/canvas Skill
SkillFiles & storageCreate and edit Obsidian Canvas files for architecture visualization
Available today. Use it from your connected AI after setup.
No other account needed.
Connect ahel once, and every AI you use reads what you have installed.
Then ask your AI: use the /canvas Skill skill
What this skill tells your AI
The instructions your AI receives, as published by davidroliverba/architectkb in .claude/skills/canvas/SKILL.md and read by ahel’s review.
Create and edit Obsidian Canvas files (.canvas JSON format) for architecture visualization.
When to Use This Skill
Use when creating or editing .canvas files (Obsidian Canvas format):
- Create architecture context diagrams (C4 Level 1)
- Design system landscape maps
- Visualize data flow architectures
- Show AWS infrastructure layouts
- Compare scenarios side-by-side
- Map system dependencies
⚠️ CRITICAL: Linter Configuration Warning
The Obsidian linter will corrupt .canvas files if enabled. If you have a linter active (prettier, eslint, etc.):
- DISABLE it for
.canvasfiles - Or DISABLE it globally during canvas editing
- Canvas files use a specific JSON format that lingers may reformat incorrectly
The linter will delete nodes with empty text fields and reduce complex diagrams to 2 nodes.
Canvas JSON Structure
Correct Format
Every node MUST use text property for content (NOT label):
{
"nodes": [
{
"id": "unique-id",
"type": "text",
"x": 0,
"y": 0,
"width": 200,
"height": 100,
"color": "3",
"text": "Node Content\nMultiline supported\nWith \\n character"
}
],
"edges": [
{
"id": "edge-id",
"fromNode": "node-id-1",
"toNode": "node-id-2",
"fromSide": "right",
"toSide": "left",
"label": "Edge Label"
}
],
"metadata": {
"version": "1.0-1.0",
"frontmatter": {}
}
}
Property Reference
Node Properties:
| Property | Required | Type | Values | Notes |
|---|---|---|---|---|
id | ✅ | string | Any unique ID | Use kebab-case |
type | ✅ | string | "text" | Currently only text type supported |
x | ✅ | number | Any integer | Canvas X coordinate |
y | ✅ | number | Any integer | Canvas Y coordinate |
width | ✅ | number | Any positive int | Node width in pixels |
height | ✅ | number | Any positive int | Node height in pixels |
text | ✅ | string | Any text | USE THIS FOR NODE CONTENT (not label) |
color | ❌ | string | "1"-"6" | Color: 1=red, 2=orange, 3=yellow, 4=purple, 5=cyan, 6=green |
Edge Properties:
| Property | Required | Type | Values |
|---|---|---|---|
id | ✅ | string | Any unique ID |
fromNode | ✅ | string | ID of source node |
toNode | ✅ | string | ID of target node |
fromSide | ✅ | string | "top", "right", "bottom", "left" |
toSide | ✅ | string | "top", "right", "bottom", "left" |
label | ❌ | string | Any text |
Metadata:
"metadata": {
"version": "1.0-1.0",
"frontmatter": {}
}
Common Mistakes
❌ WRONG: Using label for node content
{"id":"node1","type":"text","x":0,"y":0,"width":100,"height":50,"label":"Node Text"}
Result: Text appears on edges, boxes are empty
✅ CORRECT: Using text for node content
{"id":"node1","type":"text","x":0,"y":0,"width":100,"height":50,"text":"Node Text"}
Result: Text displays inside the box
❌ WRONG: Empty text field
{"id":"node1","type":"text","x":0,"y":0,"width":100,"height":50,"text":""}
Result: Obsidian deletes the node on next edit
✅ CORRECT: No empty fields
Only include properties with values. Don't include text field if empty.
