document-processor-plus

Built on affaan-m/ECC by @affaan-m (259,939 stars, MIT). All credit for the original idea to them. This fork improves and repackages it; upstream license preserved in UPSTREAM_LICENSE.

A Claude Code skill that helps you convert, read, scan, edit, sign, and fill documents with the Nutrient DWS API.

Built for developers who want Claude Code to handle PDF, DOCX, XLSX, PPTX, HTML, and image files.

💡 Why

Document APIs can be hard to use safely.

File names must match. API keys must stay private. OCR can miss text. File changes can harm fonts or page layout. A redaction can look safe while the old text is still there.

This skill gives Claude Code clear steps, safe rules, and ready-to-use API calls. It plugs into your current Claude Code flow.

📦 Install

Run one command:

mkdir -p ~/.claude/skills/nutrient-document-processing && curl --fail-with-body https://raw.githubusercontent.com/YOUR_GITHUB_NAME/document-processor-plus/main/skill/SKILL.md -o ~/.claude/skills/nutrient-document-processing/SKILL.md

You also need a Nutrient DWS API key. Set it in your shell:

export NUTRIENT_API_KEY="pdf_live_..."

Keep this key out of code, logs, and saved files.

🚀 Usage

Ask Claude Code in plain English:

Use the nutrient-document-processing skill to turn report.docx into report.pdf.
Keep the source file. Check that the API call worked, then tell me where the PDF was saved.

Expected output:

Created report.pdf.
The API request succeeded.
The source file report.docx was kept.
Please check the PDF for font and layout changes.

The skill can also help you:

• Read text or tables from a PDF.
• Run OCR on scans and images.
• Hide private text with redaction.
• Add a text watermark.
• Sign a PDF.
• Fill PDF form fields.
• Change common document and image file types.

🔧 What we changed vs upstream

• The improved file is rewritten from Japanese into clearer, plain English with a more scannable task-oriented structure.
• It adds substantial safety guidance for API keys, private files, output validation, redaction testing, retries, and source-file preservation.
• All curl examples now use --fail-with-body, making HTTP failures easier to detect and diagnose.
• It adds practical warnings about conversion fidelity, OCR prerequisites, and possible table-extraction errors.
• The OCR example switches from english to the standard eng code and explains common language codes more plainly.

📄 License

This project uses the MIT License. See LICENSE.

The upstream MIT license and credit are preserved in UPSTREAM_LICENSE.