This is a single skill, canvas-design. It tells an AI agent how to make original static visual art, such as posters and art pieces, as .png or .pdf files. The agent first writes a design philosophy as a .md file and then expresses it on a canvas. It is meant for people who want an agent to produce art-quality visual work.
Why read it
Before making anything, the agent writes a 4-6 paragraph design philosophy and saves it as a .md file. That document then becomes the brief for the canvas stage. Settling the aesthetic direction in words first makes the visual stage follow a consistent vision, and the reasoning is saved alongside the art.
The skill gives five named example philosophies, such as "Concrete Poetry" and "Chromatic Language". Each pairs a one-line philosophy with a description of how it looks. A note after them says these are shorter than the real output should be. This shows the agent the style while still keeping the length requirement clear.
The skill tells the agent to repeat phrases like "meticulously crafted" inside the philosophy it writes. The skill itself repeats craftsmanship language in several sections. This keeps the quality bar in view throughout a long generation.
The final step assumes the user has already said the work is not good enough. It then limits what the revision can do: refine what exists rather than add more. A STOP question redirects the urge to draw new shapes, which helps keep the second pass from cluttering the piece.
To explain how the user's topic should show up in the art, the skill compares it to a jazz musician quoting another song. One image like this tells the agent how hidden the reference should be more clearly than an abstract rule would.
The skill gives wide creative freedom, but a few rules are marked non-negotiable: no overflow, no overlap, proper margins. Making these specific failures non-negotiable gives the agent a concrete checklist even inside an open-ended task.
Text rules bend to context. The skill uses a punk venue poster and a minimalist ceramics studio to show how type size should change, and gives a default ("Most of the time, font should be thin"). It also points the agent to a local fonts directory to use.
A model read each skill and found no calls or hand-offs between them.
| SKILL | KIND | STARTED BY | DESCRIPTION | LINKS | LICENCE |
|---|---|---|---|---|---|
| canvas-design | skill | You or the agent | Create beautiful visual art in .png and .pdf documents using design philosophy. You should use this skill when the user asks to create a poster, piece of art, design, or other static piece. Create original visual designs, never copying existing artists' work to avoid copyright violations. | 0 | Apache-2.0 |
From anthropics/skills at 2c5958e, 2026-11-06. Skills read by Claude Sonnet 5.5 on 2026-11-07. Summary and techniques by Claude Opus 5.5 on 2026-11-07.