Workflow
Phase 1: Define Canvas Structure
Plan the canvas:
- Sections/clusters - Group related nodes (e.g., "SOURCE SYSTEMS", "KAFKA EVENT BUS")
- Nodes - Individual boxes with content
- Edges - Connections with labels
Example structure:
SECTION HEADERS (120px high, wide)
↓
MAIN NODES (100+ px high, sized to fit content)
↓
EDGES with labels connecting them
Phase 2: Calculate Coordinates
- X axis: Left to right (0 → +)
- Y axis: Top to bottom (0 → +)
- Group related items with consistent spacing
- Use 20-30px padding between nodes
Example grid:
(0,0) (300,0) (600,0) (900,0)
A B C D
(0,100) (300,100) (600,100) (900,100)
E F G H
Phase 3: Determine Node Heights
Size nodes based on content:
- Section headers: 40px height
- Simple labels: 50px height
- Multi-line (2-3 lines): 70px height
- Multi-line (4+ lines): 80-100px height
- Large info boxes: 100px+ height
Add ~10-15px padding for text overflow.
Phase 4: Create JSON
Use text property for all node content. Use label property for edge labels only.
Phase 5: Validate & Test
- Save
.canvasfile - Verify JSON syntax:
node -e "JSON.parse(require('fs').readFileSync('file.canvas'))" - Open in Obsidian and verify:
- All boxes display with text inside (not on edges)
- Edge labels appear on connections
- Layout matches intended design
- Colours display correctly
Examples from This Vault
All of these use correct JSON Canvas format:
Canvas - C4 Context Diagram- 9 nodes, 9 edgesCanvas - System Landscape- 18 nodes, 14 edgesCanvas - Data Flow Diagram- 22 nodes, 18 edgesCanvas - AWS Architecture- 24 nodes, 11 edgesCanvas - Scenario Comparison- 19 nodes, 2 edgesCanvas - Data Platform Data Flow- 38 nodes, 68 edges
All use text property for node content and label property for edge labels.
Color Reference
Canvas supports 6 colors:
- "1" - Red (Critical, Production)
- "2" - Orange (High Priority, Warning)
- "3" - Yellow (Source Systems, Starting Point)
- "4" - Purple (Access/Gateway, Tools)
- "5" - Cyan (Transformation, Processing)
- "6" - Green (Analytics, Destination, Data Lake)
Size Guidelines
Width:
- Section headers: 160-200px (full width of section)
- Small nodes: 100-120px
- Medium nodes: 150-160px
- Large nodes: 200-260px
- Full-width containers: 240-300px+
Height:
- Thin header: 40px
- Simple label: 50px
- 2-3 lines: 70px
- 4-5 lines: 80px
- 5+ lines or containers: 100px+
Testing Checklist
Before considering a canvas file "done":
- JSON syntax is valid
- All node content uses
textproperty - All edge labels use
labelproperty - No empty
textfields exist - File opens in Obsidian without errors
- All boxes display text inside (not on edges)
- All edge labels appear on connections
- Colours display correctly
- Layout matches intended design
- Spacing and alignment looks good
Related Skills
/diagram- Generate diagrams programmatically using Python/scenario-compare- Create side-by-side scenario comparison canvases/impact-analysis- Visualize system impact on canvas
File Naming Convention
Canvas - {{Diagram Name}}.canvas
Canvas - C4 Context Diagram.canvas
Canvas - System Landscape.canvas
Canvas - Data Flow Diagram.canvas
Canvas - AWS Architecture.canvas
JSON Validation
Quick validation command:
node -e "try { JSON.parse(require('fs').readFileSync('file.canvas', 'utf8')); console.log('✓ Valid JSON'); } catch(e) { console.log('✗ Error:', e.message); }"
Key Principle: text = node content, label = edge labels. Never use label for node content.
Key Warning: Disable linters before editing .canvas files - they corrupt the JSON structure.
Signals
- GitHub stars
- 52
- Forks
- 12
- Last commit
- Mar 2026
Advanced
- Catalog kind
- skill
- Gateway key
canvas-davidroliverba- Source
- github.com/davidroliverba/architectkb