AgentSkillsCN

doc-to-text

支持从多种文件格式中提取文本内容，涵盖PDF、Excel、Word、图片（OCR）、电子邮件及ZIP压缩包等多种类型，可用于二进制文件的摘要提炼或深度分析。

SKILL.md
--- frontmatter
name: doc-to-text
description: Extract text content from various file formats. Supports PDF, Excel, Word, Images (OCR), Email, and ZIP Archives. Use for summarizing or analyzing binary files.
status: implemented

Document to Text Converter

Overview

This skill acts as a universal converter to extract plain text and structured data from various binary and complex file formats. It enables Gemini to "read" files that are otherwise inaccessible.

Capabilities

1. Document Extraction

  • •PDF (.pdf): Extracts plain text.
  • •Excel (.xlsx): Converts sheets to CSV and performs OCR on embedded images.
  • •Word (.docx): Extracts text and performs OCR on embedded images.
  • •PowerPoint (.pptx): Extracts slide text and performs OCR on embedded images.

2. Image OCR

  • •Images (.png, .jpg, .jpeg, .webp): Uses Tesseract.js to perform OCR (Optical Character Recognition) and extract text from images. Supports English and Japanese.

3. Data & Archives

  • •Email (.eml): Parses headers (From, To, Subject) and body text.
  • •ZIP Archive (.zip): Lists files and extracts content of text-based files within the archive without extracting to disk.

Usage

To read a file, execute the extract.cjs script with the file path.

bash
node scripts/extract.cjs <path/to/file>

Example: User: "What does the error screenshot say?" Action: node scripts/extract.cjs error.png

Dependencies

This skill requires Node.js packages. Run npm install in the skill directory before using.

Knowledge Protocol

  • •This skill adheres to the knowledge/orchestration/knowledge-protocol.md. It automatically integrates Public, Confidential (Company/Client), and Personal knowledge tiers, prioritizing the most specific secrets while ensuring no leaks to public outputs.