AGENT SKILL

ppt-implement

implement ppt(powerpoint) project with best practices, start's with "ppt" template. Trigger keywords include "web ppt", "网页ppt", "html ppt", "生成ppt", "制作ppt", "制作教案", "write a ppt about", or any request to create presentations on specific topics.

License 许可
MIT
最近更新
2026-11-10
Stars
★ 1
Forks
0

01 前言 Front Matter

name
ppt-implement
description
implement ppt(powerpoint) project with best practices, start's with "ppt" template. Trigger keywords include "web ppt", "网页ppt", "html ppt", "生成ppt", "制作ppt", "制作教案", "write a ppt about", or any request to create presentations on specific topics.

02 资源文件 500

03 SKILL.md 正文

⚠️ CRITICAL: DO NOT Explore Project Structure

.
├── docs
│   ├── product
│   └── project.json
└── frontend
    ├── 404.html
    ├── index.html
    ├── package.json
    ├── postcss.config.js
    ├── public
    │   └── assets
    │       └── images
    ├── scripts
    │   └── build-slides.js
    ├── src
    │   ├── js
    │   │   ├── ppt-controller.js # PPT controller, DO NOT modify
    │   │   └── route-handler.js # Route handler, DO NOT modify
    │   ├── main.js # Entry file, DO NOT modify
    │   ├── slides # PPT slides, stores slide-N.js files, no manual loading needed - framework auto-loads them
    │   └── styles
    │       └── main.css
    ├── tailwind.config.js
    ├── vite.config.js
    └── yarn.lock

The project framework is pre-configured and ready to use. No analysis required.

  • ❌ DO NOT use code-explorer subagent to analyze the project

  • ❌ DO NOT use search_file / search_content / list_dir to explore directory structure

  • ❌ DO NOT read package.json, vite.config, or other config files to "understand the project"

  • ❌ DO NOT spend time figuring out "how the project works"

  • ✅ START directly from Phase 1 (Material Collection)

  • ✅ FOLLOW the workflow strictly: Phase 1 → Phase 9

  • ✅ ONLY READ files explicitly specified in each phase

Reason: The project structure is standardized. All necessary paths and configurations are documented in the phases below. Any exploration is a waste of time.

Path Variables

Path variables used in this document:

  • CORE_DIR: Project core directory, i.e. ${CODEBUDDY_PLUGIN_ROOT}
  • WORKSPACE_DIR: Project root directory, i.e. ${CODEBUDDY_PROJECT_DIR}

General Principles

Key Commands

# Debugging
python3 "${$WORKSPACE_DIR}/.genie/scripts/python/fetch_monitor_errors.py"

Launch Configuration

.cloudstudio file defines startup - DO NOT modify. Always use unified process script.

Development Server Guidelines

  • DO NOT automatically start the development server (e.g., npm run dev, vite, etc.) after completing tasks
  • DO NOT output localhost addresses to the user (e.g., http://localhost:5173/)

PPT Creation Workflow

Process Overview

Complete Workflow:

  1. Phase 1: Material Collection (Data Gathering, Chapter Planning) - This Skill
  2. Phase 2: Style Definition (Main Visual System) - This Skill
  3. Phase 3: Generate Page Images - This Skill
  4. Phase 4: Generate PPT Outline - This Skill
  5. Phase 5: Environment Setup - This Skill
  6. Phase 6: Template Planning (Select templates for each page) - This Skill
  7. Phase 7: Parallel PPT Generation (Batch invoke sub-agents) - This Skill orchestrates, Sub-agents execute
  8. Phase 8: Take PPT Screenshots - This Skill
  9. Phase 9: Output PPT Usage Instructions - This Skill
  10. Phase 10: Handling User Modifications - Only triggered when user requests changes

⚠️ Key Principles:

  • Phase 1-9: First-time generation flow, execute sequentially then STOP
  • Phase 10: Only triggered when user requests modifications after initial generation
  • Phase 7: Skill orchestrates parallel sub-agent calls, each sub-agent generates ONE page
  • Sub-agent only handles single page HTML generation (content filling)
  • 🚫 NO USER INTERACTION: Execute ALL phases autonomously without asking user for confirmation, clarification, or approval. Make reasonable decisions based on context.

Phase 1: Material Collection

1.1 Data Gathering

Goal: Search and collect materials for PPT theme, output to ${WORKSPACE_DIR}/docs/product/material.md.

Step 1: Generate Search Queries (8-12 queries, parallel execution)

Layer Query Pattern Example
Foundation [Theme] definition/overview "AI ethics overview"
Context [Theme] history/background "AI ethics development"
Evidence [Theme] statistics/research 2024 "AI ethics research 2024"
Application [Theme] case study/examples "AI ethics real cases"
Trends [Theme] trends/future "AI ethics future challenges"

Tips: Add year (2024/2025) for time-sensitive topics; mix CN/EN queries when needed.

