plantuml-diagrams
Skill0 starsChanged 6 months agoMITView on GitHub
Use when creating or refining PlantUML block, architecture, component, deployment, state, or flow diagrams where rendered readability and maintainable text sources both matter.
In this repository5 other files it gives its agents.
- anti-ai-tellsSkill
- contract-auditSkill
- devils-advocate-loopSkill
- karpathy-guidelinesSkill
- spec-to-shipSkill
The file
We split this skill into the rules it gives an agent
12 rules. About 0 are common, 7 are distinctive, and 5 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈7
Distinctive · estimate
This author's own choices. Worth a look.
5
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 8
Create diagrams that communicate before they decorate.
- Distinctive: line 12
Prefer PlantUML when the diagram should be diffable, reproducible, and maintained beside code or specifications.
- Distinctive: line 16
Before writing syntax, inventory:
- Very specific: line 29
Default to left to right direction for short pipelines and system boundaries.
- Distinctive: line 30
Keep the primary path obvious.
- Very specific: line 33
Avoid deeply nested containers until a flat version has rendered well.
- Distinctive: line 35
Use restrained semantic fills.
- Distinctive: line 36
Keep edge labels short. Move explanations into nodes or surrounding prose.
- Very specific: line 37
Use typography inside labels: <b> for the scan path, <i> for supporting context, and <color:#hex> for identifiers or state accents.
- Very specific: line 38
Use \n for deliberate line breaks.
- Distinctive: line 39
Avoid shadows, decorative gradients, and large icon libraries unless the user requests them.
- Distinctive: line 53
Use the installed CLI or JAR and enable Java headless mode when no display server is available:
≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
PlantUML Diagrams
Create diagrams that communicate before they decorate. A .puml file parsing is necessary but not sufficient: render it and inspect the picture.
Choose PlantUML deliberately
Prefer PlantUML when the diagram should be diffable, reproducible, and maintained beside code or specifications. Prefer a manual vector editor when exact placement or brand-level composition dominates. Prefer Mermaid when native Markdown embedding matters more than layout control.
Build from semantics
Before writing syntax, inventory:
- nodes and their responsibilities;
- directed relationships and edge labels;
- boundaries, trust zones, or lifecycle states;
- facts that must remain visible at the intended display size.
Preserve the source architecture. Do not invent components to make the drawing symmetrical.
Visual hierarchy
Layout tweak mechanisms nudge Graphviz's automatic placement; they are not for exact positioning. Use them sparingly and prefer a flat, well-ordered diagram over layout hacks.
- Default to
left to right directionfor short pipelines and system boundaries. - Keep the primary path obvious. Use cycles or secondary arrows only when they convey real behaviour.
- Steer direction per edge with
-up->,-down->,-left->,-right->(or-u->/-d->/-l->/-r->). Use--[hidden]>to position an element without implying a real relationship, and--[norank]>to de-emphasize a connection for layout purposes only. - Element definition order also affects placement — reorder the source before reaching for a layout hack.
- Avoid deeply nested containers until a flat version has rendered well; Graphviz ranking can distort nested layouts. Use
togetherto cluster related elements for layout purposes without adding a semantic boundary. - Fix cramped or sprawling spacing with
skinparam nodesepandskinparam rankseprather than guessing at container structure. - Use restrained semantic fills. Reuse each colour for one meaning.
- Keep edge labels short. Move explanations into nodes or surrounding prose.
- Use typography inside labels:
<b>for the scan path,<i>for supporting context, and<color:#hex>for identifiers or state accents. - Use
\nfor deliberate line breaks. Do not force every line to the same emphasis. - Avoid shadows, decorative gradients, and large icon libraries unless the user requests them.
Example label:
rectangle "<color:#2F8F5B><b>2 WAKE + MEASURE</b></color>\n\n<b>Read the environment</b>\n<i>SHT40 and other sensors</i>" as measure #EAF8EFFor additional text and layout syntax, consult the PlantUML Hitchhiker's Guide or current official PlantUML documentation rather than guessing.
Render and iterate
Produce the editable .puml and at least one vector export (.svg). Add .png when previews or visual inspection tools require it.
Use the installed CLI or JAR and enable Java headless mode when no display server is available:
java -Djava.awt.headless=true -jar plantuml.jar -checkonly diagram.puml
java -Djava.awt.headless=true -jar plantuml.jar -tsvg diagram.puml
java -Djava.awt.headless=true -jar plantuml.jar -tpng diagram.pumlInspect the rendered output at its intended size. Check reading order, clipped text, weak contrast, crossing connectors, ambiguous arrow direction, excess whitespace, and labels that disappear on transparent backgrounds. For crossing connectors, try skinparam linetype polyline or skinparam linetype ortho before restructuring the diagram — but with ortho, re-check that edge labels still sit on their line; they can drift away from it. Correct the source and render again.
Completion standard
Report the source and rendered files, renderer/version used, validation performed, remaining automatic-layout compromises, and whether the user still needs to approve the visual result. Never claim success from -checkonly alone.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

