PPT Design Skill

Setup (first run)

The pptx tooling this skill calls — scripts/office/*, add_slide.py, clean.py, thumbnail.py — is not vendored in this repo (its license forbids redistribution). Fetch it from the upstream repository once, before first use:

python3 scripts/setup_upstream.py           # no-op when already installed
python3 scripts/setup_upstream.py --check   # verify installed

The script pins an upstream commit for reproducibility (--ref to override) and also fetches upstream/SKILL.md (design-ideas reference) plus the upstream LICENSE.txt that governs the fetched files. Python deps for the fetched tooling: pip install defusedxml lxml Pillow "markitdown[pptx]".

Quick Reference

Reading Content

# Text extraction
python -m markitdown presentation.pptx

# Visual overview
python scripts/thumbnail.py presentation.pptx

# Raw XML
python3 -c "import sys,zipfile; zipfile.ZipFile(sys.argv[1]).extractall('unpacked')" presentation.pptx

Editing Workflow

Read [editing.md](editing.md) for full details.

1. Analyze template with thumbnail.py
2. Extract → structural changes (add_slide.py / <p:sldIdLst> / clean.py) → edit content → zip → validate.py --original

Creating from Scratch

Creation workflow（三段式，顺序固定）：

Step 1 · 规划（必做，禁止跳过）

Read [references/design-playbook.md](references/design-playbook.md) — 叙事弧、页面规划表、明暗节奏、版式决策树、自适应规则、中文排版、密度控制。

选主题：python3 scripts/theme.py list（内置 9 套 + 项目 ./ppt-themes/ + 用户级）。发现项目/用户级主题 → 追问用户是否优先使用；用户自带品牌色 → references/custom-themes.md 的 extract 流程。

产出页面规划表（页码 → 版式 → 明暗 → 内容形状 → 视觉元素），落成 outline.json 并过规划门禁后再写代码：

python3 scripts/qa/outline_check.py outline.json   # exit 0=PASS；主题名/版式 id/多样性/连续重复全查

Schema 与注册版式 id 见 references/outline-schema.md。门禁不过先改 outline.json——规划阶段改一行，代码阶段返工一页。

Step 2 · 生成

Read [pptxgenjs.md](pptxgenjs.md) for API details and corruption gotchas.

• 主题与配色：design-system.md（9 主题，只用安全字体）
• 版式代码：杂志风 templates/layouts/magazine-layouts.md / 瑞士风 templates/layouts/swiss-layouts.md / 分析模型 templates/layouts/analysis-models.md（SWOT/PEST/画布/双钻/竞争定位）
• 组件配方：templates/components.md
• 瑞士风硬约束：references/swiss-layout-lock.md
• ≥15 页走分模块管线：python3 scripts/scaffold_deck.py outline.json --out-dir deck --theme <name> 生成 slide-NN.mjs + compile.mjs，契约与子代理并行填充见 references/modular-decks.md

Step 3 · QA

确定性门禁（先跑，全 PASS 再进视觉检查；exit 0=PASS / 1=有违规，可入 CI）：

python3 scripts/qa/cjk_overflow_check.py output.pptx   # 文字溢出/越界（CJK 1em·半角 0.5em 估宽）
python3 scripts/qa/contrast_check.py output.pptx       # WCAG 对比度（4.5:1 正文 / 3:1 大字）
python3 scripts/qa/token_check.py slides/*.mjs --theme ink-classic  # 主题 token 门禁（硬编码色/字体即 fail）
python3 scripts/qa/font_check.py output.pptx                        # Office 安全字体 + 本地可用性（fc-list）
pdftoppm -png -r 150 output.pdf slide \
  && python3 scripts/qa/edge_check.py slide-*.png      # 渲染后边缘裁切/空页（纯 stdlib，PNG）

五个脚本纯 stdlib、独立解析（不依赖上游拉取件）。再按 references/checklist.md P0-P3 检查（见下方 QA 节）。

Design Ideas

Full design-ideas checklist — inspiration palettes, typography rules, spacing, and the common-mistakes list — lives in the upstream reference fetched at setup: read `upstream/SKILL.md` (§ Design Ideas) after running scripts/setup_upstream.py. The sections below are this repo's own design system layered on top.

Core principles (short form):

• One dominant color carries the deck; accents stay sharp and rare
• Dark backgrounds for title + closing slides, light for content — or dark throughout for a premium feel
• Pick ONE visual motif and repeat it on every slide
• Every slide gets at least one visual element (image, chart, icon, shape) — no text-only slides
• Vary layouts across the deck; never repeat one layout more than 3 slides in a row

Design System Quick Start

For pre-built themes and layout coordinates, see design-system.md. It provides:

• 9 built-in themes (5 magazine + 4 Swiss) with ready-to-use PptxGenJS color configs
• 32 layout templates — 完整代码见 templates/layouts/（杂志 10 + 瑞士 22）
• Office 安全字体 pairings for both magazine (serif) and Swiss (sans-serif) styles

规划先行：写代码前先读 references/design-playbook.md，产出页面规划表。版式坐标是基线不是牢房——按 playbook 第 4 节的自适应规则变形。

7-Question Clarification (optional best practice when user has only a vague idea):

1. Audience & scenario: Roadshow / internal share / tech launch / portfolio / academic?
2. Duration & page count: 10-12 / 15-20 / 25+ slides?
3. Existing materials: Documents / links / old PPT / data / images?
4. Theme choice: See Built-in themes below — magazine (elegant/humanistic) or Swiss (minimal/rational)?
5. Style preference: Serif titles with warm tones, or sans-serif with high-contrast accent colors?
6. Language: Chinese / English / bilingual?
7. Deliverables: PPTX only / also need cover images / also need web version?

