pptx

name	pptx
description	Use this skill whenever the user wants to do anything with PowerPoint presentations — reading, extracting text, creating professional slide decks.
metadata	{"yiyi":{"emoji":"📊","requires":{"pip":["python-pptx"]}}}

PowerPoint Processing

Prerequisites

Before running any script in this skill, install the Python dependency:

pip_install(packages=["python-pptx"])

Call the pip_install tool (not shell). packages is an array of strings (see the tool's schema). It's idempotent — safe to call even if already installed.

Reading & Extracting Text

PPTX files are ZIP archives containing XML. Extract content with shell tools:

# List all slides
unzip -l presentation.pptx | grep slide

# Extract all text from all slides
unzip -p presentation.pptx ppt/slides/slide*.xml | sed 's/<[^>]*>//g'

# Extract specific slide
unzip -p presentation.pptx ppt/slides/slide3.xml | sed 's/<[^>]*>//g'

Or use Python for structured extraction:

from pptx import Presentation
prs = Presentation("presentation.pptx")
for i, slide in enumerate(prs.slides):
    print(f"--- Slide {i+1} ---")
    for shape in slide.shapes:
        if shape.has_text_frame:
            print(shape.text)

---

Creating Professional Presentations

Use the create_pptx.py script to generate polished, themed slide decks.

Step 1: Build a JSON structure file

Create a JSON file describing the presentation. Include rich content — detailed bullet points, not one-word items.

{
  "title": "Presentation Title",
  "subtitle": "Descriptive subtitle",
  "author": "Author Name",
  "date": "2024-01-15",
  "theme": "dark",
  "slides": [
    { "type": "title", "title": "Main Title", "subtitle": "Subtitle" },
    { "type": "section", "title": "Section Name" },
    { "type": "content", "title": "Slide Title", "body": ["Point 1", "Point 2"] },
    ...
  ]
}

Step 2: Run the script

python3 scripts/create_pptx.py structure.json output.pptx

Available Themes

Theme	Style
dark	Dark navy background, blue/cyan accents — modern tech feel
light	White background, navy/blue accents — clean and professional
corporate	White background, navy/red accents — formal business style

Slide Types

Title Slide (opening slide)

{ "type": "title", "title": "Main Title", "subtitle": "Subtitle text", "author": "Name", "date": "2024-01-01" }

Section Divider

{ "type": "section", "title": "Section Name", "subtitle": "Optional description" }

Content Slide (bullet points)

{ "type": "content", "title": "Slide Title", "body": ["Detailed point one with explanation", "Point two with supporting data", "Point three with context"] }

Two-Column Layout

{ "type": "two_column", "title": "Comparison Title", "left_title": "Option A", "left": ["Left point 1", "Left point 2"], "right_title": "Option B", "right": ["Right point 1", "Right point 2"] }

Table Slide

{ "type": "table", "title": "Data Overview", "headers": ["Name", "Role", "Status"], "rows": [["Alice", "Engineer", "Active"], ["Bob", "Designer", "Active"]] }

Quote / Callout

{ "type": "quote", "text": "An impactful quote or key insight.", "author": "Attribution" }

Image Slide

{ "type": "image", "title": "Architecture Diagram", "path": "/absolute/path/to/image.png", "caption": "Figure 1: System overview" }

Key Metrics / KPI Dashboard

{ "type": "key_metrics", "title": "Performance KPIs", "metrics": [{"label": "Revenue", "value": "$10M", "change": "+15%"}, {"label": "Users", "value": "50K", "change": "+30%"}, {"label": "Uptime", "value": "99.9%", "change": "+0.2%"}] }

• change: Prefix with + for green (positive) or - for red (negative)
• Best with 2–4 metrics per slide

Timeline / Roadmap

{ "type": "timeline", "title": "Project Roadmap", "events": [{"date": "Q1", "text": "Phase 1 launch"}, {"date": "Q2", "text": "Feature expansion"}, {"date": "Q3", "text": "Scale to production"}] }

• Best with 3–5 events

Thank You / Closing

{ "type": "thank_you", "title": "Thank You", "message": "Questions or feedback welcome", "contact": "[e-mail]" }

Blank Slide

{ "type": "blank" }

Content Guidelines

1. Bullet points should be substantive: Each bullet should be a full phrase or short sentence, not just a keyword. "Revenue grew 43% driven by new product launch" is better than "Revenue +43%".

