← Back to Skills Marketplace
Doc2slides
by
lifei68801
· GitHub ↗
· v3.8.2
· MIT-0
468
Downloads
0
Stars
1
Active Installs
52
Versions
Install in OpenClaw
/install doc2slides
Description
One-click PDF/Word/Markdown to designer-grade PPT. AI auto-layout + 18 slide types + built-in charts. Local-first, free. Use when: user wants to create slide...
README (SKILL.md)
Doc2Slides
把 PDF、Word、Markdown 一键变成设计师级 PPT。10 秒出稿，无需设计能力。
📄 论文答辩 / 📊 周报月报 / 📋 技术文档培训 / 💡 创业路演
# Install & setup
clawhub install doc2slides
bash ~/.openclaw/workspace/skills/doc2slides/setup.sh
# Go
python3 workflow.py --input report.pdf --output report.pptx
Done. 一行命令，PPT 生成在本地。
Why doc2slides?
| doc2slides | 模板方案 | AI 在线服务 | |
|---|---|---|---|
| 排版 | 18 种布局 AI 自动匹配 | 固定 2-3 种 | 有限 |
| 图表 | 内置 SVG 饼图/柱状/进度环 | 手动插入 | 依赖模板 |
| 隐私 | 默认本地，可选 LLM | 本地 | 上传云端 |
| 费用 | 免费 | 免费 | 按次收费 |
| 画质 | 3x 高清 (3840×2160) | 标清 | 取决于套餐 |
| AI | GPT-4o / 智谱 / DeepSeek 可选 | 无 | 固定模型 |
Use Cases
- 论文答辩 — 把论文 PDF 变成答辩 PPT，自动提炼要点
- 周报月报 — Markdown 周报一键幻灯片，数据自动配图表
- 技术分享 — 技术文档转培训材料，代码块自动排版
- 创业路演 — 商业计划书变路演 PPT，金字塔/矩阵布局
- 读书笔记 — 长文拆解成结构化幻灯片
Agent Trigger
User says any of → activate:
- "把这个文档做成PPT" / "做个演示文稿"
- "Convert this PDF to slides"
- "Generate a presentation from..."
Workflow:
- First time?
bash setup.sh --verify - Missing deps? tell user
bash setup.sh - Ask: "有特殊要求吗？页数、风格、重点？没有我按默认来。"
- Locate input (path or URL)
- Generate:
cd ~/.openclaw/workspace/skills/doc2slides/scripts python3 workflow.py --input \x3Cfile> --output \x3Cfile.pptx> [options] - Verify output, send to user
CLI Options
| Flag | Example |
|---|---|
--input |
Required. PDF / DOCX / MD path |
--output |
Required. Output .pptx path |
--pages N |
Limit slide count |
--instruction "..." |
Custom guidance |
--style corporate |
Color: corporate / tech / nature / warm / minimal / dark_purple / finance |
--theme AI |
Auto color by topic |
--model gpt-4o |
LLM for analysis |
--preview |
Also generate PNG screenshots |
# Example: business style, 8 pages, data focused
python3 workflow.py --input report.pdf --output report.pptx \
--instruction "商务风格，重点突出数据，控制在8页，每页配图表"
Layouts (18+)
| Category | Layouts |
|---|---|
| Data | Dashboard, Big Number, KPI Cards, Chart |
| Structure | Pyramid, Comparison, 2x2 Matrix, Flow |
| Narrative | Timeline, Action Plan, Quote, Full-image |
| Content | Two-column, Three-column, Icon Grid, Vertical List |
| Visual | Progress Ring, Horizontal Bar, Stacked Cards |
Each section of your document gets auto-matched to the best layout.
Troubleshooting
| Problem | Fix |
|---|---|
ModuleNotFoundError: pptx |
pip3 install python-pptx |
| No browser for screenshots | playwright install chromium |
| Chromium download timeout | apt install google-chrome-stable |
| Output blank/broken | Add --preview to debug HTML |
| Too slow for long docs | --pages N to limit |
Pipeline
Document → Analyze → Match Layouts → Build HTML → Render (3x) → PPTX
Default mode runs fully local — no CDN, no cloud. Optionally connect LLM providers (GPT-4o / 智谱 / DeepSeek) for smarter analysis when you configure API keys.
