name: pdf-pro description: professional pdf creation, editing, review, conversion, extraction, visual verification, forms, redaction, cli automation, and presentation-quality layout design. use when chatgpt needs to create a new pdf, modify an existing pdf, inspect or summarize a pdf, extract text/tables/images, combine/split/rotate/crop/watermark pages, fill or flatten forms, redact content, convert documents to pdf, validate outputs, or improve the visual design of a pdf using ui/ux and design-system guidance.

PDF Pro

Use this skill to handle PDF work as a verified production workflow, not a one-step file operation. Prioritise accuracy, visual verification, safe PDF operations, and clean design.

Core operating rule

Always use this loop unless the user only asks for a high-level explanation:

1. Identify the PDF task: create, review, extract, edit, convert, redact, form-fill, visual redesign, or CLI automation.
2. Choose the safest authoring path:
   • Text-heavy reports, policies, letters, and long documents: author in DOCX or Markdown first, then convert to PDF when tooling supports it.
   • Slide-like layouts, posters, one-pagers, certificates, dashboards, and visual reports: use fixed-layout PDF creation or slide authoring, then export and verify.
   • Existing PDF edits: inspect and render pages before editing when layout matters.
   • CLI/automation tasks: use the bundled scripts or the pdf-pro CLI interface when the package is installed.
3. Inspect or render the input before changing it when working with an existing PDF.
4. Perform the smallest safe operation that satisfies the request.
5. Re-render or inspect the output.
6. Validate the final PDF using scripts/validate_pdf.py when a file is produced or modified.
7. Deliver the final PDF plus a short summary of what changed, what was verified, and any limitations.

Never claim a PDF is correct without checking the output file exists and can be opened or inspected. For final deliverables, follow references/validation-protocol.md and record any limitation clearly.

Bundled helpers

Use these scripts when they fit the task. Run scripts from inside the skill folder or pass absolute paths.

• scripts/pdf_ops.py: inspect, merge, split, rotate, crop, watermark, redact text, extract text, and render pages to PNG for verification.
• scripts/create_visual_pdf.py: generate a polished PDF from a JSON content specification using design tokens.
• scripts/validate_pdf.py: validate a final PDF by checking file existence, parser readability, page count, optional text extraction, forbidden/required phrases, and optional page renders.
• scripts/validate_package.py: validate that the repository contains the expected skill, platform, CLI, script, reference, and asset files.
• README.md: user and maintainer documentation, including ownership, licence, dependencies, platform compatibility, CLI usage, and examples.
• LICENSE: MIT licence for PDF Pro.
• NOTICE.md: attribution and third-party dependency notices, including UI/UX Pro Max by NextLevelBuilder.
• references/pdf-workflows.md: exact workflows for creation, editing, extraction, forms, redaction, conversion, and verification.
• references/visual-design-system.md: UI/UX Pro Max inspired design rules for high-quality PDF layouts.
• references/quality-checklist.md: final QA checklist before delivery.
• references/validation-protocol.md: formal verification rules for final PDF outputs.
• references/cli-integration.md: CLI and editor integration guidance for VSCode, Antigravity, Claude Code, Copilot, and other coding agents.
• references/platform-compatibility.md: platform-specific usage guidance for ChatGPT Skills, Claude Skills, Gemini Gems, CLI agents, and local LLM workflows.
• references/compatibility-checklist.md: checklist for release validation across supported targets.
• assets/design-tokens-starter.json: reusable colour, spacing, type, radius, and shadow tokens.
• assets/pdf.png: skill icon.

