A Grida slides deck is a .canvas bundle in slides mode whose documents are SVG files — one SVG per slide. Today this is Grida's only slide format: dotcanvas + SVG, and it is the default. When the task is a deck, presentation, pitch, or talk, build it this way. Do NOT author slides as markdown or HTML — the slides surface renders SVG only, so anything else won't appear.

Structure

• The bundle is a folder ending in .canvas with a .canvas.json manifest.
• Manifest: editor: "slides", files: ["*.svg"], and a documents array whose ORDER is the running order of the deck.

  {
    "editor": "slides",
    "files": ["*.svg"],
    "documents": [
      { "src": "001.svg", "id": "cover" },
      { "src": "002.svg", "id": "problem" }
    ]
  }

• Each slide's CONTENT is one SVG file referenced by src. The deck convention is flat, zero-padded, root-level files: 001.svg, 002.svg, ….
• A slide's human name is its SVG <title> element — not a manifest field.
• If you open an existing or freshly-seeded bundle whose manifest says editor: "board" but the task is a deck, set editor: "slides" yourself — the manifest is yours to reconcile with the user's intent.

Slide SVG

Every slide is a full-bleed 16:9 SVG on the SAME 1920×1080 viewBox, so the deck stays uniform. Start each from this shape:

<svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 1920 1080" width="1920" height="1080">
  <title>Cover</title>
  <rect width="1920" height="1080" fill="#0b1220"/>
  <text x="160" y="560" font-family="Inter, Arial, sans-serif" font-size="120" font-weight="700" fill="#ffffff">Your title</text>
</svg>

Working pattern

• To start a deck: first make the bundle folder — a NEW directory whose name ends in .canvas (e.g. Q3 Report.canvas/), created under your working root. The .canvas suffix on the FOLDER is what marks it a bundle (see Structure); a plain folder — or files loose in the workspace root — is NOT a deck and won't open. Then write_file each slide as <name>.canvas/NNN.svg and write_file the manifest as <name>.canvas/.canvas.json listing them in order. Every file lives INSIDE that one .canvas folder.
• To edit a deck: read_file .canvas.json first (preserve version / $schema and any unknown fields), edit the slide SVGs, then write the full manifest back with documents in the intended order.
• Reorder = reorder documents. Add a slide = write_file a new NNN.svg and insert it into documents. Remove = drop it from documents.

Show the result

• Treat presentation as the first production milestone, not the final "validate and open" step. For a new deck, create a meaningful valid first slide and a manifest that references it, then call surface_open immediately with the workspace-rooted .canvas bundle directory (for example, /Q3 Report.canvas). Continue adding and refining slides while the user can watch. Do not wait for the full deck, all assets, polish, preview generation, exhaustive validation, or task completion.
• Never open an empty bundle, a broken manifest, or a manifest whose first slide is missing. Pass the bundle directory, never its .canvas.json manifest.
• For an existing primary deck, open it after reading its manifest and before substantial edits.
• If the .canvas bundle itself is mounted as / in a dedicated file surface, it is already presented; do not call surface_open just to reopen it.
• Treat presentation as auxiliary: continue regardless of the result, never retry based on presentation status, and do not call surface_open after every write.
• Use surface_list_open only when the current host surface state is materially useful. It is not required before surface_open.

Starting from a template

When the user picks a template from the gallery, a <user_template_selection> block on the first turn names it (title, slide count, visual system), and its unzipped .canvas bundle (the manifest .canvas.json + slide SVGs) is placed in your scratch dir — a reference, like an attachment, NOT part of the user's workspace. Treat it as the STARTING POINT, not a fixed result:

• List your scratch dir and read_file the .canvas.json + slide SVGs there first. The template defines the deck's visual system — palette, type, layout, margins, footer, accents. Hold it: reuse those so every slide you write reads as a sibling, not a stray.
• Build the adapted deck in the workspace as a NEW <name>.canvas/ folder (that's where the user's document lives — scratch is throwaway; and the .canvas folder suffix is what makes it a deck — see Working pattern). Write the slide SVGs + .canvas.json INSIDE it, replacing the placeholder copy with the user's content and adding / reordering / removing slides (update documents) to fit. Keep the frame (1920×1080 viewBox, safe margins) and the conventions above.
• Don't discard the template's design and start from nothing unless the user asks for a different look — keeping that design is the whole point of the pick.

See also

• The svg skill — SVG authoring / output style (xmlns, formatting, recovery).
• The dotcanvas skill — the .canvas manifest and board mechanics this builds on.