2. Structure the presentation well:

   • Start with a title slide
   • Use section slides to divide major topics
   • Mix slide types — don't use only content slides
   • End with thank_you

3. Recommended flow (8–15 slides):

   • Title → Section → Content/Metrics → Section → Two-Column/Table → Quote → Timeline → Thank You

4. Use key_metrics for numbers: When presenting KPIs or statistics, use key_metrics instead of putting numbers in bullet points.

5. Use two_column for comparisons: Before/after, pros/cons, option A vs B — use two_column instead of a single list.

6. Chinese content: python-pptx uses system fonts. Chinese text renders correctly on systems with CJK fonts installed.

Example: Complete Presentation

{
  "title": "2024 年度技术报告",
  "subtitle": "人工智能与机器学习应用进展分析",
  "author": "技术部",
  "date": "2024-12-01",
  "theme": "dark",
  "slides": [
    { "type": "title", "title": "2024 年度技术报告", "subtitle": "AI 应用进展分析", "author": "技术部", "date": "2024-12-01" },

    { "type": "section", "title": "核心项目进展", "subtitle": "本年度 12 个 AI 系统成功部署" },

    { "type": "content", "title": "智能客服系统", "body": [
      "基于大语言模型构建对话引擎，覆盖 85% 常见客户咨询",
      "RAG 架构结合企业知识库，实现精准上下文问答",
      "累计处理 230 万轮对话，平均响应时间 1.2 秒",
      "首次解决率 78%，较人工客服成本降低 62%"
    ]},

    { "type": "key_metrics", "title": "关键业务指标", "metrics": [
      {"label": "对话处理量", "value": "230万", "change": "+180%"},
      {"label": "平均响应", "value": "1.2s", "change": "-85%"},
      {"label": "客户满意度", "value": "94.2%", "change": "+12%"}
    ]},

    { "type": "two_column", "title": "数据平台升级对比", "left_title": "升级前", "left": [
      "手动查询，平均响应 45 秒",
      "日活跃用户仅 120 人",
      "报表生成需要 2 小时"
    ], "right_title": "升级后", "right": [
      "自然语言查询，3 秒响应",
      "日活跃用户增至 580 人",
      "报表 10 分钟自动生成"
    ]},

    { "type": "table", "title": "项目进度总览", "headers": ["项目", "状态", "完成率", "负责人"], "rows": [
      ["智能客服系统", "已上线", "100%", "张三"],
      ["数据分析平台", "测试中", "85%", "李四"],
      ["推荐引擎", "开发中", "60%", "王五"]
    ]},

    { "type": "quote", "text": "好的架构不是一次设计出来的，而是在持续迭代中逐步演进的。我们的架构每季度评审优化一次。", "author": "技术架构组" },

    { "type": "timeline", "title": "2025 年路线图", "events": [
      {"date": "Q1", "text": "多模态模型评估与选型"},
      {"date": "Q2", "text": "边缘设备适配与性能优化"},
      {"date": "Q3", "text": "MLOps 平台 2.0 上线"},
      {"date": "Q4", "text": "全面评估与下一年规划"}
    ]},

    { "type": "thank_you", "title": "谢谢", "message": "如有问题请随时联系技术部", "contact": "[e-mail]" }
  ]
}

---

Advanced Operations

Convert PPTX to PDF

# Using LibreOffice
soffice --headless --convert-to pdf presentation.pptx

Extract Images from PPTX

unzip -j presentation.pptx ppt/media/* -d ./extracted_images/

Modify Existing PPTX

from pptx import Presentation
prs = Presentation("existing.pptx")
slide = prs.slides[0]
slide.shapes.title.text = "Updated Title"
prs.save("modified.pptx")

Quick Reference

Task	Approach
Extract text	Shell unzip -p + sed
Extract images	unzip -j from ppt/media/
Create professional deck	create_pptx.py script
Modify existing PPTX	Python python-pptx
Convert to PDF	LibreOffice CLI