Skill: PDF Pro (Standard 2026)

🎯 Primary Objectives

1. Semantic Extraction: Move beyond raw text to structured JSON using LLM-assisted OCR and layout analysis.
2. High-Fidelity Generation: Use Puppeteer/Playwright for pixel-perfect HTML-to-PDF conversion with CSS Print Support.
3. PDF 2.0 Compliance: Implement AES-256 encryption, UTF-8 metadata, and accessible (Tagged) PDF structures.
4. Edge-Ready Processing: Use lightweight libraries like

   unpdf

   for serverless and edge environments.

🏗️ The 2026 Toolbelt

1. Bun-Native & JS Libraries (Primary)

• pdf-lib: Byte-level modification, merging, splitting, and form filling.
• unpdf: Ultra-lightweight extraction for Edge/Serverless.
• Puppeteer/Playwright: The gold standard for generating PDFs from React templates.
• Mistral/OpenAI OCR: Semantic extraction for complex layouts and handwriting.

2. Forensic Utilities (Legacy/Advanced)

• qpdf: CLI tool for structural repairs and decryption.
• poppler-utils: Fast C-based text and image extraction.

🛠️ Implementation Patterns

1. High-Fidelity Generation (Next.js 16.2)

// app/api/generate-pdf/route.ts
import puppeteer from 'puppeteer';

export async function POST(req: Request) {
  const { htmlContent } = await req.json();
  const browser = await puppeteer.launch({ headless: true });
  const page = await browser.newPage();
  
  await page.setContent(htmlContent, { waitUntil: 'networkidle0' });
  const pdfBuffer = await page.pdf({
    format: 'A4',
    printBackground: true,
    margin: { top: '20px', bottom: '20px' }
  });

  await browser.close();
  return new Response(pdfBuffer, {
    headers: { 'Content-Type': 'application/pdf' }
  });
}

2. AI-Driven Semantic Extraction

import { unpdf } from 'unpdf';
import { generateObject } from 'ai'; // AI SDK 2026

async function extractInvoice(buffer: Buffer) {
  const { text } = await unpdf.extractText(buffer);
  
  const { object } = await generateObject({
    model: myModel,
    schema: invoiceSchema,
    prompt: `Extract structured data from this PDF text: ${text}`
  });
  
  return object;
}

🔒 PDF 2.0 Security & Integrity

AES-256 Encryption

qpdf

# Secure a PDF with 2026 standards
qpdf --encrypt user-pass owner-pass 256 -- input.pdf secured.pdf

Digital Signatures (PAdES)

🚫 The "Do Not List" (Anti-Patterns)

1. NEVER use

   pypdf

   for complex layout extraction; it fails on multi-column or overlapping text. Use

   pdfplumber

   or AI OCR.
2. NEVER generate PDFs using

   canvas

   drawing commands if HTML/CSS templates are an option. Maintenance is a nightmare.
3. NEVER store unencrypted PDFs containing PII (Personally Identifiable Information) in public S3 buckets.
4. NEVER rely on

   window.print()

   for automated server-side generation. It is non-deterministic.

🛠️ Troubleshooting Guide

poppler

-enc UTF-8

ghostscript

pdf-lib

pdf-lib

AcroForm

📚 Reference Library

• AI Extraction Patterns: Mastering semantic document understanding.
• High-Fidelity Generation: HTML-to-PDF at scale.
• Legacy Utilities: When to reach for Python/CLI tools.

📜 Standard Operating Procedure (SOP)

1. Requirement Check: Is the goal Creation, Extraction, or Modification?
2. Tool Selection:
   • Creation -> Puppeteer.
   • Extraction -> AI SDK + unpdf.
   • Modification -> pdf-lib.
3. Environment Check: Is this running in an Edge Function? (If yes, avoid Puppeteer).
4. Implementation: Build with strict TypeScript typing.
5. Audit: Verify PDF 2.0 metadata and accessibility (A11y) tags.

📈 Quality Metrics

• Extraction Accuracy: > 98% (Measured against ground truth JSON).
• Generation Speed: < 2s for a 10-page document.
• Security Audit: Zero weak crypto algorithms (Verified via

  qpdf

  ).

🔄 Last Refactor Details

• By: Gemini Elite Conductor
• Date: January 22, 2026
• Version: 1.1.0 (2026 Standard)
• Focus: Shift from Python-centric to JS-centric AI-integrated document engineering.