axton-obsidian-visual-skills: Claude Code Skills That Write Obsidian Diagrams

What the Visual Skills Pack actually solves

Drawing a diagram by hand in Obsidian is slow, and pasting model output into a canvas usually means fixing JSON by hand. This project sits in that gap. It ships three Claude Code skills, each a set of instructions that Claude loads when a prompt matches, and each one ends in a file that Obsidian can open: an Excalidraw drawing, a Mermaid diagram, or a .canvas board.

The audience is narrow and specific. You need Claude Code installed, and you need Obsidian. The README lists the Excalidraw plugin as a prerequisite for the Excalidraw skill only. If you write in Obsidian and already pay for Claude Code, the marginal cost of trying this is one marketplace command. If you use neither tool, nothing here applies to you.

The author is direct about scope. The README carries an Experimental badge and a Status section stating that this is a public prototype that works for the author's demos but does not cover all input scales and edge cases. That sentence should shape how you read everything else on this page.

Three skills, three output formats, one prompt interface

Skills in Claude Code are markdown files loaded on demand, which the README contrasts with MCP servers that need complex setup. There is no daemon and no server process. You type a request, Claude matches it against the skill descriptions, loads the relevant instructions, and writes a file.

The Excalidraw skill has three output modes. Obsidian is the default and produces a .md file that opens directly in Obsidian. Standard produces a .excalidraw file for excalidraw.com. Animated also produces a .excalidraw file, intended for the excalidraw-animate tool. The README lists eight diagram types for this skill, including flowchart, mind map, hierarchy, relationship, comparison, timeline, matrix and freeform, and notes hand-drawn styling with Excalifont at fontFamily 5 plus full Chinese text handling.

The Mermaid skill covers process flow, circular flow, comparison, mindmap, sequence and state diagrams. Its distinguishing feature is stated as built-in syntax error prevention for common pitfalls: list conflicts, subgraph naming and special characters. Anyone who has debugged a Mermaid parse error inside a note will recognize why that matters.

The Canvas skill emits JSON Canvas with two layout modes. MindMap arranges nodes radially from a center; Freeform allows custom positioning and connections. The README mentions smart node sizing based on content length, automatic labeled edges, six preset colors plus custom hex values, spacing rules to prevent overlap, and group nodes.

Installing the pack and generating your first Mermaid diagram

The README gives two installation paths. The plugin marketplace is marked Recommended. It is two slash commands, and the README says to restart Claude Code afterward so the skills load.

/plugin marketplace add axtonliu/axton-obsidian-visual-skills
/plugin install obsidian-visual-skills

If you prefer to keep the files visible, the manual route clones the repository and copies the three skill folders into your Claude Code skills directory. The README shows all three copies; you can copy only the ones you want.

# Clone the repository
git clone https://github.com/axtonliu/axton-obsidian-visual-skills.git

# Copy skills to Claude Code directory
cp -r axton-obsidian-visual-skills/excalidraw-diagram ~/.claude/skills/
cp -r axton-obsidian-visual-skills/mermaid-visualizer ~/.claude/skills/
cp -r axton-obsidian-visual-skills/obsidian-canvas-creator ~/.claude/skills/

After a restart, the skills activate on trigger words rather than a command. The README lists the Mermaid triggers as Mermaid, visualize, flowchart, sequence diagram and the Chinese term 可视化. A prompt such as "Visualize this process as a Mermaid diagram" is the documented shape of a request. You should end up with a diagram block in your note that Obsidian renders, and the README states the output is compatible with Obsidian, GitHub and other Mermaid renderers.

For Canvas, the trigger words are Canvas, mind map, visual diagram and 思维导图, and the documented example is "Turn this article into an Obsidian Canvas". The result is a .canvas file that opens directly in Obsidian. For Excalidraw, the default trigger set is Excalidraw, diagram, flowchart, mind map, 画图 and 流程图; the README notes that adding the word 标准 or standard switches to the .excalidraw output, and 动画 or animate switches to the animated variant.

