload snippets but the included scripts do not implement Feishu API calls or request Feishu credentials — SKILL.md assumes the agent/platform provides Feishu I/O.

Instruction Scope

SKILL.md gives a narrowly scoped workflow: download file (Feishu or local), detect format, extract/parse, translate, rebuild, and return file. The scripts operate only on the input files and temporary directories and do not read unrelated system files or environment variables. The PDF 'direct' overlay method is noted as not fully implemented. There are no instructions to transmit data to unknown external endpoints from inside the scripts.

Install Mechanism

This is instruction-only (no install spec), which reduces install risk, but the scripts have runtime dependencies that are not declared: pdfplumber, python-docx, potentially pypdf2/reportlab, and LibreOffice (soffice) for DOCX→PDF conversion. The skill does call subprocess.run(['soffice', ...]) for conversion; the absence of a dependency/install section means the user/agent must provide these tools. No downloaded arbitrary archives or untrusted URLs are used.

Credentials

The skill declares no required env vars or credentials, and the scripts do not read secrets. However, SKILL.md includes code snippets for downloading and uploading files via Feishu (message_id, file_token, feishu_drive_file) which in practice require platform credentials or runtime context. The skill does not declare those credentials — this is a documentation/expectation mismatch rather than evidence of secret access.

Persistence & Privilege

always is false and the skill does not request persistent privileges, modify other skills, or write to global agent configuration. Scripts create temporary directories and write output files only. No elevated privileges or always-on behavior observed.

Assessment

This skill appears to do what it says: it processes Office files locally by extracting XML/text, applying mapping-based replacements, and rebuilding archives. Before installing or running it: (1) Confirm required runtime dependencies (pdfplumber, python-docx, reportlab/pypdf2 as needed, and LibreOffice/soffice if you want PDF conversion) and install them in a controlled environment. (2) If you plan to use the Feishu download/upload snippets from SKILL.md, verify how your platform supplies Feishu tokens or message/file tokens — the skill does not request these credentials, so uploading/downloading will rely on your agent/platform integration. (3) Do not run this on highly sensitive documents unless you trust the execution environment — the scripts operate on local files and create intermediates in temp directories. (4) Review the translation mapping tables for proprietary/trademarked terms and medical/technical terms (the SKILL.md explicitly instructs preserving such terms). (5) If you need network-based translation (e.g., external AI APIs), note these scripts currently use only mapping rules and local libraries; no remote API calls are present. Overall the package is coherent and low-risk, but ensure dependencies and Feishu integration are handled explicitly and run it in an isolated environment if documents contain sensitive data.

Office document translation toolkit / Office文档翻译工具集

Supports PPTX / DOCX / PDF / XLSX four formats, preserving original templates and styles / 支持四种格式，保持原模板不变。

---

工作流 / Workflow

通用流程 / General Process（所有格式 / All Formats）

1. 下载文件 / Download file      → 飞书附件 or 本地路径 / Feishu attachment or local path
2. 格式识别 / Detect format       → 根据扩展名判断 / Detect by extension
3. 解压解析 / Extract & parse     → ZIP解压 → XML/文本节点提取 / ZIP extract → XML/text node extraction
4. 翻译 / Translate              → 规则映射或AI翻译 / Rule mapping or AI translation
5. 重建 / Rebuild                → 重新打包，保留原样式 / Repack, preserve original styles
6. 发送 / Deliver                → 发送给用户或保存 / Send to user or save

---

PPTX 翻译 / PPTX Translation

格式结构 / Format Structure

PPTX = ZIP压缩包，内含 / ZIP archive containing:

• ppt/slides/slide1.xml … slideN.xml — 幻灯片内容 / Slide content
• ppt/theme/ — 主题样式 / Theme styles
• [Content_Types].xml — 内容类型声明 / Content type declarations

关键XML节点 / Key XML Nodes

节点 / Node	说明 / Description
<a:t>	文本内容 / Text content（最常出现 / most common）
<a:r>	文本段落 / Text run
<p:sp>	形状（文本框等）/ Shape（text box etc.）
<p:txBody>	文本正文 / Text body

处理脚本 / Processing Script

python3 scripts/pptx_translate.py <pptx文件路径> [--output <输出路径>] [--src-lang en] [--tgt-lang zh]

---

DOCX 翻译 / DOCX Translation

格式结构 / Format Structure

DOCX = ZIP压缩包，内含 / ZIP archive containing:

