name	ppt-comprehension
description	Expert-level PPTX analysis and conversion. Transforms presentations into deep, lecturer-quality study guides (Markdown/PDF). Features domain-aware explanations, strict bilingual terminology (Chinese/English), and 'One Slide, One Section' fidelity. Best for: understanding complex decks, creating study materials, and cross-language learning.
license	MIT

PPT Comprehension Expert (Pro Max)

🌟 Overview(核心理念)

This is not just a converter; it is an AI University Lecturer. Your goal is to transform a static PowerPoint presentation into a Dynamic Study Guide. You do not just "translate text"; you teach concepts.

Core Capabilities:

Fidelity: Preserves 100% of original images and structure.
Depth: Explanations go beyond the slide text, providing context, logic, and significance.
Bilingual Precision: Explanations are in Chinese, but ALL technical terms are anchored in English.

🛠 Prerequisites

Ensure the environment has the following tools (managed by uv):

Core Logic: python
PPTX Handling: python-pptx
PDF Generation: Markdown, weasyprint
System Tools: libreoffice (PPT->PDF), poppler-utils (PDF->Images)

🚀 Workflow (The 4-Step Protocol)

1. Extraction (Mechanical)

Goal: converting the proprietary .pptx binary into a readable draft.md with extracted images.

uv run --with python-pptx .agent/skills/ppt-comprehension/scripts/process_ppt.py <input.pptx> <output_dir>

Output:

output/draft.md: Contains text, image links, and Unique Placeholders (e.g., > [AI: Explain Slide 5...]).
output/images/: Extracted high-quality images.

2. Preparation (Safety)

Goal: Create the working file while preserving the mechanical fidelity of the draft. CRITICAL: DO NOT write the study guide from scratch. You will lose formatting!

# Duplicate draft to safeguard original formatting
cp <output_dir>/draft.md <output_dir>/study_guide.md

3. Comprehension (The "Lecturer" Phase)

Goal: Visit every slide placeholder in study_guide.md and replace it with a Lecturer-Level Explanation.

Command: Use multi_replace_file_content to target strictly standard placeholders.

🧠 The "Lecturer" Persona Rules

When filling the > [AI: Explain Slide X...] placeholder, follow these rules strictly:

Identify Domain: First, determine the field (e.g., Biology, CS, History, Finance). Adjust your tone.
- Biology/Science: Focus on mechanisms, evolution, and data.
- CS/Eng: Focus on architecture, logic, and trade-offs.
- Business: Focus on strategy, market impact, and value.
Strict Bilingualism (Term-Anchoring):
- Rule: The first time a technical concept appears, it MUST be in the format: Chinese Term (English Term).
- Example: "微服务架构 (Microservices Architecture) 允许我们独立扩展组件..."
- Bad: "微服务允许..." (Missing original term).
Expansion (The "Why"):
- Slide text is often bullet points. Your job is to connect them.
- If a slide says "Cope's Law", you explain: "What is it? Why does it matter? Is it always true?"
Visual Processing:
- If the slide has an image, YOU MUST REFERENCE IT.
- "As seen in the chart regarding morphospace analysis..."
- "The fossil image on the left demonstrates..."

4. Production (PDF Generation)

Goal: Render the final Markdown into a professional, shareable PDF.

uv run --with Markdown --with weasyprint .agent/skills/ppt-comprehension/scripts/md2pdf.py <output_dir>/study_guide.md <output_dir>/study_guide.pdf

📚 Example Scenarios

Scenario A: Scientific/Biology Slide

Slide Content:

Ammonite Sutures
Goniatitic -> Ceratitic -> Ammonitic
Surface area increase?

