pdf-pro_skill

Overview

This skill transforms PDF engineering into a production-ready, AI-first workflow for extraction, generation, and secure modification. It combines semantic OCR, pixel-perfect HTML-to-PDF generation, and byte-level forensic edits to meet PDF 2.0 compliance and edge-ready deployment needs. The design favors Bun/JavaScript patterns while preserving advanced utilities for legacy or heavy-duty tasks.

How this skill works

PDF Pro inspects PDF bytes, page layout, and embedded resources to produce structured JSON via LLM-assisted OCR and layout analysis. For generation it renders React/HTML using Puppeteer or Playwright to create high-fidelity PDFs with CSS print support. For modifications it performs low-level edits with pdf-lib and enforces PDF 2.0 security (AES-256, UTF-8 metadata, tagged PDFs). Edge-friendly extraction uses unpdf and serverless-friendly libraries when headless browsers are unavailable.

When to use it

• Extract validated, schema-typed data from invoices, contracts, and forms using AI-driven OCR.
• Generate pixel-perfect PDFs from React components when visual parity with the web app is required.
• Encrypt or digitally sign documents to meet PDF 2.0 security and regulatory requirements.
• Perform forensic edits, merge/split, or programmatic form-filling at the byte level.
• Run lightweight extraction in Edge/Serverless environments where Puppeteer is not allowed.

Best practices

• Select tools by task: Puppeteer/Playwright for creation, unpdf+LLM for extraction, pdf-lib for modification.
• Embed fonts (WOFF2/Google Fonts) in generation containers to avoid missing-font issues.
• Validate extraction output against strict schemas (Zod or similar) and measure accuracy versus ground truth.
• Never store unencrypted PII-containing PDFs in public buckets; always use AES-256 or HSM-backed signatures.
• Avoid canvas-drawing PDF generation when HTML/CSS templates exist — use headless rendering for maintainability.

Example use cases

• Auto-extract invoice line items and map them to accounting systems as JSON with >98% accuracy.
• Render monthly statements from React templates into searchable, tagged PDFs for accessibility.
• Securely encrypt and PAdES-sign contracts using OIDC/HSM integration before distribution.
• Optimize document processing on edge functions by using unpdf for text extraction and falling back to AI OCR for complex layouts.