carnot-tech/jinba-consulting-pptx-skill

What it solves

This project provides a structured system for AI agents (specifically Claude Code) to generate professional, boardroom-quality presentation decks. It solves the problem of AI-generated slides often lacking logical rigor, consistent design, and adherence to professional consulting-style design rules, which usually requires extensive manual correction.

How it works

The system uses a "rules-first" approach rather than relying on simple templates. It combines several components:

1. Design Rulebook: A comprehensive text file (slide-rules.md) that defines strict layout and logic rules (e.g., "conclusions must be in the title").
2. HTML-based Workflow: Slides are first constructed as HTML (16:9) using a library of 62 pre-defined parts (basic and advanced layouts). This allows for faster iteration and easier mechanical checking.
3. Automated Validation: A Python script (check_deck.py) mechanically checks the HTML or PPTX for rule violations and placeholder remnants.
4. Review Cycle: A separate AI agent is prompted to review the content for logic and language contradictions using a specific review prompt.
5. Conversion: Once finalized in HTML, the deck is converted to an editable PowerPoint (.pptx) file using a script that maps HTML elements to PowerPoint shapes.

Who it’s for

• Professionals and consultants who need to produce high-quality decks quickly using AI.
• Teams wanting to standardize their AI-assisted presentation design across an organization.
• Users of Claude Code who want to extend its capabilities for document creation.

Highlights

• 62-Part HTML Library: A wide array of professional layouts including executive summaries, Gantt charts, and issue trees.
• Mechanical Rule Checker: Automates the detection of design rule violations, ensuring consistency without manual oversight.
• HTML-to-PPTX Pipeline: Converts HTML layouts into fully editable PowerPoint shapes rather than static images.
• Customizable Framework: Allows users to add organization-specific rules in a local directory to evolve the system's quality over time.