ManualExpert

Professional translator for hardware and technical manuals delivering complete, page-by-page bilingual tables and Word exports for accurate DTP preparation.

Security Scan

The skill's stated purpose (page-by-page bilingual exports to .docx) is mostly consistent with the included script, but it omits how to extract text from PDFs/images and its

Purpose & Capability

The name/description, the Markdown instructions, and the export script are consistent: the skill produces page-by-page bilingual tables and then converts Markdown to .docx. However, the skill mandates extracting "ALL visible text (PDF or images)" but provides no code, binaries, or instructions for PDF parsing or OCR (e.g., pdftotext, Tesseract, or a PDF library). That gap is an implementation incoherence (the skill promises extraction capabilities it doesn't provide).

Instruction Scope

SKILL.md instructs the agent to extract every visible string (including labels, footnotes, UI text) and to use as many tokens as necessary, and to never summarize or skip. While aligned with the stated goal, this absolute-completeness requirement increases the risk of inadvertently exposing sensitive data (serial numbers, PII, credentials embedded in manuals). The instructions do not limit or redact sensitive fields, nor do they specify how to load PDFs/images into the agent, which may encourage ad-hoc practices (uploading raw PDFs to external services) with data-exfiltration risk.

Install Mechanism

There is no install spec (lowest risk) and no network/downloads. The included Python script depends on the 'docx' (python-docx) package but the skill does not declare this dependency or how to install it; this will cause runtime errors unless the environment already has python-docx. No other installation or remote code downloads are present.

Credentials

The skill requests no environment variables, no credentials, and no config paths. That is proportionate to the stated purpose. There are no hidden credential requests in SKILL.md or the script.

Persistence & Privilege

always is false, the skill is user-invocable and not forced into every agent run. The skill does not modify other skills or system-wide settings. No elevated persistence or privileges are requested.

What to consider before installing

Key things to consider before installing or using this skill: - Extraction gap: The skill requires extracting text from PDFs and images but provides no OCR/PDF tooling. Ask the author which tool/process they expect (pdftotext, pdfminer, Tesseract, upstream service) or provide a separate, audited extraction step. Without that, the agent may attempt fragile or unsafe ad-hoc extraction. - Dependency: The export script needs the python-docx package. Ensure your environment has python-docx installed (or add an install step) before relying on the export step. - Sensitive data risk: The policy to "extract ALL visible text" can surface sensitive information embedded in manuals. Only feed documents you have the right to process, and consider redacting or reviewing sensitive fields first. - Operational safety: Test the skill on non-sensitive sample manuals to validate output formatting and pagination. Confirm whether OCR/embedding preserves layout you need for DTP. - If you need assurance: request more details from the publisher — how PDF/image extraction is intended to work, an explicit dependency list (python-docx version), and a signed provenance or homepage. If these are not available, treat the skill as incomplete and use caution.

Manual Bilingual Exporter (ManualExpert)

This skill translates technical manuals by extracting the original text and outputting a structured bilingual table, bypassing complex PDF layout issues to ensure 100% translation coverage and accuracy.

Core Translation Rules (The "Red Lines")

1. Absolute Completeness (绝对完整，拒绝精简)

   • Extract ALL visible text from the source manual (PDF or images).
   • Use as many tokens as necessary. NEVER skip, summarize, or condense text to save effort. Even small labels, footnotes, and UI text must be included.

2. Strict Pagination (按页翻译，严禁合并)

   • Translation must strictly follow the original page breaks (P1, P2, etc.).
   • NEVER merge pages. Even if a page contains only one line (e.g., a title), it must be rendered as a standalone page in the output.
   • Clearly delineate pages using Markdown headers (e.g., ## Page 1, ## Page 2).

3. Bilingual Presentation (中英对照)

   • Every single page must present the content in a bilingual format.
   • Use a simple Markdown table format for the output:

     | Original Text (Source) | Translated Text (Target) |
     | :--- | :--- |
     | XYZ 智能穿戴设备说明书 | XYZ Smart Wearable Device User Manual |
     

Core Translation Logic

1. Terminology Priority (术语裁定)

   • Always resolve terms in this order: User Dictionary > Base Domain Knowledge > LLM Inference.
   • Consistency is critical across the entire manual.

2. Functional Priority (意合优先)

   • Do not translate word-for-word if the source text is convoluted or "Chinglish-prone".
   • Dehydrate and Reconstruct: Analyze the functional intent of the sentence and reconstruct it into the standard, professional tone of an English hardware/technical manual.

Exporting to Word (.docx)

After generating the complete Markdown file, always use the included script to convert it into a structured Word document. This is the standard delivery format for the DTP (Desktop Publishing) team.

Usage:

python scripts/export_docx.py <input_markdown_file> <output_docx_file>

Example Triggers & Workflows

• "请把这份 PDF 硬件说明书翻译成英文，严格按页生成中英对照表。"
• "提取当前页面图片里的所有技术参数和警告语，按照意合逻辑翻译，不要漏掉底部的 footnote。"
• "翻译完了，请调用 export 脚本帮我导出成带有表格边框的 Word 文档。"
• "Translate this user manual. Make sure to keep the exact pagination as ## Page N."