What problem does it solve? Editing tldraw offline canvases by hand-placing shapes is slow and error-prone for an agent, and adding durable interactive behavior to a drawing requires knowing the app's document-script contract and local control API. This Skill teaches the agent to read and modify the open canvas through the app's localhost HTTP API and to write JavaScript document scripts that persist inside the .tldraw file. ## Core Features & Use Cases - Live canvas edits via /exec: Generate shapes, arrows, and layouts on the focused document using curl calls against the local API with the bearer token from server.json. - Durable document scripts: Edit script/main.js so reactive, animated, or clickable behavior (like a counter with MINUS/RESET/PLUS buttons) survives reloads and ships inside the .tldraw file. - Schema validation: Run scripts/validate_shapes.mjs to verify raw note/text/frame records against the real tldraw SDK v5 schema before building file snapshots. - Use Case: Ask the agent to turn an open tldraw document into an interactive wireframe with working buttons; it writes an idempotent document script, confirms script-status shows applied, and verifies clicks by dispatching pointer events. ## Quick Start Open tldraw offline with a document, then ask the agent to build a labeled architecture diagram with connected nodes on the current canvas.
