How to Integrate PPT Master with AI Tools Like Claude Code or VS Code Copilot

PPT Master runs as a skill inside AI coding assistants, allowing Claude Code, VS Code Copilot, and similar agents to generate PowerPoint files by executing a series of shell commands governed by the SKILL.md definition.

PPT Master is not a standalone application but a structured automation skill designed to run within AI-driven coding assistants. If you want to integrate PPT Master with AI tools like Claude Code or VS Code Copilot, you simply need to let the agent execute the Python scripts defined in the repository's skill definition. According to the hugohe3/ppt-master source code, the assistant acts as an orchestrator that reads skills/ppt-master/SKILL.md and runs the required conversion, design, and export pipelines.

How PPT Master Works Inside AI Coding Assistants

PPT Master follows a skill-based architecture where the AI assistant acts as the orchestrator. The repository contains no compiled binaries or graphical interface; instead, it exposes a declarative pipeline defined in skills/ppt-master/SKILL.md that any agent with file-system and shell access can execute.

The Core Skill Architecture

The integration relies on four components:

• skills/ppt-master/SKILL.md — Declares the entire pipeline, enforces strict serial execution, and lists the scripts that implement each step.
• skills/ppt-master/scripts/*.py and *.cjs — Python and Node utilities that handle conversion, project management, SVG generation, quality checks, and final export.
• templates/ — Layout JSON, chart SVGs, and design-spec references used during page construction.
• AI IDE integration layer — The assistant must support (a) reading/writing files in the repository, (b) running shell commands, and (c) maintaining conversation context. No additional SDK is required.

Because every step is a pure shell command, any AI coding assistant that can invoke a terminal can drive the complete workflow.

Execution Flow (Serial Pipeline)

The SKILL.md file defines a rigid execution order that the AI agent must follow:

1. Source → Markdown — Convert PDF, DOCX, or URLs using scripts/source_to_md/pdf_to_md.py, doc_to_md.py, or web_to_md.py.
2. Project Initialization — Run scripts/project_manager.py init to create a structured project folder.
3. Design Specification — The Strategist role produces design_spec.md and a machine-readable spec_lock.md.
4. Optional Image Generation — Execute scripts/image_gen.py (multi-provider) if the spec requests AI-generated images.
5. Executor Phase — The assistant reads spec_lock.md sequentially, builds each SVG page, and runs scripts/svg_quality_checker.py.
6. Post-processing & Export — Run scripts/total_md_split.py, scripts/finalize_svg.py, and finally scripts/svg_to_pptx.py to produce the editable .pptx file.

Setting Up the Integration

Integrating PPT Master with Claude Code, VS Code Copilot, Cursor, or similar agents requires only local repository access and Python 3.

Prerequisites

Ensure your AI assistant has:

• File system access — Scripts read/write only within the project folder.
• Shell command execution — All actions are standard python3 invocations; no custom RPC or plugins needed.
• Model-agnostic compatibility — The skill works with Claude, GPT, Gemini, or any model capable of following structured instructions.

Loading the Skill

1. Clone the repository to your local machine:

git clone https://github.com/hugohe3/ppt-master.git && cd ppt-master

2. Set the working directory to the skill location:

cd skills/ppt-master

3. Prompt your AI assistant with a natural language request, such as: "Create a 12-page PPT from /home/user/report.pdf." The agent will parse the request, consult SKILL.md, and execute the required commands automatically.

Claude Code has built-in support for skill execution and will reference the CLAUDE.md notice to read SKILL.md first. VS Code Copilot, when used in the terminal pane or Chat view, executes identical commands because it also possesses file-system access.

Step-by-Step Command Execution

Below are the precise shell commands your AI assistant will run when you integrate PPT Master with AI tools like Claude Code or VS Code Copilot.

