Skip to content

Canvas Design skill

Anthropic's skill for original posters and static art: it writes a design philosophy first, then renders it as a PNG or PDF.

Official · Anthropic
AuthorAnthropic
LicenceApache-2.0
Sourceanthropics/skills · SKILL.md on GitHub
Works inClaude Code · claude.ai · Claude API · Codex, Gemini CLI, Cursor, Copilot
Read from the source on

The canvas-design skill is Anthropic's official Claude skill for posters, covers and other single-page art. Instead of drawing straight away, it has Claude write a short design philosophy, an imaginary art movement with a name, and then express that philosophy on a canvas that you get back as a PNG or a PDF.

It is part of Anthropic's public skills repository (Apache-2.0) and ships with 54 open-licence font files, so the type on the poster is chosen rather than left to whatever is installed. It suits visual pieces with little text; for slides, documents or brand-accurate layouts, other skills fit better.

From the skill's SKILL.md: Create one single page, highly visual, design-forward PDF or PNG output (unless asked for more pages).

This is an official skill from Anthropic. Like any skill it gives the agent instructions and may include scripts it runs, so install it from the source linked here, not from a copy elsewhere.

What it does

  1. Reads your request as a light starting point, not a layout to copy.
  2. Names a visual movement in one or two words and writes its philosophy (four to six paragraphs on space, colour, scale, rhythm and hierarchy) as a Markdown file.
  3. Finds a subtle reference to your topic and hides it in the composition rather than spelling it out.
  4. Draws the canvas with code, picking fonts from the bundled canvas-fonts folder, keeping every element inside the margins.
  5. Takes a second refining pass on the same composition instead of adding more elements, and can extend it to several pages on request.

What you get

  • One design philosophy as a .md file.
  • The artwork as a single-page .png or .pdf (more pages, bundled in one PDF or several PNGs, only if you ask).

When to use it

  • You want a poster, cover, flyer or art print where the image carries the message.
  • You want something that looks designed, with a deliberate palette and type, rather than a generic AI picture.
  • You want a series of related pages that share one visual language.

When not to use it

  • The piece is mostly text (a report, a menu with prices, a CV): text is kept to a few words by design.
  • You need exact brand colours and logos: give it your brand rules, or pair it with a brand skill such as brand-guidelines.
  • You need an editable vector file for print production: the output is PNG or PDF, not an .ai, .fig or layered file.
  • You want a photo-realistic image: it composes shapes, patterns and type in code; it does not generate photographs.

What it needs

  • An agent that can run code and write files: Claude Code, claude.ai with code execution on, or the Claude API with the code execution tool.
  • Nothing to install for the fonts: they ship inside the skill folder (about 5.5 MB in total).
  • No account beyond the one you use for Claude; no API keys.

Files in the skill

File or folderWhat it is for
SKILL.mdThe instructions: the two-step method (philosophy, then canvas), craft rules and the multi-page option.
canvas-fonts/54 .ttf font files (families such as Big Shoulders, Bricolage Grotesque, Crimson Pro, IBM Plex Mono, Young Serif), each with its Open Font License text.
LICENSE.txtApache License 2.0.

Licence as found in the repository on 2026-11-14: Apache-2.0.

Example prompts

Type one of these after installing; naming the skill makes sure it is used.

  • Use the canvas-design skill to make an A3 poster for a jazz night called "Late Set", Friday 21:00, Hall B. PNG, mostly visual, only those words.
  • Create a book cover for a short-story collection about tide pools. Give me the design philosophy first, then a PDF.
  • Make a three-page PDF art series on the theme of city noise at night, each page a variation of the same visual language.
  • Design a minimalist concert poster in black, off-white and one accent colour. Use a condensed font from the bundled fonts and keep generous margins.

How to install canvas-design

Installed with Anthropic's example skills

In Claude Code, canvas-design comes inside the example-skills plugin of Anthropic's marketplace, which installs eleven other example skills with it (a plugin installs whole). If you want only this one, copy the folder instead (the skills-folder steps below).

On claude.ai it is one of the example skills Anthropic already provides: switch it on under Customize > Skills; there is nothing to upload.

On the Claude API

The folder is about 5.5 MB with its fonts, well under the 30 MB upload limit of the Skills API. The API sandbox has no internet access, so the skill can use only the bundled fonts there, not download others.

A skill is a folder with a SKILL.md file. You install it once; the agent reads its short description at the start of each session and loads the full instructions only when a request matches. Pick the place you use:

Claude Code: from the plugin marketplace

  1. Start Claude Code in a terminal with claude.
  2. Add the marketplace (once per computer):
    /plugin marketplace add anthropics/skills
  3. Install the plugin. Claude Code opens its details first: check what it will add (skills, hooks, MCP servers), then choose whether it is for you, for everyone on this repository, or for you in this repository only.
    /plugin install example-skills@anthropic-agent-skills
  4. Type / and look for /example-skills:canvas-design, or simply ask for the task.

From your shell instead of a session:

claude plugin marketplace add anthropics/skills
claude plugin install example-skills@anthropic-agent-skills

The plugin installs as a whole, so other skills come with it, including algorithmic-art, brand-guidelines, doc-coauthoring, frontend-design, internal-comms, mcp-builder, skill-creator, slack-gif-creator, theme-factory, web-artifacts-builder, webapp-testing.

Claude Code: copy the folder into your skills folder

Put the skill's folder in ~/.claude/skills/ to use it in every project on this computer, or in .claude/skills/ inside a repository to share it with everyone who works there (commit it).

git clone --depth 1 https://github.com/anthropics/skills.git
mkdir -p ~/.claude/skills
cp -R skills/skills/canvas-design ~/.claude/skills/

Claude Code notices new or changed skills in these folders during the session. If the skills folder did not exist when the session started, run /reload-skills once. Then type /canvas-design or just describe the task.

