Skill: slide-analyze

Analyze a PPTX template file to extract fonts, colors, layouts, slide masters, and embedded images. Produces a structured context.json, a guideline.md with design rules, and sample_code.js demonstrating the template style.

Workflow

Step 1 — Accept Input

Ask the user for the PPTX file path if not provided. Confirm the file exists.

Step 2 — Derive Template Name

Derive a filesystem-safe template name from the filename with a timestamp suffix:

• Take the filename without extension (e.g., my-corp-template.pptx → my-corp-template)
• Replace spaces and special characters with hyphens
• Convert to lowercase
• Append a timestamp in YYYYMMDD-HHmmss format (use the current date/time)
• Example: my-corp-template.pptx → my-corp-template-20260407-143022

The timestamp ensures each analyze run produces a unique folder so re-analyzed versions never overwrite each other.

Step 3 — Create Workspace

Create the template workspace directory:

slide-workspace/templates/{template-name}/

Copy the original PPTX file to:

slide-workspace/templates/{template-name}/original.pptx

Step 4 — Check Dependencies

Check that python-pptx is available:

python -c "import pptx" 2>/dev/null || python3 -c "import pptx" 2>/dev/null

If the check fails, run the appropriate setup script:

• Unix/macOS: bash shared/scripts/setup_deps.sh
• Windows: powershell -ExecutionPolicy Bypass -File shared/scripts/setup_deps.ps1

Step 5 — Extract Template Context

Run the extraction script:

python shared/scripts/extract_template.py \
  "slide-workspace/templates/{template-name}/original.pptx" \
  "slide-workspace/templates/{template-name}/"

Or on Windows:

python shared/scripts/extract_template.py "slide-workspace/templates/{template-name}/original.pptx" "slide-workspace/templates/{template-name}/"

This produces:

• slide-workspace/templates/{template-name}/context.json — full structured context
• slide-workspace/templates/{template-name}/images/ — extracted image files

If exit code is non-zero, report the error and stop.

Step 6 — Read Extracted Context

Read context.json and note:

• presentation.slide_width_inches and slide_height_inches
• fonts_summary.families_used — font names used
• theme.color_scheme — theme colors
• slide_masters — number of masters and their names
• slide_layouts — number of layouts and their names
• images_manifest — list of extracted images

List the files in images/ directory to confirm images were extracted.

Step 7 — Read Prompt Rules

Read the analysis prompt from shared/prompts/analyze-rules.md.

Read the PptxGenJS API reference from shared/docs/pptxgenjs-api.md.

Step 8 — Generate Guideline and Sample Code

Using the extracted context.json content, images metadata, and the prompt rules from analyze-rules.md as your system instructions, generate two outputs:

guideline.md — A markdown document covering:

• Typography (exact font families, sizes, weights from context)
• Color palette (exact hex codes from theme and shape fills)
• Layout patterns (slide types detected, with descriptions)
• Spacing and alignment conventions
• Imagery style guidelines
• Brand/style rules

sample_code.js — A complete, runnable PPTXGenJS script that:

• Recreates the template structure
• Uses exact font names, colors, and positions from context
• References images via images['filename']
• Includes one example slide per layout type
• Follows all rules in analyze-rules.md

The JSON response format from analyze-rules.md is { "guideline": "...", "pptxgenjs_code": "..." }.

Step 9 — Save Outputs

Write the generated content to:

• slide-workspace/templates/{template-name}/guideline.md
• slide-workspace/templates/{template-name}/sample_code.js

Step 10 — Report Summary

Report to the user:

• Template name and workspace path
• Fonts found: list from fonts_summary.families_used
• Theme colors: list from theme.color_scheme
• Masters: count and names
• Layouts: count and names
• Images extracted: count from images_manifest
• Files created: guideline.md, sample_code.js, context.json, images/

Output Structure

slide-workspace/
  templates/
    {template-name}/
      original.pptx        ← Copy of input file
      context.json         ← Structured extraction (fonts, colors, layouts, masters)
      guideline.md         ← Design guidelines in Markdown
      sample_code.js       ← Sample PPTXGenJS code demonstrating the template
      images/              ← Extracted image files
        master0_Logo.png
        layout0_bg.jpg
        ...

Error Handling

• File not found: Ask user to verify the path
• Not a .pptx file: Inform user only .pptx files are supported
• extract_template.py exits with code 3: PPTX parse error — file may be corrupted
• No images directory: Template has no embedded images (normal for simple templates)
• LLM output not valid JSON: Retry or ask user to try again

Notes

• The images variable in generated code is pre-injected by the runner; never re-declare it
• effective_background on each slide is the definitive background — use it, not the raw background field
• Multiple slide masters are common in corporate templates — each gets its own MASTER_N name