What problem does it solve? Manually drawing architecture and data flow diagrams is slow and error-prone, and keeping them in sync with a codebase is tedious. This Skill lets an AI agent place shapes, arrows, and annotations on a live Excalidraw canvas through MCP tools, producing accurate diagrams directly from components, sample mockups, or textual descriptions. ## Core Features & Use Cases - Batch Diagram Creation: Create shapes, arrows, zone backgrounds, and labels in a single batch_create_elements call with ID-based arrow binding and auto-routing. - Self-Critique Loop: Automatically snapshot, validate geometry and visuals via screenshots, fix overlaps or clipped labels, and roll back if fixes regress. - Multiple Export Formats: Export finished diagrams to PNG, SVG, editable .excalidraw JSON files, or shareable Excalidraw URLs. - Use Case: Ask the agent to trace a parameter through five architectural layers; it produces a data flow diagram with split/converge paths, a decision diamond, layer labels, and data-form annotations, then exports it as a PNG for your design doc. ## Quick Start Ask the agent to draw an architecture diagram of your system on the Excalidraw canvas, listing the components and how they connect.