The canvas-design skill: posters and art as PDF or PNG

Source: anthropics/skills · Maintainer: Anthropic · License: Apache 2.0 · Checked: 24 October 2026

What is the canvas-design skill?

The canvas-design skill has an agent make a static visual piece, such as a poster or an art print, in two steps: it writes a short design philosophy as Markdown, then renders that philosophy as a single-page PDF or PNG. It suits developers and makers who want a striking one-off image from Claude Code or Claude.ai rather than a layout template.

Its description in the SKILL.md reads: "Create beautiful visual art in .png and .pdf documents using design philosophy."

It lives in Anthropic's anthropics/skills repo, which showed 183.3K GitHub stars on its skills.sh listing. The skill's skills.sh page listed 121.5K installs (snapshot, 24 October 2026) and files it under Design & UI.

What does the canvas-design skill do?

The description tells the agent to use it "when the user asks to create a poster, piece of art, design, or other static piece." It also says to create original work and never copy existing artists. Once loaded, the SKILL.md runs a fixed sequence:

1. Design philosophy. The agent names an "aesthetic movement" in one or two words and writes four to six paragraphs on how it shows up in space, form, color, scale, rhythm and hierarchy. The file gives condensed examples such as "Concrete Poetry," "Chromatic Language" and "Geometric Silence." The philosophy is saved as a .md file.
2. A subtle reference. Before drawing, the agent works out a quiet link to your original request and weaves it into the piece without stating it. The SKILL.md compares this to "a jazz musician quoting another song."
3. The canvas. The agent renders one page as a PDF or PNG: "90% visual design, 10% essential text." It favors repeating patterns, a limited palette, thin type and strict margins, with nothing overlapping or falling off the page.
4. A refinement pass. The last step tells the agent to polish what exists instead of adding more shapes or effects.

If you ask for more pages, it treats the first as one page of a book and makes the rest as variations, bundled in one PDF or as several PNGs.

What is in the folder?

• SKILL.md: the instructions above. No scripts.
• canvas-fonts/: 54 .ttf files, including IBM Plex Mono and Serif, JetBrains Mono, Instrument Sans and Serif, Lora, Work Sans, Crimson Pro and Big Shoulders, plus 27 OFL license texts (related families such as IBM Plex Mono and Serif share one).
• LICENSE.txt: Apache 2.0.

The SKILL.md tells the agent to "Search the ./canvas-fonts directory" when it sets text, and to download other fonts if the design needs them.

How do you install the canvas-design skill?

Three documented routes. Pick one per agent.

Claude Code plugin. The anthropics/skills README registers the repo as a plugin marketplace. In a Claude Code session:

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

The example-skills plugin installs the whole example set (12 skills in the repo's marketplace file, including theme-factory, brand-guidelines and frontend-design), not just this one. See Claude Code plugins for how plugins work.

The skills CLI. To install only this skill, use the command from its skills.sh page:

npx skills add https://github.com/anthropics/skills --skill canvas-design

The skills CLI README says it installs into the current project by default; add -g for all projects or -a claude-code to target Claude Code. The folder lands in .claude/skills/canvas-design/ or, globally, ~/.claude/skills/canvas-design/, the standard Claude Code skill locations. Codex and Cursor read the same SKILL.md format; see Codex skills and Cursor skills.

Claude.ai. The README says the example skills "are all already available to paid plans in Claude.ai." To add a copy yourself, Anthropic's help article on using skills in Claude says to package the folder as a ZIP and upload it under Customize, Skills, with code execution turned on. Claude.ai is a natural fit here, since the result is a downloadable file.

How do you use it?

You don't need to name the skill; ask for a poster or art piece and the agent should pick it up. In Claude Code you can also call it directly: /canvas-design if you used the skills CLI, or /example-skills:canvas-design with the plugin, since plugin skills are namespaced. Some prompts that fit what it does (suggestions, not quotes from the repo):

• "Make a poster for a jazz night at a small bar. Dark, sparse, one bold word."
• "Create an abstract art print about long-distance running, as a PNG."
• "Design a cover image for our engineering blog post on database migrations."
• "Make a three-page PDF art book about tide tables, same philosophy on each page, different compositions."
• "Create a minimalist poster for a ceramics studio open day. Thin type, lots of space."

Keep requests short. The skill treats your words as a starting point, not a spec: the SKILL.md says user input "should not constrain creative freedom." If you need exact text, dates or a logo placed precisely, say so plainly and check the result.

Requirements

The SKILL.md names no library. The agent writes its own drawing code, so it needs a code execution environment that can produce PDF or PNG files: Claude.ai with code execution on, or a local agent with a language and imaging tools installed. Fonts come from the bundled folder; downloading extra fonts needs network access.

Tips and limits

• Expect art, not layouts. It is built for "art objects, not documents with decoration." For slides or documents with a consistent look, theme-factory or pptx fit better.
• Text is minimal by design. If you need paragraphs, prices or a full event schedule, this skill will fight you. Ask for the key words only, or use another tool for text-heavy pieces.
• It refines rather than adds. The final step tells the agent to polish the existing composition. If you want a bigger change, ask for a new philosophy instead.
• It runs code it writes. There are no bundled scripts, but the agent writes and runs rendering code and may download fonts. Review those steps as you would any agent shell command.
• Licenses. The skill is Apache 2.0. The fonts are separate works under the Open Font License texts in canvas-fonts/. Anthropic's README calls the example skills "provided for demonstration and educational purposes only."
• Audits. As of 24 October 2026, the skills.sh page showed Pass from all three security audits listed (Gen Agent Trust Hub, Socket and Snyk).

Save it in your Kipt library

If you keep a set of skills you rely on, Kipt holds them in one place. Open Skills in Kipt and paste the skill's GitHub folder link into the import box; Kipt reads the SKILL.md and its files, and for a repo with many skills it shows a list to pick from. You can also drop the folder or a .zip. Your skills live in your own private GitHub repo and install into Claude Code or Codex with one command:

npx skills add https://github.com/<you>/<repo>.git --all -g

Kipt is not affiliated with Anthropic.

Similar skills

• theme-factory skill: ten ready color and font themes for slides, docs and pages.
• brand-guidelines skill: apply a fixed brand palette and type instead of inventing one.
• frontend-design skill: Anthropic's design guidance for interfaces rather than posters.
• web-artifacts-builder skill: build interactive HTML artifacts when a static image isn't enough.