drawio-skill

Tracked

Generate draw.io diagrams from natural language with 6 presets, vision self-check, multi-round refinement, and 10,000+ official shapes including 321 AI/LLM brand logos.

Author Agents365-ai Open Sourced 2026-03-03 Last Commit Unknown

Overview

drawio-skill is a Claude Code skill that generates professional draw.io diagrams from natural language descriptions. It features 6 preset diagram types, a vision-based self-check mechanism, up to 5-round automatic refinement, codebase-to-diagram conversion, and a library of 10,000+ official shapes including 321 AI/LLM brand logos for consistent visual branding.

Key Features

  • 6 Diagram Presets: Architecture, flowchart, UML, ERD, sequence, and network diagrams with optimized layout defaults
  • Vision Self-Check + Refinement: Automatically validates generated diagrams visually and iterates up to 5 rounds for quality
  • Codebase-to-Diagram: Analyzes source code repositories and generates architectural diagrams automatically
  • 10,000+ Shapes Library: Includes official draw.io shapes and 321 AI/LLM brand logos for industry-standard visuals
  • Multi-Format Export: PNG, SVG, PDF, and JPG output formats

Use Cases

  • System Design Documentation: Generate architecture diagrams from natural language descriptions during design sessions
  • Codebase Visualization: Automatically produce architectural overviews from existing code repositories
  • Technical Documentation: Create consistent, branded diagrams for blog posts, presentations, and internal docs

Technical Details

  • Built as a Claude Code skill using the skill-md specification compatible with SkillsMP and OpenClaw ecosystems
  • Uses vision-based feedback loop to validate diagram correctness before final output