← Back to Skills Marketplace
drawio-AI-maker
by
wuliwenjing
· GitHub ↗
· v1.0.2
· MIT-0
127
Downloads
1
Stars
0
Active Installs
3
Versions
Install in OpenClaw
/install drawio-ai-maker
Description
将自然语言或文档（docx/pdf/txt）转换为 draw.io 可编辑 XML 图表（.drawio 文件）。当用户说"画个流程图"、"生成图表"、"画架构图/时序图/网络拓扑图"、"帮我画xxx流程"、"生成draw.io"等时自动触发。使用流程：接收描述→生成JSON→用户确认→调用gen.py生成.dr...
README (SKILL.md)
drawio-generator
如何使用（标准流程）
Step 1：接收用户的流程描述（文字或文件上传）
Step 2：根据描述生成结构化 JSON（nodes + edges），将 JSON 展示给用户确认
Step 3：用户确认后，调用生成脚本：
python3 skills/drawio-generator/scripts/gen.py "图表标题" '{"title":"...","nodes":[...],"edges":[...]}' [类型]
Step 4：交付 .drawio 文件到输出目录，并告知用户：
用 draw.io 打开此文件（网页版 https://app.diagrams.net 或桌面版 App），如需微调请手动调整节点位置后保存。
⚠️ 生成结果非完美，可能存在线条重叠或间距不理想，需要在 draw.io 中手动微调。
输出目录：
- 默认：
/Users/owen/Desktop/drawio-generator/ - 可通过环境变量
DRAWIO_OUTPUT_DIR或命令行--output-dir自定义
JSON 结构
{
"title": "流程名称",
"type": "flowchart",
"nodes": [
{"id": "0", "type": "start", "label": "开始"},
{"id": "1", "type": "process", "label": "处理步骤"},
{"id": "2", "type": "decision", "label": "判断条件？"},
{"id": "3", "type": "end", "label": "结束"}
],
"edges": [
{"source": "0", "target": "1"},
{"source": "1", "target": "2"},
{"source": "2", "target": "3", "label": "是"},
{"source": "2", "target": "1", "label": "否"}
]
}
节点 type 值
| type | 形状 |
|---|---|
start |
椭圆 |
end |
椭圆 |
process |
圆角矩形 |
decision |
菱形 |
document |
文档形状 |
data |
平行四边形 |
edges.label 值
| label | 含义 |
|---|---|
是 / Y |
条件为真 |
否 / N |
条件为假 |
| 空 | 普通顺序流 |
支持的图表类型
flowchart | sequence | network | architecture | hierarchy | function | deployment
文件说明
| 文件 | 说明 |
|---|---|
scripts/gen.py |
标准生成脚本（必须使用，禁止直接调用 generator.py） |
scripts/generator.py |
底层 XML 生成器（gen.py 内部调用） |
scripts/parser.py |
输入解析（txt 直接读取，docx/pdf 需额外安装库） |
references/drawio-xml-spec.md |
draw.io XML 格式规范 |
依赖说明：
- 解析 .docx 需要：
pip install python-docx - 解析 .pdf 需要：
pip install pdfplumber
布局参数
| 参数 | 默认值 | 说明 |
|---|---|---|
CANVAS_WIDTH |
850 | 画布宽度 |
_LAYER_GAP_Y |
120 | 层间垂直间距 |
MAIN_X |
425 | 主轴 X 坐标 |
Usage Guidance
This skill appears to implement the described draw.io generator, but I found a few problems you should consider before installing or running it:
- Missing module: scripts/__init__.py imports renderer functions (render_to_png, render_to_svg, check_drawio_available) but there is no renderer.py in the package. This likely causes ImportError or runtime failures in some paths. Ask the author for the missing file or remove references.
- Hard-coded output path: the default output directory is /Users/owen/Desktop/drawio-generator/. If you run the skill without setting DRAWIO_OUTPUT_DIR or --output-dir it will attempt to write there. Override the directory explicitly and run in a safe directory (or a sandbox) to avoid unexpected file writes.
- Minor internal inconsistencies: comments and some schema/validator notes don't line up exactly with constants used in code (e.g., ID ranges and comment text). These are probably bugs, not malicious intent, but they can cause the tool to reject valid outputs or misbehave.
- File parsing dependencies: to parse .docx/.pdf you must install python-docx and pdfplumber locally; the code will raise ImportError if not present. That behavior is documented.
- No network/exfiltration seen: I did not find network calls, external endpoints, or credential access. The skill asks the host agent to call an LLM (MiniMax) to produce JSON — make sure you trust the agent/LLM and are comfortable that any sensitive diagram content would be sent to it.
