sci-html

Compare original and translation side by side

🇺🇸

Original

English

🇨🇳

Translation

Chinese

Sci HTML

Use this skill to create browser-based academic slide decks inside the Aut_Sci_Write suite. It can work from prepared Markdown/JSON, or run the full paper workflow from PDF to structured insights, extracted figures, and HTML deck.

Use

sci-html

for shareable browser output. Use

sci-ppt

when the user needs a real

.pptx

. Use

sci-extract

when the user only wants a text analysis without a presentation artifact.

本技能用于在Aut_Sci_Write套件中创建基于浏览器的学术幻灯片集。它可以基于已准备好的Markdown/JSON文件运行，也可以执行从PDF到结构化见解、提取图表再到HTML幻灯片集的完整论文处理流程。

如需可共享的浏览器输出，请使用

sci-html

。若用户需要真实的

.pptx

文件，请使用

sci-ppt

。若用户仅需要文本分析而无需演示成品，请使用

sci-extract

。

When To Use

使用场景

Use this skill when the user asks for HTML slides, clickable presentation webpages, browser-based academic presentations, Markdown to slide deck, paper summary to web presentation, PDF paper to HTML report, or a PPT-like HTML file with left/right navigation.

当用户需要HTML幻灯片、可点击演示网页、基于浏览器的学术演示文稿、Markdown转幻灯片集、论文摘要转网页演示文稿、PDF论文转HTML报告，或带左右导航的类PPT HTML文件时，可使用本技能。

Output

输出内容

Always create a deck directory containing

index.html

deck.json

assets/style.css

assets/deck.js

, and copied local figures under

assets/figures/

Do not create single-file HTML. Directory output is the standard mode because it keeps the presentation maintainable and avoids oversized embedded assets.

始终生成包含

index.html

、

deck.json

、

assets/style.css

、

assets/deck.js

以及复制到

assets/figures/

下的本地图表的幻灯片目录。

请勿生成单文件HTML。目录输出为标准模式，因为这种模式便于维护演示文稿，且可避免嵌入资源过大的问题。

Workflow

工作流程

Parse the user's source content into a deck model.
Add cover, table of contents, section, summary, and ending slides when appropriate.
Validate image paths and slide structure.
Render HTML, CSS, and JS.
Tell the user the output path and navigation controls.

将用户的源内容解析为幻灯片模型。
酌情添加封面、目录、章节、摘要和结尾幻灯片。
验证图片路径和幻灯片结构。
渲染HTML、CSS和JS。
告知用户输出路径和导航控件。

PDF Paper Workflow

PDF论文处理流程

For a source PDF, use the integrated

paper

command rather than requiring separate

sci-extract

sci-figure

skills:

bash

python scripts/generate_html_deck.py paper paper.pdf -o out/paper-report --author "Presenter"

When working from the Aut_Sci_Write repository root, run:

bash

python skills/sci-html/scripts/generate_html_deck.py paper paper.pdf -o skills/sci-html/out/paper-report --author "Presenter"

This writes

structured_insights.json

figure_manifest.json

, extracted files under

figures/

, a generated

{paper-stem}-deck.md

, and the final

index.html

The PDF workflow uses bundled integrations copied from

sci-extract

and

sci-figure

. It requires

PyMuPDF

pdfplumber

numpy

opencv-python

Pillow

, and

pytesseract

; Tesseract OCR is optional and only affects subfigure OCR.

The integrated

sci-extract

logic detects original research versus review literature. Original research decks emphasize problem, method, results, innovation, applications, and limits. Review, survey, systematic review, scoping review, and meta-analysis decks emphasize scope, taxonomy, evidence base, consensus, controversies, evidence quality, research gaps, future directions, and important figures or tables. Use

--paper-type auto|research|review

when the user wants to override automatic routing.