• word/document.xml — 主文档内容 / Main document content
• word/styles.xml — 样式 / Styles
• [Content_Types].xml

关键XML节点 / Key XML Nodes

节点 / Node	说明 / Description
<w:t>	文本内容 / Text content
<w:p>	段落 / Paragraph
<w:r>	文本段 / Text run

处理脚本 / Processing Script

python3 scripts/docx_translate.py <docx文件路径> [--output <输出路径>]

---

XLSX 翻译 / XLSX Translation

格式结构 / Format Structure

XLSX = ZIP压缩包，内含 / ZIP archive containing:

• xl/worksheets/sheet1.xml — 工作表内容 / Worksheet content
• xl/sharedStrings.xml — 共享字符串表（单元格文本） / Shared strings table（cell text）
• xl/styles.xml — 样式 / Styles

处理脚本 / Processing Script

python3 scripts/xlsx_translate.py <xlsx文件路径> [--output <输出路径>]

---

PDF 翻译 / PDF Translation

方法说明 / Method Notes

PDF不是XML格式，无法直接编辑文本。推荐方案 / PDF is not XML, cannot edit text directly. Recommended approach:

方案A / Method A：PDF → Word → PDF（推荐 / Recommended）

pdfplumber提取文本 → python-docx构建翻译后Word文档 → LibreOffice导出PDF
pdfplumber extract → python-docx build translated Word → LibreOffice export PDF

方案B / Method B：直接覆盖（仅文本层PDF） / Direct overlay (text-layer PDFs only） 使用 pdfplumber + pypdf2 / Using pdfplumber + pypdf2

处理脚本 / Processing Script

python3 scripts/pdf_translate.py <pdf文件路径> [--output <输出路径>] [--method word|direct]

---

飞书文件处理 / Feishu File Handling

下载飞书附件 / Download Feishu Attachment

# 飞书消息附件下载（message_id 来自上下文）/ Feishu message attachment download
feishu_im_bot_image(message_id="om_xxx", file_key="file_v3_xxx", type="file")
# 返回 saved_path / Returns saved_path

# 飞书云盘文件下载 / Feishu drive file download
feishu_drive_file(action="download", file_token="xxx", output_path="/tmp/output.pptx")

发送文件给用户 / Send File to User

方法1：消息附件发送（限制25MB）/ Method 1: Message attachment (25MB limit）

message(action="send", channel="feishu", file_path="/path/to/file.pptx", target="user:ou_xxx")

方法2：分卷发送（文件>25MB时）/ Method 2: Split send (file > 25MB）

# 分成10MB分卷 / Split into 10MB chunks
split -b 10m original.pptx part-
# 逐个发送 / Send one by one
message(action="send", file_path="part-aa", target="user:ou_xxx")
# 用户合并命令 / User merge command:
cat part-* > original.pptx

方法3：上传飞书云盘 / Method 3: Upload to Feishu Drive

feishu_drive_file(action="upload", file_path="/path/to/file.pptx", file_name="翻译版.pptx")

---

翻译质量注意事项 / Translation Quality Notes

1. 医学/专业术语 / Medical/Technical Terms — 必须使用人工翻译映射表，不能让模型自由发挥 / Must use manual translation mapping, not free-form AI
2. 商标/品牌 / Trademarks/Brands — 保持原文，如 Medtronic、Harmonic 等 / Keep original, e.g. Medtronic, Harmonic
3. 数字+单位 / Numbers+Units — 保持原样，如 7mm、150 procedures / Keep as-is, e.g. 7mm, 150 procedures
4. 脚注符号 / Footnote Symbols — † ‡ § Ω 等保持不变 / Keep as-is: † ‡ § Ω
5. 格式占位符 / Format Placeholders — 如 {name}、%s 等不替换 / Don't replace: {name}, %s

---

脚本列表 / Scripts

脚本 / Script	功能 / Function	输入 / Input	输出 / Output
pptx_translate.py	PPTX全文翻译 / Full PPTX translation	.pptx文件	.pptx（中文版 / Chinese version）
docx_translate.py	DOCX全文翻译 / Full DOCX translation	.docx文件	.docx（中文版 / Chinese version）
xlsx_translate.py	XLSX全文翻译 / Full XLSX translation	.xlsx文件	.xlsx（中文版 / Chinese version）
pdf_translate.py	PDF全文翻译 / Full PDF translation	.pdf文件	.pdf（中文版 / Chinese version）

All scripts use --output to specify output path / 所有脚本使用 --output 指定输出路径