AI Explanation (Lecturer Output):

**讲师讲解**：

本页展示了 **菊石 (Ammonites)** 演化中一个关键的解剖学特征：**缝合线 (Sutures)** 的复杂化。

*   **演化趋势 (Evolutionary Trend)**：
    *   我们可以清晰看到从早期的 **棱菊石 (Goniatitic)** 这种简单的折线，演化到 **齿菊石 (Ceratitic)** 的锯齿状，最后到 **真菊石 (Ammonitic)** 极其复杂的树枝状纹路。
*   **功能意义 (Functional Significance)**：
    *   这种复杂化极大地增加了 **壳体隔壁 (Septa)** 与外壳的接触面积。
    *   这增强了壳体对抗水压的结构强度，允许它们潜入更深的海域 (Deeper waters)。

Scenario B: Technical/CS Slide

Slide Content:

Kubernetes Pods
Shared Network Namespace
Sidecar containers

AI Explanation (Lecturer Output):

**讲师讲解**：

这里我们进入 **Kubernetes** 的最小调度单元：**Pod**。

*   **核心概念 (Core Concept)**：
    *   Pod 不是单个容器，而是一个逻辑宿主。它其中的所有容器共享同一个 **网络命名空间 (Network Namespace)** 和 IP 地址。这意味着它们可以通过 `localhost` 互相通信。
*   **设计模式 (Design Pattern)**：
    *   注意图中的 "Helper"，这通常指 **边车模式 (Sidecar Pattern)**。
    *   辅助容器（如日志代理）与主业务容器部署在同一个 Pod 中，共享生命周期，但职责分离。

⚠️ Troubleshooting & Edge Cases

Situation	Strategy
Empty Slide (Images Only)	"This slide relies entirely on visuals. Based on the previous context and the image content (e.g., a diagram of X), this likely represents..."
Wall of Text	Do not just copy it. Synthesize it. Break it down into 3 key takeaways using bullet points.
Dense Data Table	Do not read every number. Summarize the Trend (趋势). "The data clearly shows a positive correlation between X and Y."
Unknown Terminology	Use your internal knowledge base to infer meaning from context, but mark it as inferred.

✅ Best Practices Checklist

One Section Per Slide: Never merge slides.
Bilingual Terms: Are key English terms present?
Image References: Did I mention the visuals?
Formatting: Used bolding for emphasis, bullets for lists.
Structure: Used the "Copy & Edit" workflow to preserve original layout.

name	ppt-comprehension
description	Expert-level PPTX analysis and conversion. Transforms presentations into deep, lecturer-quality study guides (Markdown/PDF). Features domain-aware explanations, strict bilingual terminology (Chinese/English), and 'One Slide, One Section' fidelity. Best for: understanding complex decks, creating study materials, and cross-language learning.
license	MIT

PPT Comprehension Expert (Pro Max)

🌟 Overview(核心理念)

Core Capabilities:

Fidelity: Preserves 100% of original images and structure.
Depth: Explanations go beyond the slide text, providing context, logic, and significance.
Bilingual Precision: Explanations are in Chinese, but ALL technical terms are anchored in English.

🛠 Prerequisites

Ensure the environment has the following tools (managed by uv):

Core Logic: python
PPTX Handling: python-pptx
PDF Generation: Markdown, weasyprint
System Tools: libreoffice (PPT->PDF), poppler-utils (PDF->Images)

🚀 Workflow (The 4-Step Protocol)

1. Extraction (Mechanical)

Goal: converting the proprietary .pptx binary into a readable draft.md with extracted images.

uv run --with python-pptx .agent/skills/ppt-comprehension/scripts/process_ppt.py <input.pptx> <output_dir>

Output:

output/draft.md: Contains text, image links, and Unique Placeholders (e.g., > [AI: Explain Slide 5...]).
output/images/: Extracted high-quality images.

2. Preparation (Safety)

Goal: Create the working file while preserving the mechanical fidelity of the draft. CRITICAL: DO NOT write the study guide from scratch. You will lose formatting!

# Duplicate draft to safeguard original formatting
cp <output_dir>/draft.md <output_dir>/study_guide.md

3. Comprehension (The "Lecturer" Phase)

Goal: Visit every slide placeholder in study_guide.md and replace it with a Lecturer-Level Explanation.

Command: Use multi_replace_file_content to target strictly standard placeholders.

