carnot-tech/consulting-pptx-skill

AIにまじなPPTXを作らせるClaude Codeスキル — スライド規約+62型スライド型カタログ(SlideSpec 36型+自由記述27パーツ)+生成パイプライン+機械チェック

What it solves

This project provides a structured framework for AI agents (specifically Claude Code) to generate high-quality, professional "boardroom-quality" presentation decks. It solves the problem of AI-generated slides often lacking professional design rigor, logical consistency, and adherence to strict corporate formatting rules by providing a comprehensive rulebook and a library of standardized components.

How it works

The system uses a multi-step pipeline to ensure quality:

  1. Rule Adherence: The AI is required to read a detailed rulebook (slide-rules.md) containing approximately 110 design and logic constraints before starting.
  2. Component-Based Assembly: Instead of free-form generation, the AI selects from a library of 62 HTML-based slide archetypes (divided into basic and additional parts) to build the deck. These are assembled into a single HTML file via a Python script.
  3. Automated Validation: A Python script (check_deck.py) mechanically checks the output for rule violations and placeholder remnants, while a Node.js script (check_layout.mjs) checks for rendering issues like footer overlaps or content overflow.
  4. Human-like Review: A separate AI agent, unaware of the creation process, is prompted to review the final deck for linguistic and logical contradictions.
  5. Export: The final HTML (16:9) is printed to PDF via Chrome.

Who it’s for

Consultants, business professionals, and AI developers who want to automate the creation of professional-grade presentation materials while maintaining strict control over design and logical quality.

Highlights

  • Extensive Rulebook: A living document of ~110 practical slide design rules derived from real-world professional reviews.
  • 62-Part HTML Library: A wide array of pre-defined slide layouts (e.g., executive summaries, Gantt charts, issue trees, matrices) to speed up the drafting process.
  • Mechanical Quality Control: Automated scripts to detect rule violations and layout errors before the final export.
  • Customizable: Users can append their own company-specific rules to the rulebook and modify CSS variables for branding colors and fonts.

Related

  • Project
  • Project
  • Project
  • Project