Canvas User Docs

Guide for pages under docs/editor/ — user-facing documentation for the Grida Canvas editor at grida.co/docs.

Not for: docs/wg/, docs/reference/, forms, CLI, or architecture docs.

Directory & Build

Item	Value
Source	docs/editor/
Site	Docusaurus in apps/docs/
Product	Canvas editor at http://localhost:3000/canvas
Build	turbo build --filter=editor && turbo start --filter=editor

See docs/AGENTS.md for linking rules and MDX caveats.

Writing

• Direct, second-person, present-tense. "You can resize by dragging the handle."
• Neutral. No hype, no marketing, no exclamation marks.
• Concise. One idea per sentence. Short paragraphs.
• UI labels in bold. Shortcuts in <kbd> tags. Values/filenames in backticks.
• Links to editor pages use universal routing: https://grida.co/_/<path>.
• Add format: md frontmatter unless the page uses JSX/MDX features.
• No bare angle brackets in prose (MDX compatibility).

Page Structure

---
title: Feature Name
description: One-line summary.
format: md
---

# Feature Name

Brief intro (1-2 sentences). What is it, why use it?

## How to Use / Creating a \_\_\_ / When to Use

Primary workflow or step-by-step.

## How It Differs from X (if applicable)

Comparison table.

## Nesting / Hierarchy Rules (if applicable)

## Default Appearance (if applicable)

## Keyboard Shortcuts (if applicable)

Headings: ## for sections, ### for subsections. No ####+.

Screenshots

Property	Value
Output size	960 x 960 px (1:1)
Format	WebP q90 (preferred), PNG fallback
Theme	Default light
Naming	<feature>-<description>.webp (kebab-case)
Location	Co-located with the doc or sibling img/

960x960 is the cropped output, not the viewport. Capture at any size, then crop to the area of interest and resize. Crop aggressively — a single panel or dialog is better than a full-window capture.

Preparing Demo State

Prefer scripting over manual interaction, in this order:

1. Fixture file. Load a .grida or .svg from examples/fixtures/ or test/.
2. Scripting API. globalThis.grida exposes the editor instance (via WindowGlobalCurrentEditorProvider). Use javascript_tool or console to create/modify nodes, set state, rename layers programmatically.
3. URL parameters. Playground accepts ?src=<url>, ?backend=canvas|dom.
4. Manual interaction. Fallback only.

Before capture: name layers meaningfully, deselect (<kbd>Escape</kbd>), zoom to fit. Use realistic labels ("Login", "Dashboard") — never lorem ipsum.

Requirements

• Production build (localhost), not dev mode
• No browser chrome, system notifications, or debug panels
• Show hover/focus states only when they're the subject

Custom Graphics

For workflows, before/after, or relationships a screenshot can't show.

• 960 x 960 px (1:1) or 1280 x 720 px (16:9)
• WebP preferred, SVG for vector diagrams
• Annotations: red (#FF3B30) circles/arrows, 2px stroke, 1-3 callouts max

For SVG figures specifically — gestures, alignment, before/after diagrams that recreate canvas UI — use the docs-svg-kit skill. It provides a starter template, a primitives catalog (selection chrome, size badges, anchor pins, resize cursors, ripples), and finished examples to crib from. Saves you from re-deriving stroke widths, colors, and the half-pixel alignment tricks.

Pre-Publish Checklist

• title and description in frontmatter
• format: md (unless using MDX)
• Screenshots are 960x960 WebP, cropped to subject, with alt text
• No broken links, no bare angle brackets
• Page renders: pnpm --filter docs start

Pitfalls

• Writing for yourself. Explain what things do, not how they're implemented.
• Screenshot overload. One cropped shot beats three full-page captures.
• Stale screenshots. When updating a feature, verify screenshots match current UI.
• Wrong doc type. Architecture and design decisions belong in docs/wg/.