Creation workflow

For a new PDF:

1. Decide the output type: report, handout, visual one-pager, packet, form, certificate, invoice, worksheet, dashboard, or portfolio.
2. Load references/visual-design-system.md when the user expects professional visual design, branding, layout polish, colour palette, typography, charts, or a PDF that should look presentation-ready.
3. Use scripts/create_visual_pdf.py for quick fixed-layout PDFs from structured JSON, or use a richer document/slide workflow when the content is complex.
4. Render the finished PDF using scripts/pdf_ops.py render and check the images for clipped text, overlaps, broken glyphs, poor margins, and weak hierarchy.
5. Validate the finished file with scripts/validate_pdf.py. Use --render for visual or client-facing PDFs.
6. If the PDF is visual or client-facing, run the final review against references/quality-checklist.md.

Default visual standard: restrained, readable, accessible, and polished. Avoid decoration that makes the document harder to scan.

Editing workflow

For an existing PDF:

1. Inspect first:

   python scripts/pdf_ops.py inspect input.pdf
   

2. Render pages when layout matters:

   python scripts/pdf_ops.py render input.pdf --out-dir renders/input --dpi 160
   

3. Apply the smallest safe edit: merge, split, rotate, crop, watermark, extraction, form filling, or redaction-specific workflow.
4. Inspect and render the output.
5. Summarise exactly which pages changed.

For true redaction, never cover text with a black rectangle only. Use a redaction-capable workflow that removes underlying text or pixels, then verify by text extraction and rendering.

Review and extraction workflow

For review, summary, or extraction:

1. Inspect the file metadata, page count, encryption status, and text availability.
2. Extract text when the PDF has a text layer.
3. Render pages when the user asks about visual layout, diagrams, tables, charts, signatures, scans, stamps, screenshots, or placement.
4. Flag OCR uncertainty for scanned pages. Do not overstate confidence when the text layer is missing or poor.
5. For tables, compare extracted data with page renders when accuracy matters.

UI/UX Pro Max integration

Use design-system thinking for any PDF with a visible layout. Apply these priorities:

1. Accessibility: readable type, sufficient contrast, clear hierarchy, labelled charts, and no colour-only meaning.
2. Layout: consistent grid, generous margins, aligned edges, predictable spacing, and no horizontal crowding.
3. Typography: limited font families, clear scale, line height around 1.4-1.7 for body copy, and tabular figures for data.
4. Colour: semantic palette with primary, secondary, background, foreground, border, muted, success, warning, and destructive tokens.
5. Visual polish: restrained shadows, consistent radius, intentional whitespace, and chart choices matched to the data.
6. Verification: render and inspect the PDF as images before delivery.

For premium visual PDFs, treat text as part of the composition. Use short headings, clear callouts, and purposeful visual rhythm instead of dense paragraphs.

CLI and agent integration

When the package is installed locally, the same helpers can be called with the pdf-pro command:

pdf-pro inspect input.pdf
pdf-pro render input.pdf --out-dir renders/input --dpi 160
pdf-pro validate output.pdf --render --out-dir validation/output
pdf-pro visual examples/visual-spec.json --out output.pdf

For VSCode, Antigravity, Claude Code, Copilot, and other coding agents, load AGENTS.md, CLI.md, and references/cli-integration.md when the user asks for editor automation, terminal workflows, package installation, or repository integration.

Platform compatibility

Use this package directly as a ChatGPT Skill. For Claude Skills, Claude Code, Gemini Gems, Copilot, Antigravity, and local agent workflows, use the platform notes in CLAUDE.md, GEMINI.md, AGENTS.md, CLI.md, and platforms/.

Do not overclaim a platform feature. If the target assistant cannot execute code, provide the commands for a local terminal or coding-agent environment instead.

Response format after completion

Use this structure in the final message when a PDF is created or edited:

• State the output file link.
• Summarise the main changes or contents.
• State what verification was completed.
• Mention limitations only when relevant, such as missing fonts, scanned pages, OCR uncertainty, password protection, unavailable source files, or third-party dependency limits.

Keep the user-facing summary concise. Do not expose internal script logs unless the user asks.

PDF Pro

PDF Pro is a production-ready PDF workflow package for AI assistants and local automation.