对于源PDF文件，请使用集成的

paper

命令，无需单独使用

sci-extract

或

sci-figure

技能：

bash

python scripts/generate_html_deck.py paper paper.pdf -o out/paper-report --author "Presenter"

在Aut_Sci_Write仓库根目录下运行：

bash

python skills/sci-html/scripts/generate_html_deck.py paper paper.pdf -o skills/sci-html/out/paper-report --author "Presenter"

该命令会生成

structured_insights.json

、

figure_manifest.json

、

figures/

下的提取文件、自动生成的

{paper-stem}-deck.md

以及最终的

index.html

。

PDF处理流程使用从

sci-extract

和

sci-figure

复制的捆绑集成组件。它依赖

PyMuPDF

、

pdfplumber

、

numpy

、

opencv-python

、

Pillow

和

pytesseract

；Tesseract OCR为可选组件，仅对子图表OCR产生影响。

集成的

sci-extract

逻辑可区分原创研究文献与综述文献。原创研究幻灯片集重点突出问题、方法、结果、创新点、应用场景和局限性。综述、调研、系统综述、范围综述和元分析幻灯片集则重点突出研究范围、分类体系、证据基础、共识观点、争议点、证据质量、研究空白、未来方向以及重要图表。若用户需要覆盖自动分类，请使用

--paper-type auto|research|review

参数。

Presentation Markup

演示文稿标记规则

Put
```
layout: section
```
at the top of a Markdown slide block to create a chapter divider.
Put
```
subtitle: 01
```
below
```
layout: section
```
to show the chapter number.
Use
```
**key phrase**
```
or
```
==key phrase==
```
for red emphasis.
Keep bullets concise; generated bullet slides use numbered key-point cards by default.
When a slide has only one or two content bullets, render it as plain statements without visible item numbers.
Long slide and paper titles are automatically fitted by shrinking the heading and expanding its line width so the title stays on the page.
Sparse slides automatically receive subtle topic motifs from title and bullet keywords, so visually empty areas are filled without requiring extra image assets.

在Markdown幻灯片块顶部添加
```
layout: section
```
以创建章节分隔页。
在
```
layout: section
```
下方添加
```
subtitle: 01
```
以显示章节编号。
使用
```
**关键短语**
```
或
```
==关键短语==
```
实现红色高亮强调。
项目符号需简洁；生成的项目符号幻灯片默认使用带编号的要点卡片。
若幻灯片仅有1-2个内容项目符号，则渲染为无编号的普通陈述内容。
过长的幻灯片和论文标题会自动调整：通过缩小标题字号并加宽行宽，确保标题完整显示在页面内。
内容稀疏的幻灯片会从标题和项目符号关键词中自动生成细微的主题图案，无需额外图片资源即可填充视觉空白区域。

Navigation Controls

导航控件

Generated decks support mouse click, left-third click for previous slide, explicit Prev/Next buttons, left/right arrow keys, space, Home/End, F fullscreen, and URL hash state such as

#/3

生成的幻灯片集支持以下操作：鼠标点击切换、点击页面左三分之一区域返回上一张幻灯片、点击显式的“上一页/下一页”按钮、左右方向键、空格键、Home/End键、F键全屏，以及URL哈希状态（如

#/3

）定位。

Design Rules

设计规则

Keep slides academic, readable, and presentation-oriented. Avoid long paragraphs. Prefer figures, concise numbered points, section breaks, and strong hierarchy. Use decorative but unobtrusive visual patterns to reduce empty space. Keep section and ending slides centered. Escape user-provided HTML by default. Do not write API keys or

.env

values into generated decks.

幻灯片需保持学术性、可读性和演示导向。避免冗长段落。优先使用图表、简洁的编号要点、章节分隔和清晰的层级结构。使用装饰性但不突兀的视觉图案减少空白区域。章节页和结尾页需居中显示。默认对用户提供的HTML进行转义处理。请勿将API密钥或

.env

文件中的值写入生成的幻灯片集。

Aut_Sci_Write Integration

与Aut_Sci_Write的集成

```
sci-html
```
is one of the suite's output skills alongside
```
sci-ppt
```
.
The paper workflow locally reuses Aut_Sci_Write extraction and figure-processing logic, so users do not need to manually chain
```
sci-extract
```
and
```
sci-figure
```
before generating a deck.
Keep generated decks under
```
out/
```
or another user-selected output folder. These outputs are intentionally ignored by the repository/package rules.
Do not create or require skill-local
```
.env
```
files for this skill; it runs offline except for optional dependencies installed by the user.

```
sci-html
```
是套件中与
```
sci-ppt
```
并列的输出技能之一。
论文处理流程本地复用Aut_Sci_Write的提取和图表处理逻辑，因此用户在生成幻灯片集前无需手动串联
```
sci-extract
```
和
```
sci-figure
```
技能。
将生成的幻灯片集保存在
```
out/
```
目录或用户选择的其他输出文件夹中。这些输出会被仓库/包规则主动忽略。
请勿为本技能创建或要求技能本地的
```
.env
```
文件；除用户安装的可选依赖外，本技能可离线运行。