风格画廊技巧（借鉴 humanize-ppt）：用户对主题犹豫时，不要盲选——用 2-4 套候选主题各渲染一张封面给用户挑，选定后再出全 deck。

Theme usage rhythm:

Cover       →  Main theme (consistent across deck)
Section     →  Same theme, variant switch (light/dark)
Data page   →  Same theme, light variant
Highlight   →  Same theme, dark variant
Closing     →  Same theme, dark variant or cover style

Rule: Never switch to a different theme mid-deck.

Built-in Design System Themes

9 套内置主题（5 magazine + 4 Swiss）的完整色值与适用场景见 design-system.md。主题名即 token：ink-classic / indigo-porcelain / forest-ink / kraft-paper / dune / swiss-blue / swiss-yellow / swiss-green / swiss-orange（项目/用户级主题用 python3 scripts/theme.py list 查看）。

Built-in Layout Templates

Coordinates in design-system.md; complete code in templates/layouts/ — 杂志 10 版式 + 瑞士 S01-S22 + 分析模型 5（SWOT/PEST/画布/双钻/竞争定位）。outline.json 引用这些注册 id（全表见 references/outline-schema.md）；版式多样性与连续重复上限由 outline_check.py 在规划阶段强制。

中文安全字体（内置主题已锁定）

只用 Office 安全字体——用户机器上没有的字体（Inter / Playfair / Noto 系列）会 fallback 到宋体。内置主题的中英文字体配对（magazine serif / Swiss sans）已锁定在 design-system.md，直接用主题 token，不要自造字体组合（font_check.py 会查）。

QA (Required)

Assume there are problems. Your job is to find them.

Your first render is almost never correct. Approach QA as a bug hunt, not a confirmation step. If you found zero issues on first inspection, you weren't looking hard enough.

Content QA

python -m markitdown output.pptx

Check for missing content, typos, wrong order.

When using templates, check for leftover placeholder text:

python -m markitdown output.pptx | grep -iE "xxxx|lorem|ipsum|this.*(page|slide).*layout"

If grep returns results, fix them before declaring success.

File QA

python scripts/office/validate.py output.pptx                      # built from scratch
python scripts/office/validate.py output.pptx --original src.pptx  # built from a template

If the deck came from a template, always pass `--original` — it baselines schema/slide checks against the template so the template's own defects don't read as yours.

Visual QA

⚠️ USE SUBAGENTS — even for 2-3 slides. You've been staring at the code and will see what you expect, not what's there. Subagents have fresh eyes.

Convert slides to images (see Converting to Images), then send the ready-to-use inspection prompt in references/checklist.md §「子代理看图提示词」to a subagent — it checks overlaps, overflow, collisions, gaps, contrast, and leftover placeholders per slide, and reports ALL issues including minor ones.

Pre-Flight Checklist

Before declaring success, run through references/checklist.md (P0-P3 levels). P0 must all pass before delivery: no placeholder text, no overflow/cut-off, no overlapping elements, theme consistent, minimum font sizes.

Verification Loop

1. Generate slides → Convert to images → Inspect
2. List issues found (if none found, look again more critically)
3. Fix issues — 按失败模式目录的修复顺序：内容 → 结构 → 版式 → 样式，不要从样式开始修
4. Re-verify affected slides — one fix often creates another problem
5. Repeat until a full pass reveals no new issues. 最多 3 轮；3 轮后仍有 P0 问题，向用户说明剩余问题，不要无限重试。

Do not declare success until you've completed at least one fix-and-verify cycle.

Design System Reference

Converting to Images

Convert presentations to individual slide images for visual inspection:

python scripts/office/soffice.py --headless --convert-to pdf output.pptx
pdftoppm -jpeg -r 150 output.pdf slide

This creates slide-01.jpg, slide-02.jpg, etc.

To re-render specific slides after fixes:

pdftoppm -jpeg -r 150 -f N -l N output.pdf slide-fixed

Dependencies

• First run: python3 scripts/setup_upstream.py — fetches the tooling runtime (see Setup)
• pip install "markitdown[pptx]" - text extraction
• pip install Pillow defusedxml lxml - thumbnail grids / validate
• npm install -g pptxgenjs - creating from scratch
• LibreOffice (soffice) - PDF conversion (auto-configured for sandboxed environments via scripts/office/soffice.py)
• Poppler (pdftoppm) - PDF to images