Skip to content
Back to skills

Canvas Boards

ASecurity

Draw clear, editable canvases in Daedalus (flows, plans, architectures, decision trees, mind maps) with the Canvas tools, and link the operator to the board.

  • 2 stars
  • 0 votes
  • 0 copies
  • 0 views
  • Added November 11, 2026
ai-agentsgonodeawsapi

Works with

  • api

Security analysis

A100/100

Scanned November 11, 2026

npx -y skills add anchor-inference/daedalus --skill canvas-boards --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Canvas Boards?

Add the live security badge to your README. It updates with every re-scan.

Security grade badge for Canvas Boards
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/anchor-inference-canvas-boards/badge)](https://www.skillsdirectory.com/skills/anchor-inference-canvas-boards)

More formats (shields.io, HTML) on the badges page. Keep it an A: scan every change in CI with Pro.

Download with Pro
SKILL.md
---
name: canvas-boards
description: Draw clear, editable canvases in Daedalus (flows, plans, architectures, decision trees, mind maps) with the Canvas tools, and link the operator to the board.
requires: canvas
---
# Canvas boards

A canvas is a living board the operator opens at `/app/canvas/<id>` and edits by hand: cards with
Markdown, groups, links and arrows between card sides, in Obsidian's JSON Canvas format. It is not a
picture. The tools compute card sizes, placement and edge sides; you decide the content.

## Workflow

1. **Lay out first.** Anything with flow or structure (steps, components, a decision tree) goes
   through `CanvasCreate` with `nodes` and `edges`, or `CanvasEdit` with `layout_graph` to add a laid-out
   part beside an existing board. Do not place cards by coordinates unless asked.
2. **Tweak with `CanvasEdit`:** change a card's words (`{id, text}`), recolour it (`{id, color}`), add
   a card (it goes right of the board, sized for its words), connect two cards (`edges: [{from, to,
   label}]`), group cards (a `group` node whose box surrounds them), delete (`delete_ids`; a card's
   edges go with it, a group's contents stay).
3. **Read before every edit.** `CanvasRead` gives the version, the ids, places and what each group
   holds. Pass that `version`. The operator may be moving cards right now: your edit still saves when
   it touches other items, and is refused only when the operator changed the same item; then read
   again and redo the change on what is there.
4. **Link it.** Finish with a markdown link in this form:
   `[Payment flow](/app/canvas/0123456789abcdef0123456789abcdef)`. The id is 32 hex characters; the
   chat turns that form into an in-app link.

A message from the canvas's comment bar names the canvas and its id, and may quote an element of a
frame the operator pointed at: read the canvas and act on it there.

## Content

- **Short card text:** a bold first line, detail below (`"**Validate order**\nschema + stock"`).
  Cards are Markdown: lists, links and code work.
- **At most about 15 cards** per board section. Split a large system into groups, or an overview
  board and one board per part.
- **One direction per flow:** `right` for processes and pipelines, `down` for hierarchies and
  decision trees. A loop back is fine; label it.
- **Edge labels** only where they carry meaning: the branch of a decision (`yes` / `no`), a protocol,
  a data name.
- **Stable, readable ids** (`api`, `queue`, `worker`), so later edits can name them.

## Colours

`color` is a preset `"1"` to `"6"` (red, orange, yellow, green, cyan, purple), which each app draws in
its own shade, or `#rrggbb`. Use a handful, consistently: green for the happy path's end, red for
failure or risk, yellow for a decision or something needing attention, cyan or purple for extra groups
when a legend card says what they mean. Leave most cards uncoloured.

## Scope

A canvas belongs to this chat (the default) or to its project (`scope: "project"`), where every chat of
the project sees it. Choose the project for a board the whole project works from (a roadmap, an
architecture); the chat for anything that belongs to this conversation alone.

Attribution

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.

View all in ai-agents →

Comments

Loading comments…

Newsletter

Get the weekly list of safe new skills, and the ones we pulled.