Where the pack breaks down

The README does not hide the failure modes. Its Status section says output quality varies based on model version and input structure, and that results may fluctuate. That is a real constraint for anything you intend to keep: the same prompt can produce a different layout after a model update, and there is no version pinning described anywhere in the README.

Scale is the other boundary. The status note says the project does not yet cover all input scales and edge cases. Long articles pushed into the Canvas skill are exactly the case where node sizing and spacing heuristics would be stressed, and the README does not document what happens when the content exceeds what the layout can hold.

The maintenance stance deserves a plain reading. The author writes that the primary focus is demonstrating how tools and systems work together, not maintaining this codebase. The repository is not archived and the last push was on 2026-06-26, so it has not been abandoned, but the stated intent is demonstration rather than support. Issue reports are requested as reproducible cases with input, output file and steps. That is a reasonable bar, and it also means vague bug reports are unlikely to go anywhere.

Finally, the Excalidraw skill depends on a third-party Obsidian plugin for its default output. The Mermaid and Canvas skills do not carry that prerequisite, which makes them the lower-friction half of the pack.

How this differs from a general diagramming skill set

The closest comparison in the same search space is Kepano's obsidian-skills repository, which also appears in related searches alongside this project. The difference is the target format. Kepano's set is built around Obsidian's own conventions and markdown-first output. This pack is built around three external formats that Obsidian happens to render: Excalidraw, Mermaid and JSON Canvas.

That distinction decides which one you want. If your notes are markdown with embedded diagrams, the Mermaid skill here overlaps heavily with any general diagramming skill and the choice comes down to prompt phrasing and syntax-error handling. If you want editable hand-drawn drawings or interactive canvas boards, this pack covers ground a markdown-only skill set does not, because .excalidraw and .canvas are file formats with their own structure, not markdown blocks.

The three output modes for Excalidraw are the sharpest differentiator. A general skill that emits Mermaid has no equivalent of writing a .excalidraw file for excalidraw.com or an animated variant for excalidraw-animate. Whether that matters depends entirely on whether you share drawings outside Obsidian.

Licence, releases and what upgrades cost you

The project is MIT licensed, and the LICENSE file sits at the repository root. MIT permits commercial use and modification with attribution; it also means no warranty accompanies the code. Nothing in the README suggests any additional terms, and this is not legal advice. If you redistribute the skills inside a product, read the licence text rather than this summary.

Releases are visible and small in number. v0.1.0 was the initial release on 2026-01-16. v1.2.0 added three output modes and marketplace support on 2026-02-02. v1.2.1, described as enhanced design rules, followed on 2026-02-12. The last push to the repository was on 2026-06-26, roughly four months after that release.

Upgrade cost is low in the mechanical sense: skills are copied folders or marketplace installs, with no compiled artifacts and no configuration files to migrate. The cost that does exist is behavioural. Because the skills are prompts rather than code, a new release can change how diagrams are laid out without any API change to notice. If you have a note whose diagram you liked, keep the generated file; regenerating it after an upgrade is not guaranteed to reproduce the same result, and the README's own warning about fluctuating output is the reason.

Editorial conclusion

Adopt it if you already run Claude Code and keep notes in Obsidian, because the skills are markdown files that load on demand and the Canvas and Mermaid outputs open without extra plugins. Do not adopt it if you need a stable, documented interface: the README labels the project experimental, says output quality varies with model version and input structure, and states that maintaining the codebase is not the author's primary focus. Verify first that the three skill folders land in your skills directory, that the Excalidraw plugin is installed if you want the default .md output, and that a small test prompt produces a file Obsidian actually opens before you point it at a large document.

Frequently asked questions

Official sources

1. axtonliu/axton-obsidian-visual-skills on GitHub
2. Issues
3. License: MIT
4. README
5. Releases