Recommendations: run this in a sandbox or isolated environment first; set DRAWIO_OUTPUT_DIR to a directory you control; inspect/fix the missing renderer dependency before relying on it; and if you plan to process sensitive documents, ensure the LLM invocation path (the 'main agent' call) conforms to your privacy requirements.
Capability Analysis
Type: OpenClaw Skill
Name: drawio-ai-maker
Version: 1.0.2
The drawio-generator skill bundle is a legitimate tool designed to convert natural language descriptions or documents (PDF/DOCX) into draw.io XML diagrams. The code implements a sophisticated multi-step process: parsing input via 'parser.py', generating a structured layout using a custom engine in 'generator.py', and performing rigorous validation of both the intermediate JSON and the final XML in 'validator.py'. The skill uses a standard OpenClaw pattern where it provides prompts for the main agent to execute via an LLM. No evidence of data exfiltration, unauthorized execution, or malicious prompt injection was found; notably, the developer even implemented shell-escaping logic in '__main__.py' to safely handle command-line retries.
Capability Assessment
Purpose & Capability
Name/description (natural language/doc → draw.io XML) aligns with included modules (parser, designer, generator, validator, gen.py). The code implements JSON→XML conversion, layout, validation and file output which are expected for this functionality.
Instruction Scope
SKILL.md and __main__.py restrict actions to: accept text/files, ask host agent/LLM to produce JSON, validate JSON/XML, write a .drawio file to an output directory. The runtime instructions do not attempt to read unrelated secrets or send data to external endpoints. Note: default output directory is a hard-coded absolute path (/Users/owen/Desktop/drawio-generator/) which will write to the user's filesystem unless overridden.
Install Mechanism
No install spec; the skill is instruction + code only. There are optional Python library dependencies (python-docx, pdfplumber) declared in SKILL.md for docx/pdf parsing — these are reasonable and documented. No downloads or network installers are present.
Credentials
The skill does not request credentials or environment variables except an optional DRAWIO_OUTPUT_DIR to override the default output directory. That is proportionate. However the default hard-coded path targets a specific user's Desktop (suggests packaging from a developer machine) — this is a usability/incoherence concern (may create files in unexpected locations).
Persistence & Privilege
always is false and the skill does not declare or attempt to persist or modify other skills or system-wide agent settings. It only writes output files to a directory (expected for this functionality).
How to Use
- Make sure OpenClaw is installed (local or Docker)
- Run the install command in chat:
/install drawio-ai-maker - After installation, invoke the skill by name or use
/drawio-ai-maker - Provide required inputs per the skill's parameter spec and get structured output
Version History
v1.0.2
- Added support for direct import of docx, pdf, and txt files for flow description input.
- Output directory for generated .drawio files can now be customized via the DRAWIO_OUTPUT_DIR environment variable or --output-dir command line option.
- Improved documentation and usage instructions.
- Detailed dependency requirements for file parsing are now listed in the documentation.
v1.0.1
- Improved trigger description in the skill metadata for better automatic activation based on user intent.
- Clarified and streamlined the usage instructions, now presented in steps for easier understanding.
- Condensed technical explanations and highlighted the key user workflow with clearer callouts.
- Expanded and clarified the table for supported diagram types.
- Removed outdated details about the legacy execution pipeline.
- Updated descriptions for edge labels, node types, and usage of draw.io for manual adjustment.
v1.0.0
Initial release of drawio-generator:
- Converts natural language process descriptions or documents into editable draw.io (XML) diagrams (.drawio files).
- Supports flowcharts, architecture diagrams, sequence diagrams, and network topology diagrams based on user input.
- Follows a structured flow: receive input → generate structured JSON → get user confirmation → generate .drawio file.
- Informs users to open and manually adjust .drawio files with the draw.io web or desktop app.
- Includes clear documentation for JSON structure, element types, and layout parameters.
Metadata
Frequently Asked Questions
What is drawio-AI-maker?
将自然语言或文档（docx/pdf/txt）转换为 draw.io 可编辑 XML 图表（.drawio 文件）。当用户说"画个流程图"、"生成图表"、"画架构图/时序图/网络拓扑图"、"帮我画xxx流程"、"生成draw.io"等时自动触发。使用流程：接收描述→生成JSON→用户确认→调用gen.py生成.dr... It is an AI Agent Skill for Claude Code / OpenClaw, with 127 downloads so far.
How do I install drawio-AI-maker?
Run "/install drawio-ai-maker" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.
Is drawio-AI-maker free?
Yes, drawio-AI-maker is completely free, licensed under MIT-0. You can download, install and use it at no cost.
Which platforms does drawio-AI-maker support?
drawio-AI-maker is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).
Who created drawio-AI-maker?
It is built and maintained by wuliwenjing (@wuliwenjing); the current version is v1.0.2.
More Skills