Step 2: Extract & Organize into 6 Dimensions

  1. Overview - Definition, significance
  2. Background - History, milestones, current status
  3. Key Info - Facts, data, expert views
  4. Evidence - Cases, research, visual refs
  5. Analysis - Multiple perspectives, comparisons
  6. Outlook - Trends, recommendations

Step 3: Output Format (${WORKSPACE_DIR}/docs/product/material.md)

# Material: [Theme]

## 1. Overview
- [Point]

## 2. Background
- [Point]

## 3. Key Info
- [Point]

## 4. Evidence
- Case: [Name] - [Desc] (Has Images: Yes/No)

## 5. Analysis
- [Viewpoint]

## 6. Outlook
- [Trend]

## Summary
- High-authority: [N], Gaps: [List or None]

Quality Rules: Prioritize academic/official sources; exclude unverified/outdated (>5yr) content; ensure ≥3 high-authority sources and ≥5 different domains.

1.2 Chapter Planning (Pyramid Principle)

Page Count: Default 14-20 pages when unspecified. Single chapter: 1-5 pages.

Core Principle: Apply Pyramid Principle — structure content with clear hierarchy, each level supporting the one above.

Step 1: Identify PPT Type → Select Pyramid Structure

PPT Type Pyramid Structure Chapter Flow
Report/Summary Classic (Conclusion First) Result → Evidence → Details → Next Steps
Product/Sales Problem-Solution Pain Point → Solution → Value → Proof → CTA
Training/Education Progressive (Shallow→Deep) Basics → Core → Advanced → Practice
Pitch/Fundraising SCQA Story Situation → Complication → Question → Answer
Event/Promo Emotional Hook → Resonance → Climax → Action

Step 2: Build Chapter Hierarchy

[PPT Core Message]                    ← Answer: What's the ONE takeaway?
    ├── [Chapter 1: Sub-conclusion]   ← Support core message
    │       ├── Page: Evidence/Data
    │       └── Page: Example/Case
    ├── [Chapter 2: Sub-conclusion]
    │       ├── Page: Evidence/Data
    │       └── Page: Example/Case
    └── [Chapter 3: Sub-conclusion]
            └── ...

Step 3: Validate Structure (MECE Check)

  • Mutually Exclusive: No overlap between chapters
  • Collectively Exhaustive: Chapters fully support the core message
  • Logical Order: Time sequence / Importance / Structure

Based on collected materials (${WORKSPACE_DIR}/docs/product/material.md), output chapter structure to ${WORKSPACE_DIR}/docs/product/chapters.md.

Chapter Planning Format:

## Page N: Page Name
- **Page Type**: Cover/TOC/Transition/Content/Ending
- **Page Title**: XXX (recommended within 15 characters, can add subtitle: XXX)
- **Selected Template**: (Leave blank, to be filled during Phase 6 Template Planning)

- **Content Structure**: (Choose appropriate structure based on page purpose)

  > **IMPORTANT - Content Richness Requirements**:
  > - Each main point MUST have 2-3 sentences of detailed explanation
  > - Each page MUST contain at least 3-5 specific data points (numbers, percentages, years)
  > - Each page body content should be 100-200 words minimum
  > - Include supporting details: comparisons, context
  
  **Content Structure by Page Type** (select appropriate structure):

  | Page Type | Core Elements |
  |-----------|---------------|
  | **Argument** | Core Argument + 3-4 Evidence (Data/Case/Quote) + Key Insight |
  | **Data** | Data Headline + 4-6 Metrics (Number\|Label\|Context) + Data Story |
  | **Process** | Goal + 4-6 Steps (Action→Output→Duration) + Success Metrics |
  | **Case Study** | Header (Company/Timeline/Scale) + Challenge + Solution + Results (Before→After) + Lessons |
  | **Comparison** | Framework + Options A/B/C (Dimensions + Best For + Limits) + Recommendation |
  | **Concept** | Definition (Simple/Technical) + 3-5 Components + Applications + Misconceptions |
  | **Problem/Solution** | Problem (Scope/Urgency) + Impacts + Root Causes + Solutions (Approach/Benefit/Risk) |
  | **Trend** | Baseline + 3-5 Trends (Desc/Evidence/Drivers/Timeline) + Projections + Actions |
  | **Summary** | Executive Summary + 4-6 Takeaways + Key Numbers + Call to Action |

  > **Universal Requirements**: Each page needs 100-200 words body content, 3-5 specific data points.

- **Content Density**: Light (≤3 main points) / Medium (4-6 points) / Heavy (7+ points or complex data)
- **Narrative Role**: [How this page advances the overall story] (e.g., "Establishes credibility with data", "Introduces key challenge", "Provides solution framework")

- **Image Requirements**:
  **Image Requirements Planning Principles**:
  - **Default to No Images**: Unless images are truly necessary, use text layout. **Control the overall proportion of pages with images to no more than 40% of total pages** (e.g., 10-page PPT with images on no more than 4 pages).
  - **Strict Evaluation Criteria** (Must meet at least one):
    1. **Complexity**: Information has inherent complexity (spatial/temporal/relational) that text cannot efficiently express
    2. **Visual Priority**: Visual form is the primary information carrier (e.g., network relationships, geographic distribution)
    3. **Efficiency Gain**: Images can significantly reduce cognitive load (e.g., trend comparisons, structural hierarchies)
    4. **Scenario Necessity**: Specific scenarios inherently require images (e.g., character introductions, visual effects demonstrations)
  