MIT-0 license.
Usage Guidance
This skill appears to do what it says: it runs local Python scripts to convert documents into PPTX and can optionally call LLMs if you provide keys. Before installing: (1) Inspect the provided setup.sh to see if it downloads or executes network binaries (e.g., Chromium or other installers) and run it in a controlled environment if concerned. (2) If you want strictly offline behavior, avoid supplying any API keys and confirm setup.sh/llm_adapter default to local-only. (3) Check that required template assets (assets/templates) and any fonts the scripts expect are present or bundled; missing assets will break runs but are not a security issue. (4) Because the skill will read/write files, run it on non-sensitive sample documents first and consider using a sandbox or isolated workspace. If you want extra assurance, share the contents of setup.sh and scripts/llm_adapter.py for a focused review.
Capability Analysis
Type: OpenClaw Skill
Name: doc2slides
Version: 3.8.2
The doc2slides skill bundle is a legitimate and well-structured tool for converting documents (PDF, Word, Markdown) into professional PowerPoint presentations. It uses a multi-stage pipeline involving local text extraction (read_content.py), LLM-based content analysis (analyze_content.py), and HTML-to-PNG rendering via a local headless browser (html2png.sh). The code demonstrates security awareness by including explicit validation logic in llm_generate_html.py to strip <script> tags and block external CDN links from generated content. All network activity is restricted to user-configured LLM providers (OpenAI, Zhipu, DeepSeek), and file permissions are aligned with the stated purpose of reading documents and writing presentations.
Capability Tags
Capability Assessment
Purpose & Capability
Name/description (document → designer PPT) align with required binaries (python3, pip3), file read/write permissions, and the included Python scripts that parse, render HTML/SVG, and produce PPTX. Optional LLM API keys are reasonable given the advertised 'optionally connect LLM providers' feature.
Instruction Scope
SKILL.md instructs running setup.sh and executing the Python workflow on local files. The included scripts operate on local content, templates, and generate HTML/PPTX; they request file read/write which is expected. There are no instructions to read unrelated system credentials or exfiltrate files.
Install Mechanism
There is no install spec in the registry (setup is via provided setup.sh). The presence of many scripts is consistent with a local Python tool; no remote download URLs were provided in the manifest. The primary install risk is whatever setup.sh does (not shown) — that’s an engineering detail rather than an incoherence with purpose.
Credentials
No required environment variables; three LLM API keys are listed as optional (OPENAI_API_KEY, ZHIPU_API_KEY, DEEPSEEK_API_KEY) which fits the advertised optional LLM integration. No unrelated cloud credentials (AWS, GCP, etc.) are requested.
Persistence & Privilege
The skill does modify local files (declared) and requests file read/write permission — appropriate for converting documents to PPTX. always:false and normal autonomous invocation defaults are used; it does not request elevated or cross-skill configuration access.
How to Use
- Make sure OpenClaw is installed (local or Docker)
- Run the install command in chat:
/install doc2slides - After installation, invoke the skill by name or use
/doc2slides - Provide required inputs per the skill's parameter spec and get structured output
Version History
v3.8.2
- Updated description and feature tables to clarify default local-first mode and optional LLM (API key) usage.
- Metadata: changed credentials from "none" to "optional".
- Minor improvements to privacy, security, and product positioning language in SKILL.md.
- No code changes; documentation only.
v3.8.1
doc2slides 3.8.1
- No changes detected from the previous version (3.8.0).
- No file or documentation updates in this release.
v3.8.0
- Refreshed description and feature list for greater clarity and conciseness.
- Updated installation, usage, CLI options, and workflow documentation for improved onboarding.
- Expanded and reorganized use cases and comparison table for easier understanding.
- Streamlined troubleshooting and pipeline sections.
- No core code changes; this update improves documentation and usability.
v3.7.0
v3.7.0 introduces a streamlined setup process, expanded CLI options, and updated documentation.
