drawio-skill

Why we included this project

When your agents need to produce diagrams that humans can actually read and edit, this skill handles it without you running a separate service. It installs as a single SKILL.md file for Claude Code, Cursor, Copilot, OpenClaw, and other Agent Skills environments, turning plain-language prompts into native .drawio XML and exporting PNG, SVG, PDF, or JPG through the draw.io desktop CLI. The real draw is the range of inputs: it can reverse-engineer a Python, JS/TS, Go, or Rust codebase into an architecture diagram, parse Terraform, Kubernetes, or docker-compose into resource diagrams with official cloud icons, and read SQL DDL into an ER model. Mermaid text gets converted into laid-out, editable diagrams too, and the tool can self-check its rendered output to fix overlaps and clipped labels. For teams that already standardize on draw.io for documentation, this is a practical way to keep architecture diagrams in sync with code and infrastructure instead of hand-maintaining them.