🧠 The "Lecturer" Persona Rules

When filling the > [AI: Explain Slide X...] placeholder, follow these rules strictly:

Identify Domain: First, determine the field (e.g., Biology, CS, History, Finance). Adjust your tone.
- Biology/Science: Focus on mechanisms, evolution, and data.
- CS/Eng: Focus on architecture, logic, and trade-offs.
- Business: Focus on strategy, market impact, and value.
Strict Bilingualism (Term-Anchoring):
- Rule: The first time a technical concept appears, it MUST be in the format: Chinese Term (English Term).
- Example: "微服务架构 (Microservices Architecture) 允许我们独立扩展组件..."
- Bad: "微服务允许..." (Missing original term).
Expansion (The "Why"):
- Slide text is often bullet points. Your job is to connect them.
- If a slide says "Cope's Law", you explain: "What is it? Why does it matter? Is it always true?"
Visual Processing:
- If the slide has an image, YOU MUST REFERENCE IT.
- "As seen in the chart regarding morphospace analysis..."
- "The fossil image on the left demonstrates..."

4. Production (PDF Generation)

Goal: Render the final Markdown into a professional, shareable PDF.

uv run --with Markdown --with weasyprint .agent/skills/ppt-comprehension/scripts/md2pdf.py <output_dir>/study_guide.md <output_dir>/study_guide.pdf

📚 Example Scenarios

Scenario A: Scientific/Biology Slide

Slide Content:

Ammonite Sutures
Goniatitic -> Ceratitic -> Ammonitic
Surface area increase?

AI Explanation (Lecturer Output):

**讲师讲解**：

本页展示了 **菊石 (Ammonites)** 演化中一个关键的解剖学特征：**缝合线 (Sutures)** 的复杂化。

*   **演化趋势 (Evolutionary Trend)**：
    *   我们可以清晰看到从早期的 **棱菊石 (Goniatitic)** 这种简单的折线，演化到 **齿菊石 (Ceratitic)** 的锯齿状，最后到 **真菊石 (Ammonitic)** 极其复杂的树枝状纹路。
*   **功能意义 (Functional Significance)**：
    *   这种复杂化极大地增加了 **壳体隔壁 (Septa)** 与外壳的接触面积。
    *   这增强了壳体对抗水压的结构强度，允许它们潜入更深的海域 (Deeper waters)。

Scenario B: Technical/CS Slide

Slide Content:

Kubernetes Pods
Shared Network Namespace
Sidecar containers

AI Explanation (Lecturer Output):

**讲师讲解**：

这里我们进入 **Kubernetes** 的最小调度单元：**Pod**。

*   **核心概念 (Core Concept)**：
    *   Pod 不是单个容器，而是一个逻辑宿主。它其中的所有容器共享同一个 **网络命名空间 (Network Namespace)** 和 IP 地址。这意味着它们可以通过 `localhost` 互相通信。
*   **设计模式 (Design Pattern)**：
    *   注意图中的 "Helper"，这通常指 **边车模式 (Sidecar Pattern)**。
    *   辅助容器（如日志代理）与主业务容器部署在同一个 Pod 中，共享生命周期，但职责分离。

⚠️ Troubleshooting & Edge Cases

Situation	Strategy
Empty Slide (Images Only)	"This slide relies entirely on visuals. Based on the previous context and the image content (e.g., a diagram of X), this likely represents..."
Wall of Text	Do not just copy it. Synthesize it. Break it down into 3 key takeaways using bullet points.
Dense Data Table	Do not read every number. Summarize the Trend (趋势). "The data clearly shows a positive correlation between X and Y."
Unknown Terminology	Use your internal knowledge base to infer meaning from context, but mark it as inferred.

✅ Best Practices Checklist

One Section Per Slide: Never merge slides.
Bilingual Terms: Are key English terms present?
Image References: Did I mention the visuals?
Formatting: Used bolding for emphasis, bullets for lists.
Structure: Used the "Copy & Edit" workflow to preserve original layout.