canvas-hand

v0.1.0

A local infinite canvas for Codex, powered by tldraw — generate images onto the canvas, annotate with arrows, and let the AI redraw images from your annotations.

Codex1 skill1 MCP serverstdio

by changcheng192updated 1 month ago

Source

$ git clone https://github.com/Tasihi89/canvas-hand

Clone the source, then follow the repository's marketplace instructions for your runtime. The repository root is the plugin root.

Layout

canvas-hand/

├── .codex-plugin/plugin.json
├── skills/canvas-hand/SKILL.md
└── .mcp.json

Skills1

canvas-handskills/canvas-hand/SKILL.md

Operate "my canvas" (a local infinite canvas) for image work — open the canvas, generate new images onto it, and redraw images based on arrow annotations drawn on the canvas. Use when the user wants to open the canvas, generate an image onto it, or modify/iterate an image already on the canvas (even without saying "canvas" explicitly — phrases like "change the background", "redraw this", "tweak this image" count too).

MCP servers1

canvas-handstdio

command: bash
args: ./scripts/start-mcp.sh
cwd: .

Manifests1

.codex-plugin/plugin.json

{
  "name": "canvas-hand",
  "version": "0.1.0",
  "description": "A local infinite canvas for Codex, powered by tldraw — generate images onto the canvas, annotate with arrows, and let the AI redraw images from your annotations.",
  "author": {
    "name": "changcheng"
  },
  "keywords": [
    "canvas",
    "tldraw",
    "whiteboard",
    "image-generation",
    "annotation"
  ],
  "skills": "./skills/",
  "mcpServers": "./.mcp.json",
  "interface": {
    "displayName": "Canvas Hand",
    "shortDescription": "Create, annotate, and generate images on a local infinite canvas in Codex.",
    "longDescription": "Canvas Hand brings a tldraw-powered infinite canvas into Codex. Open a local canvas, generate AI images onto it, draw arrow annotations with instructions, and let the AI read your annotations and redraw the image accordingly — all kept in your project workspace.",
    "developerName": "changcheng",
    "category": "Productivity",
    "capabilities": [
      "Interactive canvas",
      "Local web service",
      "AI image generation",
      "Annotation-driven editing",
      "Project-local persistence"
    ],
    "defaultPrompt": [
      "Open my canvas",
      "Generate an image on my canvas",
      "Redraw this image based on my annotations on the canvas"
    ]
  }
}