PDF Proof

What it does

PDF Proof locates specific values or phrases inside PDFs, crops tight, readable screenshots around matches, draws precise highlights using PDF coordinates, verifies the highlighted text by reading it back, and assembles a shareable HTML proof page with confidence badges and per-value proof cards. The primary output is a standalone HTML file plus PNG screenshots that form a visual audit trail.

When to use it

Use when you need visual evidence that a specific number, clause, invoice line, or contract phrase appears in a PDF — e.g., tax line items, invoice totals vs PO, indemnification caps in contracts, or verified receipts. Do NOT trigger for generic PDF tasks like merging, converting, or summarizing; this skill is for targeted verification where a highlighted, verifiable screenshot is required.

What's included

• Scripts: bundled extraction utilities (extract_proof.py) for search, crop, highlight, and verification (if has_scripts=true these would be present).
• References: template HTML and guidance for confidence badges and output structure.
• Instructions: step-by-step process: find matches, prefer heuristics (right/first), run verify mode to read back text, generate HTML proof page, and save outputs in a named subfolder.

Compatible agents

Best for agents with Python/script execution capability (Claude Code, Copilot-style agents, or local Python runners). Agents that can run PyMuPDF and optionally Tesseract OCR will get full functionality.