  - **Decision Framework** (Apply to any scenario):
    - **ASK**: Can text + layout clearly convey this? → If YES, do not use images
    - **ASK**: Is information complexity high (≥4 elements/dimensions/steps)? → If NO, prioritize text
    - **ASK**: Does visual form have unique advantages? → If NO, use text/cards
  
  - **Reference Examples** (Not exhaustive, apply evaluation criteria flexibly):
    - ✅ **Should Use Images**:
      - Complex spatial: Multi-location maps, architectural layouts
      - Multi-step processes: ≥4-node timelines, multi-branch flowcharts
      - Network relationships: ≥5-node ecosystem diagrams, dependency graphs
      - Data trends: Multi-dimensional curves, complex comparative charts
      - Character/roles: Avatars, portraits (when identity is key information)
    - ❌ **Should Not Use Images**:
      - Pure viewpoint exposition: Use lists, card layouts
      - Simple enumeration (≤3 items): Text layout is clearer
      - Concept definitions: Text + color/bold more direct
      - Simple data (≤3 data points): Data cards suffice
      - Short processes (≤3 steps): Text arrows more concise
      - Cover/TOC/ending: Text primary, images only decorative
  - **Flexibility Note**: Above examples are references, not exhaustive lists. Apply evaluation criteria based on actual content to decide if images are needed.
  
- **Page Weight**: Core page/Secondary page/Transition page (clarify page importance; core pages can allocate more design effort, e.g., key/difficult content pages in teaching set as core pages)
- **Content Page Selection Rationale**: (Required only for content pages, case pages, practical pages) Briefly explain page value (e.g., connecting previous chapter's classical Chinese knowledge points, breaking down lesson's key/difficult points, paving the way for subsequent interactive exercises; or concretizing modern application of Chinese elements through this case to enhance persuasiveness)
- **Notes**: (Optional, supplement special requirements) e.g., Page needs animation effects (text appears line by line); Reserve QR code position (linking to extended materials); Educational scenarios need to add mini-question (e.g., "What do you think is the beauty of the character 'green'?")

Page Type Descriptions:

  • Cover Page: PPT opening, contains main title, subtitle, presenter information
  • TOC Page: Chapter navigation, lists all chapter names
  • Transition Page: Chapter transitions, indicates current chapter name
  • Content Page: Main content, contains text, images, charts, etc.
  • Ending Page: PPT closing, contains thank you message, contact information, etc.

Phase 2: Style Definition

2.1 Determine PPT Style

Based on the user's input prompt, determine the PPT style and record the confirmed result in ${SlideStyle}. Currently supported styles include:

  • Minimalist
  • Chinese
  • Educational
  • Business
  • Geometric
  • Literary
  • Black-Gold
  • Cartoon
  • Tech
  • Flat

2.2 Main Visual System

Font Planning

Font Selection Principles:

  • Entire PPT should not exceed 2-3 font types (title + body + decorative optional)
  • For mixed Chinese-English text, choose fonts that support both languages
  • Prioritize Web-safe fonts or Google Fonts (for cross-platform display)

Font Style Matching: Choose fonts that match PPT tone - sans-serif (Noto Sans SC, Roboto, Inter, Montserrat) for modern/tech/business themes; serif (Noto Serif SC, Playfair Display) for traditional/premium/cultural themes; creative/handwritten fonts (Ma Shan Zheng, ZCOOL KuaiLe) for artistic/brand themes. Use Google Fonts for cross-platform compatibility.

Common Google Fonts Quick Reference:

Font Name Type Features CDN Import
Ma Shan Zhen Chinese Creative Handwritten style, highly personalized @import url('https://fonts.googleapis.com/css2?family=Ma+Shan+Zheng&display=swap');
Noto Sans SC Chinese Sans-serif Strong versatility, supports Chinese & English @import url('https://fonts.googleapis.com/css2?family=Noto+Sans+SC:wght@300;400;700&display=swap');
Noto Serif SC Chinese Serif Elegant traditional, suitable for cultural themes @import url('https://fonts.googleapis.com/css2?family=Noto+Serif+SC:wght@400;700&display=swap');
ZCOOL KuaiLe Chinese Creative Handwritten style, highly personalized @import url('https://fonts.googleapis.com/css2?family=ZCOOL+KuaiLe&display=swap');
Roboto English Sans-serif Modern simplicity, Google default font @import url('https://fonts.googleapis.com/css2?family=Roboto:wght@300;400;700&display=swap');
Poppins English Sans-serif Rounded modern, suitable for titles @import url('https://fonts.googleapis.com/css2?family=Poppins:wght@400;600;700&display=swap');
Montserrat English Sans-serif Strong geometric feel, tech-style @import url('https://fonts.googleapis.com/css2?family=Montserrat:wght@400;600;700&display=swap');
Inter English Sans-serif High readability, suitable for body text @import url('https://fonts.googleapis.com/css2?family=Inter:wght@300;400;600&display=swap');
Playfair Display English Serif Premium elegance, suitable for high-end brands @import url('https://fonts.googleapis.com/css2?family=Playfair+Display:wght@400;700&display=swap');