It is designed first as a ChatGPT Skill and structured to be adaptable for Claude Skills, Claude Code, Gemini Gems, VSCode, GitHub Copilot, Antigravity, and other coding-agent or CLI workflows.

PDF Pro helps assistants and developers create, edit, inspect, extract, render, validate, redact, and visually improve PDF files through a verification-first process. It combines practical PDF operations with design-system guidance so final documents are usable, visually checked, accurate, and ready to share.

Repository description

A professional AI workflow package for creating, editing, reviewing, extracting, validating, redacting, and visually improving production-ready PDF files across ChatGPT Skills, Claude Skills, and CLI-based coding agents.

What this package supports

• Creating new PDFs from structured content.
• Creating polished visual PDFs from JSON specifications.
• Editing existing PDFs through safe operations such as merge, split, rotate, crop, watermark, redaction, and extraction.
• Reviewing PDFs by inspecting metadata, page count, encryption status, page sizes, rotation, and text layer availability.
• Rendering PDF pages to PNG for visual verification.
• Extracting text from selectable-text PDFs.
• Validating final PDF outputs before delivery.
• Checking required and forbidden phrases during validation.
• Applying professional visual design rules for client-facing PDFs.
• Supporting local CLI workflows for editors and coding agents.
• Providing platform guidance for ChatGPT, Claude, Gemini, Copilot, Antigravity, and local assistants.

Platform compatibility

Compatibility depends on whether the target platform supports instruction files, uploaded references, filesystem access, and code execution.

Designed for

• Client-facing reports
• Branded PDF documents
• Certificates
• One-page documents
• Educational handouts
• Worksheets
• Dashboards
• Internal reports
• Policy documents
• PDF inspection workflows
• PDF extraction workflows
• Safe PDF editing workflows
• Visual review before final delivery
• Editor-based automation in VSCode, Claude Code, Copilot, and Antigravity

Core workflow

PDF Pro follows a production-style workflow:

1. Identify the PDF task.
2. Inspect or render the input where needed.
3. Choose the safest authoring or editing method.
4. Perform the smallest safe operation.
5. Inspect or render the output.
6. Validate the final file.
7. Deliver the PDF with a clear summary of what was done, what was verified, and any limitation.

The goal is to avoid treating PDF work as a one-step file operation. Every final PDF should be checked before delivery.

Author and licence

• Author: Tariq Said
• Licence: MIT
• UI/UX Design Inspiration: UI/UX Pro Max by NextLevelBuilder, MIT Licence

See LICENSE and NOTICE.md for the complete licence and attribution details.

Folder structure

pdf-pro/
├── SKILL.md
├── README.md
├── LICENSE
├── NOTICE.md
├── AGENTS.md
├── CLAUDE.md
├── GEMINI.md
├── CLI.md
├── pyproject.toml
├── agents/
│   └── openai.yaml
├── assets/
│   ├── design-tokens-starter.json
│   └── pdf.png
├── examples/
│   └── visual-spec.json
├── platforms/
│   ├── chatgpt.md
│   ├── claude.md
│   ├── gemini.md
│   └── cli-agents.md
├── references/
│   ├── pdf-workflows.md
│   ├── quality-checklist.md
│   ├── validation-protocol.md
│   ├── visual-design-system.md
│   ├── cli-integration.md
│   ├── platform-compatibility.md
│   └── compatibility-checklist.md
├── scripts/
│   ├── create_visual_pdf.py
│   ├── pdf_ops.py
│   ├── validate_pdf.py
│   └── validate_package.py
├── src/
│   └── pdf_pro/
│       ├── __init__.py
│       ├── cli.py
│       ├── create_visual_pdf.py
│       ├── pdf_ops.py
│       └── validate_pdf.py
└── tests/
    └── test_cli_smoke.py

Main entry points

ChatGPT Skills

SKILL.md is the primary ChatGPT Skill instruction file.