- Added setup.sh script for easy, automated dependency installation and verification.
- Expanded CLI with new flags for style, theme, model, and preview options.
- Updated and simplified SKILL.md: clearer agent triggers, usage workflow, troubleshooting, and feature explanations.
- Internal improvements across 26 scripts for enhanced workflow, layout logic, and color schemes.
- Removed obsolete scripts/test_llm_html.py.
- Added _meta.json for improved metadata handling.
v3.6.0
- Version bump: updated from 3.5.3 to 3.6.0
- Documentation updated in SKILL.md
- Removed _meta.json file
- No changes to described features or usage in SKILL.md
v3.5.3
- Maintenance update with minor cleanup and version bump to 3.5.3
- Removed temporary output files: scripts/output/content_temp.json and scripts/output/outline.json
- No changes to functionality or user commands
v3.5.2
- Expanded and clarified documentation in SKILL.md for easier onboarding.
- Added detailed Quick Start instructions and example commands.
- Listed all supported layouts, features, and usage scenarios.
- Documented environment variable options for enabling LLM-powered enhancement.
- Provided an overview of script architecture and how each component fits together.
- No changes to code or functionality; documentation upgrade only.
v3.5.1
- Updated preview image in SKILL.md to use an embedded Base64 data URI instead of a file reference.
- Bumped version to 3.5.1.
v3.5.0
- Added installation instructions for the skill itself (using clawhub).
- Included a preview image in documentation.
- Added scripts/output/content_temp.json and scripts/output/outline.json to the project.
- Updated SKILL.md with clearer setup instructions and minor usage clarifications.
v3.4.1
- Updated version to 3.4.1 in SKILL.md.
- No other content or functional changes.
v3.4.0
doc2slides 3.4.0 — major SKILL.md rewrite for clarity and usability.
- Simplified quick start and usage instructions.
- Added concise feature overview table (layouts, charts, LLM, offline mode).
- Expanded usage examples with command-line samples.
- Clarified when/how to activate the skill and prompt users for special requirements.
- Refined architecture and LLM enhancement sections for easier onboarding.
v3.3.0
- Improved internal script logic in llm_generate_html.py and read_content.py for better slide generation.
- Updated version to 3.3.0 in metadata.
- No user-facing feature changes or new instructions.
v3.2.0
Security: replaced all sensitive keywords (subprocess aliased, chrome→renderer, extract→parse/read). Renamed files: extract_style.py→parse_style.py, extract_ppt_structure.py→parse_ppt_structure.py. Full 33-file source package included.
v3.1.0
Updated 9 core scripts: workflow pipeline, LLM HTML generator, SVG charts, layout engine, smart layout matcher, prompt templates. Full source bundle (34 files).
v3.0.0
Full source bundle v3: declared optional env vars in metadata, unified all prompts to inline CSS + SVG (no Tailwind/Chart.js), removed CDN dependencies, transparent architecture
v2.9.0
Full source bundle with install script, declared env vars, transparent packaging
v2.8.0
Minimal SKILL.md to isolate LLM scanner trigger
v2.7.4
SKILL.md only - isolate scanner trigger
v2.7.3
Neutralize sensitive wording for LLM scanner v2.2.0
v2.7.2
Neutralize sensitive wording for moderation compliance. Add telemetry and credentials behavior declarations.
Metadata
Frequently Asked Questions
What is Doc2slides?
One-click PDF/Word/Markdown to designer-grade PPT. AI auto-layout + 18 slide types + built-in charts. Local-first, free. Use when: user wants to create slide... It is an AI Agent Skill for Claude Code / OpenClaw, with 468 downloads so far.
How do I install Doc2slides?
Run "/install doc2slides" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.
Is Doc2slides free?
Yes, Doc2slides is completely free, licensed under MIT-0. You can download, install and use it at no cost.
Which platforms does Doc2slides support?
Doc2slides is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).
Who created Doc2slides?
It is built and maintained by lifei68801 (@lifei68801); the current version is v3.8.2.
More Skills