Local Font Alternatives (no loading needed, system built-in):

  • Cross-platform Safe Fonts: Arial, Helvetica, Times New Roman, Georgia

Phase 3: Generate Page Images

3.1 Scan Image Requirements

Extract all pages with image requirements marked from ${WORKSPACE_DIR}/docs/product/chapters.md:

Extraction Logic:

# Pseudocode example
image_tasks = []

for page in chapters:
    if page.image_requirements is not None and page.image_requirements != "":
        image_task = {
            "pageIndex": page.page_number,
            "pageTitle": page.page_title,
            "visualType": identify_image_type(page.image_requirements),  # Data chart/timeline/comparison chart, etc.
            "imageRequirements": page.image_requirements,  # Raw requirements, prompt generated later if needed
            "savePath": f"${WORKSPACE_DIR}/frontend/public/assets/images/page-{page.page_number}-{index}.png"
        }
        image_tasks.append(image_task)

Image Type Identification:

  • Data Visualization: Curve charts, bar charts, pie charts, ring charts, radar charts, data cards
  • Process Display: Timelines, step diagrams, flowcharts, route maps
  • Structure Display: Four-grid, card sets, grid layout
  • Comparison Analysis: Left-right comparison, A/B testing, pros/cons comparison
  • Concept Illustration: Icon lists, illustrated text, scene schematic diagrams

3.2 Generate Detailed imagePrompt

Prompt Generation Principles:

  1. Specific Description: Include chart type, data content, annotation information
  2. Style Unity: Consistent with overall PPT style (business/tech/Chinese style, etc.)
  3. Clear Color Scheme: Specify primary color system (e.g., "blue-purple gradient", "business blue tones")
  4. Generatability: Ensure AI can understand and generate corresponding images

Execution Flow:

# Generate prompts for all image tasks
for image_task in image_tasks:
    
    # Generate detailed prompt based on image requirements
    image_task["imagePrompt"] = generate_detailed_prompt(
        page_title=image_task["pageTitle"],
        visual_type=image_task["visualType"],
        requirements=image_task["imageRequirements"]
    )

3.3 Call ImageGen Tool to Generate Images

Calling Method: Use the ImageGen tool to generate images one by one

ImageGen Tool Parameters:

{
  "prompt": "string",          // [Required] Image description text
  "size": "string",            // [Optional] Size, e.g., "1024x1024", "1024x1536"
  "n": "number",               // [Optional] Generation count, 1-10, default 1
  "quality": "string",         // [Optional] Quality: low, medium, high
  "style": "string",           // [Optional] Image style
  "background": "string",      // [Optional] Background: transparent, opaque
  "output_dir": "string"       // [Optional] Custom output directory
}

Execution Flow:

# Pseudocode example
for image_task in image_tasks:
    # Call ImageGen tool
    ImageGen({
        "prompt": image_task["imagePrompt"],
        "size": "1024x1024",
        "quality": "high",
        "output_dir": "${WORKSPACE_DIR}/frontend/public/assets/images"
    })
    
    # Verify file exists
    save_path = image_task["savePath"]
    if not exists(save_path):
        error and stop
    
    # Record image path to image task
    image_task["generatedPath"] = save_path

Key Rules:

  • ✅ Batch Calling: Generate images one by one using ImageGen tool
  • ✅ Unified Save Path: Set output_dir to ${WORKSPACE_DIR}/frontend/public/assets/images/
  • ✅ Validate Generation Results: Immediately check if file exists after each image generation

3.4 Generate Image Mapping Table

Output File: ${WORKSPACE_DIR}/frontend/public/assets/images/images-mapping.json

{
  "5": {
    "imagePath": "assets/images/page-5.png",
  },
  "7": {
    "imagePath": "assets/images/page-7.png",
  }
}

Key Notes:

  • Key is page number (string format)
  • imagePath is relative to ${WORKSPACE_DIR}/frontend/public/ path

Phase 4: Generate PPT Outline

Based on the collected information (${WORKSPACE_DIR}/docs/product/material.md, ${WORKSPACE_DIR}/docs/product/chapters.md), generate the PPT Outline (file: ${WORKSPACE_DIR}/docs/product/features.md). CRITICAL: You MUST follow the template below to generate the PPT Outline. Do NOT modify the template structure.

