canvas

Visual layer of the wiki. Add images, text cards, PDFs, wiki pages to infinite visual boards. Read _shared/canvas-spec.md before editing.

I/O

• Input: Canvas name, node type, content path/URL.
• Output: Modified/created .canvas file.

Process

1. Open: /canvas → list existing canvases. /canvas new [name] → create slug, write wiki/canvases/<slug>.canvas, append to dashboard. See references/operations.md § Status & Create.
2. Add: Select node type per references/operations.md § Add to Canvas — image, text, pdf, note, zone, list, from-banana. Dimensions and colors per references/node-templates.md.
3. Position: Per references/operations.md § Auto-positioning. Read canvas, find target zone, flow left-to-right with row wrapping. Auto-detect aspect ratio for images. Report position and zone.
4. Commit: Parse JSON, append new node, write atomically. Update wiki/index.md for new canvases only.

Rules

• Read canvas before every write. Never reuse IDs — pattern: [type]-[content-slug]-[unix-ts].
• Write tool is permitted only for *.canvas and _attachments/** paths (per _shared/vault-ops.md §5 bypass list). All other vault writes use obsidian CLI.
• Banana integration: after /banana, offer "Add generated images to canvas?".