It defines:

• When the skill should be used.
• How PDF tasks should be classified.
• Which workflows should be followed.
• When to inspect, render, extract, edit, validate, redact, or visually review a PDF.
• How final outputs should be delivered.

Claude Skills and Claude Code

Use CLAUDE.md and platforms/claude.md.

Gemini Gems

Use GEMINI.md and platforms/gemini.md.

CLI and coding agents

Use CLI.md, AGENTS.md, and references/cli-integration.md.

Install for CLI use

From the repository root:

python -m pip install -e .

Install optional PDF dependencies:

python -m pip install pypdf pymupdf reportlab

Or install with the package extras:

python -m pip install -e ".[pdf]"

CLI examples

Inspect a PDF:

pdf-pro inspect input.pdf

Extract text:

pdf-pro extract-text input.pdf --out extracted.txt

Render pages for visual review:

pdf-pro render input.pdf --out-dir renders/input --dpi 160

Merge PDFs:

pdf-pro merge output.pdf first.pdf second.pdf

Split pages:

pdf-pro split input.pdf --pages 1-3,7 --out selected-pages.pdf

Rotate pages:

pdf-pro rotate input.pdf --pages 1,3-5 --angle 90 --out rotated.pdf

Crop pages:

pdf-pro crop input.pdf --box 36,36,559,806 --out cropped.pdf

Add a watermark:

pdf-pro watermark input.pdf --text "DRAFT" --out watermarked.pdf

Redact matched text:

pdf-pro redact input.pdf --text "Confidential" --out redacted.pdf

Validate a finished PDF:

pdf-pro validate output.pdf --expected-pages 5 --render --out-dir validation/output

Create a visual PDF from JSON:

pdf-pro visual examples/visual-spec.json --out output.pdf

Script examples

The same operations are available without installing the package:

python scripts/pdf_ops.py inspect input.pdf
python scripts/pdf_ops.py render input.pdf --out-dir renders/input --dpi 160
python scripts/pdf_ops.py merge output.pdf first.pdf second.pdf
python scripts/validate_pdf.py output.pdf --render --out-dir validation/output
python scripts/create_visual_pdf.py examples/visual-spec.json --out output.pdf

Validation standard

Before a final PDF is delivered, verify at minimum:

1. The output file exists and opens.
2. The page count is correct.
3. Page sizes and rotations are intentional.
4. Text extraction is checked when text layer accuracy matters.
5. Visual renders are checked for clipping, overlap, missing glyphs, weak contrast, and layout errors.
6. Any limitation is stated clearly, especially scanned pages, OCR uncertainty, missing fonts, password protection, or poor source-file quality.

See references/validation-protocol.md for the full protocol.

Quality checklist

Before delivering a visual PDF, check:

• No clipped text.
• No overlapping elements.
• No missing glyphs.
• No broken images.
• No accidental page rotation.
• Margins are consistent.
• Headings are clear.
• Tables are readable.
• Charts are labelled.
• Colours have enough contrast.
• The PDF opens successfully.
• Rendered pages match the intended layout.

See references/quality-checklist.md for the complete checklist.

Redaction note

PDF Pro includes redaction workflow guidance and a text redaction command using a redaction-capable PDF engine where available. Do not treat a visual black box as safe redaction. True redaction must remove underlying text or pixels and must be verified through text extraction and rendering.

See references/validation-protocol.md for redaction validation rules.

Package validation

Validate the repository structure:

python scripts/validate_package.py .

Run smoke tests:

python -m pytest

Compile Python files:

python -m compileall scripts src

Attribution note

This package includes original PDF workflow instructions authored for PDF Pro.

Its visual design guidance is inspired by UI/UX Pro Max by NextLevelBuilder, which is identified as MIT-licensed. This package does not claim ownership of UI/UX Pro Max or its upstream project.

See NOTICE.md for full attribution details.

Licence

This project is released under the MIT Licence.

See LICENSE for details.

Maintainer

Created and maintained by Tariq Said.

PDF Pro is part of DXBMark’s pr