Format:

# PPT Outline

## Overview
{{PPT Overview}}

## Outline Content
{{Outline Content - Extract directly from chapters.md, but MUST remove the following template-related fields: Selected Template, Page Weight, Content Density, Narrative Role, Image Requirements, and Notes. Keep only Page Type, Page Title, Page Subtitle, and Content Structure for each page}}

## Design Style
{{Design Style}}

Phase 5: Environment Setup

5.1 Create Global Configuration File

Generate ${WORKSPACE_DIR}/docs/page-global-config.json, containing style configurations shared by all pages:

{
  "slideStyle": "${SlideStyle}",
  "colorScheme": "Modern Tech Style",
  "primaryColor": "#4285F4",
  "accentColor1": "#34A853",
  "accentColor2": "#FBBC04",
  "neutralColor": "#5F6368",
  "background": {
    "type": "color",
    "value": "#FFFFFF"
  },
  "fontTitle": "'Montserrat', 'Noto Sans SC', sans-serif",
  "fontBody": "'Inter', 'Noto Sans SC', sans-serif"
}

Background Configuration Options:

Type Example Description
color {"type": "color", "value": "#FFFFFF"} Solid color background
image {"type": "image", "value": "assets/images/bg.png"} Background image (relative to ${WORKSPACE_DIR}/frontend/public/)
gradient {"type": "gradient", "value": "linear-gradient(180deg, #f8fafc 0%, #e2e8f0 100%)"} CSS gradient

When to use background image:

  • If a global background image was generated in Phase 3, set type: "image"
  • If no background image, use type: "color" with appropriate color

Important:

  • fontTitle and fontBody values must be valid CSS font-family strings (with quotes around font names containing spaces)
  • These values will be directly inserted into CSS variables in index.html
  • Font sizes are NOT configurable globally - templates use Tailwind's built-in size classes (text-4xl, text-xl, etc.) which are optimized for each layout

Fill in corresponding values based on color scheme and font solution selected in Phase 2.

5.2 Update ${WORKSPACE_DIR}/frontend/index.html Placeholders

  • {{PPT_TITLE}} → PPT title
  • {{FONT_IMPORTS}} → Google Fonts import links (see below)
  • {{PRIMARY_COLOR}} → Primary color (hex, e.g., #4285F4)
  • {{ACCENT_COLOR_1}} → Accent color 1 (hex)
  • {{ACCENT_COLOR_2}} → Accent color 2 (hex)
  • {{NEUTRAL_COLOR}} → Neutral color (hex)
  • {{SLIDE_BACKGROUND}} → Slide background (see below)
  • {{FONT_TITLE}} → Title font (e.g., 'Montserrat', 'Noto Sans SC', sans-serif)
  • {{FONT_BODY}} → Body font (e.g., 'Inter', 'Noto Sans SC', sans-serif)

Slide Background Configuration ({{SLIDE_BACKGROUND}}):

Based on background config in page-global-config.json, generate the appropriate CSS value:

Config Type CSS Value Example
{"type": "color", "value": "#FFFFFF"} #FFFFFF
{"type": "image", "value": "assets/images/bg.png"} url('/assets/images/bg.png')
{"type": "gradient", "value": "linear-gradient(...)"} linear-gradient(180deg, #f8fafc 0%, #e2e8f0 100%)

Usage in slide HTML (Agent applies this in geniekit-ppt-creator):

<!-- Use .slide-bg class on the main slide container -->
<div class="w-[1440px] h-[810px] slide-bg relative overflow-hidden">
  <!-- slide content -->
</div>

Font Import Configuration ({{FONT_IMPORTS}}): Only import fonts that are actually used in fontTitle and fontBody. Example:

<!-- If fontTitle uses Montserrat and fontBody uses Inter -->
<link rel="preconnect" href="https://fonts.googleapis.com">
<link rel="preconnect" href="https://fonts.gstatic.com" crossorigin>
<link href="https://fonts.googleapis.com/css2?family=Montserrat:wght@400;600;700&family=Inter:wght@300;400;600&family=Noto+Sans+SC:wght@400;500;700&display=swap" rel="stylesheet">

Common Font Import URLs:

Font Name Import URL Parameter
Noto Sans SC family=Noto+Sans+SC:wght@400;500;700
Noto Serif SC family=Noto+Serif+SC:wght@400;700
Montserrat family=Montserrat:wght@400;600;700
Inter family=Inter:wght@300;400;600
Poppins family=Poppins:wght@400;600;700
Roboto family=Roboto:wght@300;400;700

Note: Always include Noto Sans SC as fallback for Chinese characters.

Color Handling:

  • Templates use Tailwind's built-in color classes (e.g., text-blue-500, bg-gray-800)
  • Each template style has its own built-in color scheme optimized for that style
  • The CSS variable colors are available via custom utility classes for additional customization:
<!-- Custom color utility classes (use CSS variables) -->
<h1 class="text-primary">使用主色 --primary-color</h1>
<div class="bg-accent-1">使用强调色1背景</div>
<span class="border-primary border-2">主色边框</span>

<!-- Slide background class -->
.slide-bg  /* Uses --slide-background, supports color/image */

When to use custom colors vs Tailwind colors:

  • Tailwind colors (default): Templates already have coordinated color schemes
  • Custom colors: When generating new content that needs to match PPT theme precisely

Phase 6: Template Planning

Objective: For each page in chapters.md, select the appropriate template and update the file with template information. This enables parallel sub-agent execution in Phase 7.