On Windows, run the commands in Git Bash or WSL, or copy the folder into the same place with File Explorer.

claude.ai

canvas-design is one of the example skills Anthropic already provides, so there is nothing to upload: check that Code execution and file creation is on under Settings > Capabilities, then switch canvas-design on under Customize > Skills.

Plans: Anthropic's Help Center lists skills on Free, Pro, Max, Ultra, Team and Enterprise plans; its developer docs list uploading custom skills on Pro, Max, Ultra, Team and Enterprise. On Team plans skills are on by default; on Enterprise an owner must allow code execution and Skills under Organization settings > Plugins & skills. Skills you enable on claude.ai also load in Claude Code when you sign in with the same account.

We could not find steps for the Claude desktop app in Anthropic's skills documentation, so they are not listed here. Skills uploaded to claude.ai are not copied to the API, and API skills are not copied to claude.ai.

Claude API

Upload the folder once with the Skills API (as a ZIP; the total must be under 30 MB uncompressed), then list its id in the container of each Messages request together with the code execution tool. Uploaded skills are shared with your whole workspace.

git clone --depth 1 https://github.com/anthropics/skills.git
cd skills/skills
zip -r canvas-design.zip canvas-design
import anthropic

client = anthropic.Anthropic()
skill = client.skills.create(files=[open("canvas-design.zip", "rb")])

response = client.messages.create(
    model="claude-opus-5-5",
    max_tokens=4096,
    container={"skills": [{"type": "custom", "skill_id": skill.id, "version": "latest"}]},
    tools=[{"type": "code_execution_20250825", "name": "code_execution"}],
    messages=[{"role": "user", "content": "Use the canvas-design skill."}],
)

Skills on the API run in a sandbox with no internet access and no package installation, so a skill that needs a web service or an MCP server will not work there.

Codex, Gemini CLI, Cursor and GitHub Copilot

These agents read the same SKILL.md folders. The one folder all four accept inside a project is .agents/skills/, so this copies the skill there:

git clone --depth 1 https://github.com/anthropics/skills.git
mkdir -p .agents/skills
cp -R skills/skills/canvas-design .agents/skills/
AgentProject folderYour user folderNotes
OpenAI Codex
.agents/skills/
~/.agents/skills/
Type $ or run /skills to pick a skill; Codex also chooses one when your task matches.
Gemini CLI
.agents/skills/
.gemini/skills/
~/.agents/skills/
~/.gemini/skills/
Or install from Git with gemini skills install <git URL> --path <folder> --consent. Google replaced Gemini CLI with Antigravity CLI for free-tier and Google One users on 18 June 2026; Antigravity uses .agents/skills/ in a project and ~/.gemini/config/skills/ for you.
Cursor
.agents/skills/
.cursor/skills/
~/.agents/skills/
~/.cursor/skills/
Type / in Agent chat to pick a skill. Cursor also reads Claude Code's .claude/skills/ folders.
GitHub Copilot
.github/skills/
.agents/skills/
.claude/skills/
~/.copilot/skills/
~/.agents/skills/
Works in Copilot CLI, the cloud agent, and agent mode in VS Code, Visual Studio and JetBrains (IDE support depends on the version). Or run gh skill install OWNER/REPOSITORY SKILL (public preview, GitHub CLI 2.90.0 or later).
Only install skills you trust. Anthropic's advice is to use skills only from trusted sources, and to read every file of any other skill before using it: a skill can direct the agent to run commands, read files or send data in ways its description does not mention.

Sources for these steps

Checked against each vendor's documentation on 2026-11-14:

How to get better results

Give it a mood and a few hard facts, not a layout. "A calm poster for a ceramics open studio, Saturday 10–4" gives the skill room to choose; a pixel-by-pixel description works against its method.

If the first canvas is close, ask it to refine rather than add: the skill's own final step is to polish the existing composition instead of piling on new shapes, and it does that best when you name what feels off (spacing, contrast, one colour too many).

Name the size you need (A3, 1080 × 1350, Letter) and the format. Without that it makes one page at a size it picks.

Questions about canvas-design

Is the canvas-design skill free?

The skill itself is free and open source under the Apache 2.0 licence. Running it uses your normal Claude plan or API usage, like any other request; it adds no separate fee.

What file formats does canvas-design produce?

A PNG or a PDF for the artwork, plus a Markdown file with the design philosophy it followed. Several pages come as one multi-page PDF or several PNGs, but only when you ask for more than one page.

Why does it write a "design philosophy" before the poster?

The skill's method is to invent a small art movement first and then express it visually, so the result follows one consistent idea about space, colour and type instead of a pile of effects. You can read the philosophy, change it, and ask for the canvas again.

Can it put a lot of text on the poster?

Not by design: it treats text as a visual accent and keeps it to a few essential words. For text-heavy layouts such as menus, reports or slides, ask for a document instead or use a skill made for documents.

Can I use the posters commercially?

The skill's licence (Apache 2.0) covers the skill's files; what you make with it is governed by your Claude terms of use. The bundled fonts are under the SIL Open Font License, which allows use in artwork. The skill tells Claude to create original work and not copy existing artists.

Does it work outside Claude?

It is a standard SKILL.md folder, so agents that read Agent Skills folders can load it (see the install guide below). It assumes the agent can run code to draw the canvas and write PNG or PDF files.

More AI skills

Related skills

Skill categories