How to use the Diagram Generator agent skill

What an agent skill is and what this one makes the agent do

An agent skill is a folder containing a SKILL.md file with instructions, plus optional reference files and scripts that an AI agent can read and use when a task matches the skill's description. This skill is for diagrams. Once it is installed, the agent can generate, refine, validate, and render diagrams from natural language, notes, code snippets, schemas, tables, or existing diagram source.

By default the agent writes the diagram as text source in Mermaid, which you can paste into Markdown or edit later. For UML-heavy engineering diagrams it can use PlantUML, for dense graph layouts Graphviz DOT, and for custom visuals SVG. Supported diagram types include flowcharts, swimlanes, sequence diagrams, state diagrams, ER diagrams, class diagrams, architecture (C4-style) diagrams, dependency graphs, Gantt charts, mind maps, user journeys, sankey-style flows, org charts, and network graphs.

Install the skill from the ZIP

Download the ZIP of the skill from this page and unzip it. Then place the diagram-generator folder in one of these locations, depending on how you use your agent:

Put it in ~/.claude/skills/ to make it available across your projects, or in a project's .claude/skills/ folder to limit it to that project.

In the Claude apps you can instead upload the ZIP under Settings → Capabilities → Skills.

• Personal folder: ~/.claude/skills/
• Project folder: .claude/skills/ inside your project
• Claude apps: Settings → Capabilities → Skills, then upload the ZIP

Ask for a diagram in natural language

You do not need a special command. Describe what you want and what material to use. The skill's description covers flowcharts, sequence diagrams, ER diagrams, Gantt charts, and similar types, so naming the diagram type helps the agent pick the right language. If you do not name one, it chooses based on your request.

Good requests include the source material and the audience. Mention whether you want source only or a rendered file. The agent will normally return editable source and a short list of assumptions it made. It only creates image or PDF files when you ask for them or when a downloadable file would clearly help.

• Turn these meeting notes into a flowchart for the approval process.
• Draw an ER diagram from this SQL schema.
• Make a Gantt chart from this table of tasks and dates.
• Render this architecture description as an SVG I can put in a slide.

Render to PNG, SVG or PDF (optional)

When you request an image, the agent creates a source file and runs the bundled script scripts/render_diagram.py. The script tries common local command-line tools and reports installation hints if one is missing. Rendering depends on Mermaid CLI, Graphviz, or PlantUML being installed on your machine, which you must do separately. Python 3 is needed to run the bundled scripts.

The agent should only report that an image was rendered when the script finished successfully and the output file exists.

• Mermaid source (.mmd) rendered with Mermaid CLI
• Graphviz source (.dot) for dense graph layouts
• PlantUML source (.puml) for UML diagrams

Safety notes: scripts and network

This skill runs scripts. The source states that it runs scripts and does not use the network. The bundled renderer calls local command-line tools on your machine, and the sample generator script is also included in the package. Read the scripts in the scripts/ folder before running them if you are working with sensitive projects, and keep rendering inside a directory you control.

Because the skill does not use the network, the source indicates the generated diagrams are produced locally. Your notes or schemas are still sent to the AI model you use in the normal way when you ask for a diagram.

• Runs scripts: yes (scripts/render_diagram.py and scripts/create_sample_diagrams.py)
• Uses network: no
• Review scripts before running them on confidential material

Tips for better results

Give the agent the messy input as it is. It normalizes entities, relationships, and order itself. If something is missing, it labels its assumptions, so check those before using the diagram.

Ask for source you can edit later when you plan to keep the diagram in documentation. Ask for an image only when you need a file. If you want a left-to-right pipeline or a swimlane, say so; the skill defaults to top-down flowcharts unless asked otherwise.

• Paste the raw notes, table, or schema rather than a summary
• Name the diagram type if you know it
• Ask for assumptions to be listed when the input is vague
• Keep one diagram per request for clearer output