6.1 Template Selection Process

Internal Checklist (Execute silently):

  1. Read ${WORKSPACE_DIR}/docs/page-global-config.json to get ${SlideStyle}
  2. Read ALL template index files for required page types | Page Type | Index File Path | |-----------|-----------------| | Cover | ${CORE_DIR}/skills/ppt-implement/references/templates/cover/cover-pages-index.md | | TOC | ${CORE_DIR}/skills/ppt-implement/references/templates/toc/toc-pages-index.md | | Transition | ${CORE_DIR}/skills/ppt-implement/references/templates/transition/transition-pages-index.md | | Content | ${CORE_DIR}/skills/ppt-implement/references/templates/content/content-templates-index.md | | Ending | ${CORE_DIR}/skills/ppt-implement/references/templates/ending/ending-pages-index.md |
  3. For EACH page, internally analyze:
    • Image requirements (yes/no, position)
    • Bullet point count
    • Content density (Light/Medium/Heavy)
  4. Apply matching rules below
  5. Verify diversity requirements (no Content template used >2 times)
  6. Update chapters.md directly with selections

6.1.1 Template Matching Rules (Priority Order)

Priority 1: Layout Structure Match (MOST IMPORTANT)

Page Characteristic Required Template Description Keywords
Has image image, photo, visual, picture
No image text-only, content-focused, text
Image on left image-left, left-image, photo-left
Image on right image-right, right-image, photo-right

Priority 2: Content Density Match

Bullet Point Count Template Style Keywords
1-2 points minimal, centered, spacious, hero
3-4 points balanced, two-column, standard
5+ points dense, list, multi-column, grid

Priority 3: Visual Diversity (ENFORCED)

  • ❌ Same Content template used consecutively = VIOLATION
  • ❌ Same Content template used >2 times total = VIOLATION
  • ✅ Each Content page should have DIFFERENT template when possible

6.1.2 Page Type Selection Guidelines

Cover Page
  • Only select templates from cover/${SlideStyle}/xxx.tpl
  • Match visual emphasis with content theme
TOC Page
  • Only select templates from toc/${SlideStyle}/xxx.tpl
  • Select layout based on chapter count:
    • 3-4 chapters: centered, whitespace, large-title
    • 5-6 chapters: list, grid, multi-column
    • 7+ chapters: multi-image-grid, dense, two-column
Transition Page
  • Only select templates from transition/${SlideStyle}/xxx.tpl
  • Reuse the SAME template for ALL transition pages
Ending Page
  • Only select templates from ending/${SlideStyle}/xxx.tpl
  • Match ending purpose (Thank-you/Q&A/Contact/Summary)
Content Page
  • Only select templates from content/${SlideStyle}/xxx.tpl
  • Apply Priority 1-3 matching rules
  • ❌ Same template used consecutively = VIOLATION
  • ❌ Same template used >2 times total = VIOLATION
## Page N: Page Name
- **Page Type**: Cover/TOC/Transition/Content/Ending
- **Page Title**: XXX
- **Selected Template**: {templatePath}
... (other existing fields)

Phase 7: Sequential PPT Generation

CRITICAL: DO NOT use any other skill/command/agent to complete the following tasks.

CRITICAL: ONE PAGE AT A TIME

You MUST generate slides one by one. For EACH page, you MUST:

  1. Read its template file
  2. Process content (adapt, handle images, check overflow)
  3. Write its slide-{N}.js file
  4. Then move to next page

Even if you decide to "batch create" or "generate all pages", you MUST still execute the full single-page workflow (7.2-7.6) for EACH page. There is NO shortcut. Skipping steps will cause broken slides.


Read inputs ONCE before generating all slides:

  • ${WORKSPACE_DIR}/docs/page-global-config.json → global styles/colors
  • ${WORKSPACE_DIR}/docs/product/chapters.md → content for each page

7.1 For Each Page (Sequential Loop)

For each page in ${WORKSPACE_DIR}/docs/product/chapters.md, execute steps 7.2-7.6:

7.2 Read Template

Read from: ${CORE_DIR}/skills/ppt-implement/references/templates/{chapters.page.selected_template}

7.3 Apply Global Background

Find outermost <div class="w-[1440px] h-[810px] ...">, remove hardcoded backgrounds, add slide-bg:

<!-- Before -->
<div class="w-[1440px] h-[810px] shadow-2xl relative overflow-hidden bg-gradient-to-br from-slate-900 via-blue-900 to-slate-900">

<!-- After -->
<div class="w-[1440px] h-[810px] shadow-2xl relative overflow-hidden slide-bg">

Remove: bg-white, bg-[#xxx], bg-slate-*, bg-gradient-to-*, etc. CRITICAL: Never use gradient colors for background or text.

