wtf is drawio-skill?

drawio-skill is a tool that lets an AI coding assistant turn a plain-English description into a professional diagram, using the free diagramming app draw.io as the engine that actually renders it. You describe what you want, such as a microservices architecture with a gateway, a few backend services, a message queue, and some databases, and it produces an editable draw.io file that can also be exported as a PNG, SVG, PDF, or JPG image. It supports eleven preset diagram types, including flowcharts, UML class and sequence diagrams, entity relationship diagrams, network diagrams, and the C4 model used for describing software architecture at different zoom levels. Beyond writing diagrams from scratch, it can read an existing codebase written in Python, JavaScript, TypeScript, Go, or Rust and lay out its structure automatically, and it can turn infrastructure files such as Terraform, Kubernetes, or docker-compose into an architecture diagram where each resource is drawn with its real cloud provider icon. It can also convert database table definitions into an entity relationship diagram with keys and relationships marked, and it understands diagrams written in the popular Mermaid text format, converting them into fully editable draw.io files. The tool works as a skill, a packaged set of instructions, for AI coding assistants like Claude Code, Cursor, Copilot, and several others that support the shared Agent Skills format. It has a self-check step where it looks at its own rendered image and automatically fixes problems like overlapping shapes or clipped text, and it can go through several rounds of refinement based on feedback. It also draws from a searchable library of over ten thousand official shape icons for major cloud providers and diagramming standards, so diagrams look accurate rather than generic. To use it, you first install the draw.io desktop application, which provides the command line tool that actually generates image files, and then install the skill itself into your AI assistant. The project is released under the MIT license, which allows free use for any purpose, including commercial projects.