Ppt Template Adaptation

PPT Template Adaptation

Use this skill when the task is to rewrite an existing PPT template into a new deck while preserving the template's visual logic. It consumes spec.json from pptx-template-parsing; do not start from a failed generated deck or manually inferred shape indices.

This skill is intentionally self-contained. Do not read old external handbooks from previous projects.

What this skill enforces

• Parse first, generate second. If a slot, card count, or style is missing, fix parsing before adding

generation fallback logic.

• Work page by page. Incremental runs keep original slide indices stable until the final full build.
• Keep one global outline as the source of truth for contents pages, section dividers, and section titles.
• Fill text before icons, logo, and images.
• Preserve parsed template style: geometry, alignment, theme colors, paragraph/run structure, and number

badges unless the user asks for redesign.

• Replace icons and images only when semantics require it and a valid slot exists.
• Validate each page before moving on.

Inputs

• Original template .pptx.
• Parsed spec.json from pptx-template-parsing.
• A content_plan.json with spec, output, asset_dir, optional brand, keep_slides, optional

outline, and pages[].

Required workflow

1. If no current spec.json exists, run pptx-template-parsing first.
2. Read the parser summary.txt; choose template slides whose roles and slot counts match the needed

story. Avoid chart/table slides unless chart/table data will be supplied.

1. Scaffold a plan:

python3 skills/ppt-template-adaptation/scripts/scaffold_content_plan.py \
  --spec workspace/specs//spec.json \
  --out workspace/plans/.content_plan.json \
  --output-pptx workspace/out/.pptx

1. Edit keep_slides, outline, and pages[]. Map each page to existing slots only: titles to

title, free labels to labels[], body boxes to body[], repeated structures to cards[], semantic glyphs to cards[].icon, and photos/screenshots to images[].

1. Validate global text coherence:

python3 skills/ppt-template-adaptation/scripts/validate_content_plan.py \
  --plan workspace/plans/.content_plan.json

1. Fill one or a few original slide indices at a time:

python3 skills/ppt-template-adaptation/scripts/apply_content_plan.py \
  --plan workspace/plans/.content_plan.json \
  --pages 0,1 \
  --render

1. Review the rendered page when visual review is available. If the user will do visual review, provide

the output files and continue with structural checks; do not use image tools when the user forbids it.

1. Repeat the page loop until selected pages pass. Then run a full build without --pages so unused

slides are dropped according to keep_slides.

Debugging order

1. Inspect parser output first: summary.txt, spec.json fill_plan, card count, style.paragraphs,

font availability, and visual obstacles.

1. If parsing is wrong, fix pptx-template-parsing, rerun parsing, and rebuild the plan or affected

pages.

1. If parsing is right but generation breaks style or layout, fix ppt-template-adaptation application

logic.

1. Use --no-clear-unfilled, --skip-global-text-check, or --skip-image-resolve only for deliberate

diagnostics. Do not use them to declare a page complete.

Reference read order

Read only the references needed for the current task:

• references/workflow.md — complete page-by-page workflow and acceptance criteria.
• references/text-coherence.md — contents/section outline consistency and text fitting.
• references/icon-fill.md — semantic icon replacement rules.
• references/image-fill.md — conservative image replacement and Tavily usage.

Example scripts

• scripts/scaffold_content_plan.py
• scripts/validate_content_plan.py
• scripts/apply_content_plan.py
• scripts/render_pages.py
• scripts/tavily_image_service.py

Usage rule

Do not jump to the next page until the current page has passed validation. For visual validation, either review the render when allowed or hand the render artifact to the user for manual review.

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: CxyZyr
• Source: CxyZyr/PPTX-Template-Skills
• License: MIT

Install and usage instructions live in the source repository linked above.