7.4 Fill Content

Content Adaptation:

Situation Strategy
Content items > template slots Condense: merge similar points, max 6 items
Content items < template slots Remove unused template elements entirely
Text too long for slot Rewrite concisely, keep core message

Text Length Limits:

Element CN EN
Title ≤12 ≤25
Subtitle ≤18 ≤35
Card Title ≤10 ≤20
Body Line ≤25 ≤50

Font Size Specs: H1: 38-42px, H2: 30-32px, H3: 20-22px, Body: 16-18px, Annotation: 12-14px

Handle Placeholder Images (CRITICAL):

Situation Action
Chapter has assigned image Replace src with path from images-mapping.json, set explicit width/height
Chapter has NO image assigned DELETE the entire <img> element and its empty container
<!-- Template placeholder -->
<div class="image-container">
  <img src="/placeholder.svg"/>
</div>

<!-- If NO image assigned → DELETE entirely (the whole div) -->

<!-- If image assigned → Replace src with explicit dimensions -->
<div class="image-container">
  <img src="/assets/images/page-5.png" width="400px" height="300px"/>
</div>

⚠️ NEVER leave placeholder images - they show as broken images!

7.5 Content Overflow Prevention

Page Constraints: Content area is 1350px × 720px.

Before writing, check and fix in priority order:

Priority Check Fix Strategy
1 Total items > 6 Reduce to max 6 items
2 Text line > limit Truncate/rewrite to fit char limits
3 Vertical stack too tall Reduce gap-*/space-y-* (e.g., gap-6→gap-3)
4 Still too tall Reduce font sizes by 1 level (e.g., text-2xl→text-xl)
5 Images too large Set explicit max-w-[Xpx] max-h-[Ypx]
6 Cards overflow Reduce p-* padding (e.g., p-6→p-4)

Quick Estimation: Text line ≈ 24-32px, Card with title + 3 lines ≈ 120-150px, Gap ≈ 16-24px

7.6 Write Slide File

Write to ${WORKSPACE_DIR}/frontend/src/slides/slide-{N}.js:

window.slideDataMap.set({N}, `
  <div class="w-[1440px] h-[810px] shadow-2xl relative overflow-hidden slide-bg">
    <!-- Filled HTML content using Tailwind CSS -->
  </div>
`);

N is integer (not string). Framework auto-loads via window.slideDataMap.

7.7 After All Slides Generated

Proceed directly to Phase 8. No manual wiring needed.

Phase 8: Take PPT Screenshots

⚠️ Phase 8 is the ONLY legitimate next step after Phase 7 completes. It MUST be executed immediately - DO NOT skip or delay!

8.1 Take Screenshots

CRITICAL: The service is already running on port 5173 by default, no need to start it again.

Execute the following command immediately (no user confirmation needed):

python3 "${CORE_DIR}/skills/ppt-implement/scripts/screenshot-ppt.py" --url http://localhost:5173 --output ${WORKSPACE_DIR}/frontend/public/assets/images/posters/pages

8.2 Update poster Field in ${WORKSPACE_DIR}/docs/pages.json

After screenshots are complete, immediately update the poster field (no user confirmation needed):

"poster": "/assets/images/posters/pages/page-{N}.png",

Example:

[
  {
    "pageKey": "ppt-1",
    "title": "",
    "url": "/index.html?page=1",
    "poster": "/assets/images/posters/pages/page-1.png",
    "pageNum": 1
  }
]

Phase 9: Final Output

⚠️ Output ONLY the following 3 lines, say nothing more:

"{PPT_TOPIC}" PPT generation complete.
"presentation.pptx" generated.

Usage: Press ← → to navigate pages, press Space for next page

🚫 NEVER output the following:

  • ❌ Tables (any form of tables)
  • ❌ Page number lists
  • ❌ File paths
  • ❌ "Summary" / "总结"
  • ❌ "All N pages completed"
  • ❌ "Files generated to..."
  • ❌ Any validation statements

Phase 10: Handling User Modifications

⚠️ This phase is ONLY triggered when users request modifications after initial PPT generation is complete.

  • First-time generation: Phase 1 → Phase 8, then STOP
  • User requests changes: Enter Phase 9

Understanding Slide File Structure

Each PPT page is stored in ${WORKSPACE_DIR}/frontend/src/slides/slide-N.js, which registers HTML content to a global Map:

// slide-N.js structure
window.slideDataMap.set(N, `
  <div class="...">
    <!-- Page N HTML content -->
  </div>
`);

The ppt-controller.js reads from slideDataMap and renders pages dynamically. Therefore, modifying page content requires editing slide-N.js files.