One-Shot Generation Workflow

This example demonstrates the complete pipeline from source PDF to final PowerPoint:

# 1. Convert source PDF to markdown

python3 skills/ppt-master/scripts/source_to_md/pdf_to_md.py /path/to/source.pdf

# 2. Initialize a new PPT-Master project (16:9 format)

python3 skills/ppt-master/scripts/project_manager.py init my_presentation --format ppt169

# 3. Import the markdown into the project

python3 skills/ppt-master/scripts/project_manager.py import-sources \
    my_presentation sources/source.md --move

# 4. Strategist phase: The assistant presents Eight Confirmations and pauses for approval

# 5. (Optional) Generate AI images if the design spec requires them

# python3 skills/ppt-master/scripts/image_gen.py "futuristic skyline" \

#     --aspect_ratio 16:9 --image_size 2K -o my_presentation/images

# 6. Execute the pipeline: quality check, split, finalize, and export

python3 skills/ppt-master/scripts/svg_quality_checker.py my_presentation
python3 skills/ppt-master/scripts/total_md_split.py my_presentation
python3 skills/ppt-master/scripts/finalize_svg.py my_presentation
python3 skills/ppt-master/scripts/svg_to_pptx.py my_presentation -s final

# 7. Retrieve the final file

ls exports/

The AI assistant automatically pauses after step 3 (import) to present the Eight Confirmations — a checkpoint covering canvas size, audience profile, visual style, and content structure — before proceeding to generation.

The Eight Confirmations Checkpoint

The Strategist role in SKILL.md enforces a strict confirmation step. When the assistant reaches the design specification phase, it will:

1. Generate design_spec.md and spec_lock.md.
2. Present eight key decisions (canvas, audience, style, etc.) for your approval.
3. Wait for explicit confirmation before invoking image_gen.py or the Executor scripts.

This ensures cost transparency and design alignment before any AI image generation or final rendering occurs.

Optional: Configuring AI Image Generation

If your design specification includes AI-generated images, you must configure API keys for the multi-provider image generation script.

Environment Variables

Create a .env file in the repository root (never commit this file; use the provided template):

# Copy the template

cp .env.example .env

Edit .env with your provider keys:

OPENAI_API_KEY=sk-...
STABILITY_API_KEY=...

Then execute the image generation step:

python3 skills/ppt-master/scripts/image_gen.py "A futuristic city skyline, neon lighting" \
    --image_size 2K --aspect_ratio 16:9 -o my_presentation/images

The scripts/image_gen.py utility reads keys from the .env file automatically and supports multiple providers, allowing the AI assistant to enrich slides with custom visuals before final export.

Summary

• PPT Master is a skill, not an app — It runs inside Claude Code, VS Code Copilot, or any AI assistant with shell access, orchestrated by skills/ppt-master/SKILL.md.
• Pure shell command execution — All functionality resides in Python scripts like pdf_to_md.py, project_manager.py, and svg_to_pptx.py; no SDK installation required.
• Strict serial pipeline — The workflow moves from source conversion → project init → design spec → optional image gen → SVG execution → final export, with explicit checkpoints.
• Eight Confirmations gate — The AI assistant pauses for design approval before costly generation steps, ensuring alignment.
• Secure API configuration — Image generation uses .env files based on .env.example, keeping keys out of version control.

Frequently Asked Questions

Can I use PPT Master with AI assistants other than Claude Code or VS Code Copilot?

Yes. Any AI coding assistant that can (a) read and write files in a local repository and (b) execute shell commands can drive PPT Master. This includes Cursor, GitHub Copilot in the terminal, or local LLM agents with file-system access. The skill is model-agnostic; the assistant simply follows the command sequence defined in SKILL.md.

Why does the AI assistant need to pause at the Eight Confirmations step?

The Eight Confirmations checkpoint, defined in the Strategist role within SKILL.md, ensures that the visual style, audience profile, and content structure are approved before the system generates SVGs or calls paid image-generation APIs. This prevents unnecessary costs and ensures the final PowerPoint matches your expectations.

Do I need to install Python dependencies to use PPT Master with my AI assistant?

The source scripts in skills/ppt-master/scripts/ rely on standard Python 3 and common libraries. Your AI assistant will typically check for dependencies and suggest installation commands (such as pip install -r requirements.txt if one exists) before executing pdf_to_md.py or svg_to_pptx.py. The repository is designed to fail gracefully with clear error messages if dependencies are missing.

Is my data secure when using PPT Master with cloud-based AI tools?

Yes. All processing happens locally on your machine; the AI assistant only orchestrates the execution of local scripts. Source files, markdown conversions, and generated SVGs remain in your local skills/ppt-master/ directory. The only external calls occur if you explicitly enable scripts/image_gen.py, which sends prompts to your chosen image provider (OpenAI, Stability, etc.) using keys from your local .env file.

{
  "mcpServers": {
    "instagit": {
      "command": "npx",
      "args": ["-y", "instagit@latest"]
    }
  }
}