What problem does it solve? Hand-authoring tldraw .tldr files is error-prone because raw records skip the editor's default fill-in and fail schema validation, while reading a messy human whiteboard back into structured data requires parsing ProseMirror richText and arrow bindings by hand. ## Core Features & Use Cases - Deterministic canvas writing: Create .tldr files and add boxes, ellipses, text, sticky notes, frames, and bound arrows via a spec JSON that clears tldraw's own validator (pinned to tldraw 5.2.5 schema). - Canvas reading and organizing: Inspect an existing board as structured JSON (shapes, text, positions, edges), then reseat shapes, add frames and arrows, and group scattered notes without altering user-authored text. - Use Case: Ask for a three-stage pipeline sketched as a hand-drawn diagram; the skill drafts a spec, runs Tldr.ts create/add/validate, and returns a .tldr file that opens cleanly in the VS Code tldraw extension, tldraw.com, or the desktop app. ## Quick Start Ask the agent to sketch your described diagram as a hand-drawn tldraw canvas and save it to a .tldr file you can open in any tldraw surface.