When users request modifications, determine which files need to be changed based on the scope of modification:

10.1 Modification Scope Identification

Modification Type Example Files to Modify Notes
Single Page Content "Change the title of page 3 to..." chapters.md → slide-3.js Content only, no template change
Add/Delete Pages "Add a page after page 5" chapters.md → Create/delete slide-N.js New pages need template selection
Single Page Template "Change layout of page 4" chapters.md(update Selected Template) → Regenerate slide-4.js Switch template within same style
Image Replacement "Replace the image on page 6" images-mapping.json → images/ directory → Update path in slide-6.js May need to regenerate images
Overall Style "Change to Chinese style" See detailed flow below Largest impact scope

10.2 Overall Style Modification Flow

When users request a style change (e.g., "change to Chinese style", "make it more tech-looking"), modify the following files in order:

Step File Changes
1 ${WORKSPACE_DIR}/docs/page-global-config.json Update slideStyle, color scheme, font scheme
2 ${WORKSPACE_DIR}/frontend/index.html Update CSS variables (colors, font import links)
3 ${WORKSPACE_DIR}/docs/product/chapters.md Re-select templates for ALL pages (Phase 6 Template Planning)
4 ${WORKSPACE_DIR}/frontend/src/slides/slide-*.js Regenerate ALL slide files (Phase 7 Parallel Generation)
5 ${WORKSPACE_DIR}/frontend/public/assets/images/ May need to regenerate some images to match new style

⚠️ Important Notes:

  • Style modification = Full redo of Phase 2 ~ Phase 7
  • Templates are organized by style (e.g., cover/chinese/, cover/business/), changing style requires re-selecting all templates
  • Transition page templates must be globally consistent, need to re-specify after style change

10.3 Modification Flow Summary

User requests modification
         │
         ▼
┌─────────────────────┐
│ Identify scope      │
└──────────┬──────────┘
           │
    ┌──────┴──────┬────────────┬────────────┐
    ▼             ▼            ▼            ▼
Single Page   Single Page   Add/Delete   Overall
 Content       Template      Pages        Style
    │             │            │            │
    ▼             ▼            ▼            ▼
  Update        Update       Update      Update
 chapters      chapters     chapters     config
    │             │            │            │
    ▼             ▼            ▼            ▼
 Regenerate   Regenerate   Create/Del   Re-select
 slide-N.js   slide-N.js   slide-N.js   all templates
                                           │
                                           ▼
                                      Regenerate
                                      ALL slide-*.js

Export PPTX File

When user requests to export a PPTX file, run the following command:

node \"${CORE_DIR}/skills/ppt-implement/scripts/export-ppt.js\" \"${CORE_DIR}\" \"${WORKSPACE_DIR}\" > /dev/null 2>&1

04 同仓库更多 Skills

YYC-Cube
YYC-Cube/YYC3-Knowledge-Base

vercel-sandbox

Run agent-browser + Chrome inside Vercel Sandbox microVMs for browser automation from any Vercel-deployed app. Use when the user needs browser automation in a Vercel app (Next.js, SvelteKit, Nuxt, Remix, Astro, etc.), wants to run headless Chrome without binary size limits, needs persistent browser sessions across commands, or wants ephemeral isolated browser environments. Triggers include "Vercel Sandbox browser", "microVM Chrome", "agent-browser in sandbox", "browser automation on Vercel", or any task requiring Chrome in a Vercel Sandbox.

★ 1
YYC-Cube
YYC-Cube/YYC3-Knowledge-Base

slack

Interact with Slack workspaces using browser automation. Use when the user needs to check unread channels, navigate Slack, send messages, extract data, find information, search conversations, or automate any Slack task. Triggers include "check my Slack", "what channels have unreads", "send a message to", "search Slack for", "extract from Slack", "find who said", or any task requiring programmatic Slack interaction.

★ 1
YYC-Cube
YYC-Cube/YYC3-Knowledge-Base

electron

Automate Electron desktop apps (VS Code, Slack, Discord, Figma, Notion, Spotify, etc.) using agent-browser via Chrome DevTools Protocol. Use when the user needs to interact with an Electron app, automate a desktop app, connect to a running app, control a native app, or test an Electron application. Triggers include "automate Slack app", "control VS Code", "interact with Discord app", "test this Electron app", "connect to desktop app", or any task requiring automation of a native Electron application.

★ 1
YYC-Cube
YYC-Cube/YYC3-Knowledge-Base

dogfood

Systematically explore and test a web application to find bugs, UX issues, and other problems. Use when asked to "dogfood", "QA", "exploratory test", "find issues", "bug hunt", "test this app/site/platform", or review the quality of a web application. Produces a structured report with full reproduction evidence -- step-by-step screenshots, repro videos, and detailed repro steps for every issue -- so findings can be handed directly to the responsible teams.

★ 1
YYC-Cube
YYC-Cube/YYC3-Knowledge-Base

agent-browser

Browser automation CLI for AI agents. Use when the user needs to interact with websites, including navigating pages, filling forms, clicking buttons, taking screenshots, extracting data, testing web apps, or automating any browser task. Triggers include requests to "open a website", "fill out a form", "click a button", "take a screenshot", "scrape data from a page", "test this web app", "login to a site", "automate browser actions", or any task requiring programmatic web interaction.

★ 1
YYC-Cube
YYC-Cube/YYC3-Knowledge-Base

shadcn

Manages shadcn components and projects — adding, searching, fixing, debugging, styling, and composing UI, including chat interfaces. Provides project context, component docs, and usage examples. Applies when working with shadcn/ui, component registries, presets, --preset codes, or any project with a components.json file. Also triggers for "shadcn init", "create an app with --preset", "switch to --preset", and for building a design system or applying a DESIGN.md or brand spec to shadcn/ui.

★ 1