agentleFS
cubxxwcubxxw/agent-kit/skills/excalidraw-architecture/SKILL.md

excalidraw-architecture

Skill2 starsChanged 3 months agoMITView on GitHub

Create, revise, render, and review editable Excalidraw diagrams for collaborative whiteboarding, open-ended architecture exploration, workshops, narrative maps, hand-drawn technical communication, and polished SVG/PNG handoff. Use for Excalidraw files, live human editing, evolving structure, alternatives and unknowns, or requests to improve diagram layout, connectors, hierarchy, and visual quality. Also use when choosing between Excalidraw and Archify; route formal, validated, repository-backed, interactive, or Before/Delta/After technical documentation to Archify.

2 other projects use a version of this.

Adopted by

Most recently taken up by

The file

Show its rules and text

We split this skill into the rules it gives an agent

23 rules. About 0 are common, 19 are distinctive, and 4 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈19

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 4

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Create diagrams that explain a decision or route, not decorated inventories.

    line 8
  2. Distinctive:

    Read tool-choice.md whenever the request could fit both Excalidraw and Archify.

    line 12
  3. Distinctive:

    Use Excalidraw when the artifact must help people think together: the structure is still moving, alternatives and questions matter, live …

    line 15
  4. Distinctive:

    Use Archify when the artifact must freeze an inspectable technical claim: typed topology, repository evidence, architecture deltas, …

    line 18
  5. Distinctive:

    Use both in sequence when discovery is genuinely needed: explore in Excalidraw, freeze the accepted facts and question, then author fresh …

    line 21
  6. Distinctive:

    If a strict offline or zero-external-request artifact is required, prefer the local Excalidraw export unless the Archify delivery has been …

    line 25
  7. Distinctive:

    Do not mix a messy workshop history and a final architecture claim in one canvas.

    line 38
  8. Distinctive:

    Use solid strokes for V1/current scope and dashed strokes for later scope.

    line 54
  9. Distinctive:

    Sketch hierarchy: Use 5-9 primary groups. Give the primary route the shortest visual path.

    line 71
  10. Very specific:

    Build the scene: Follow visual-system.md and json-contract.md.

    line 73
  11. Distinctive:

    Run these commands from this skill directory:

    line 79
  12. Distinctive:

    Prefer one strong visual argument over a grid of equal boxes.

    line 101
  13. Distinctive:

    Keep connector labels short.

    line 103
  14. Very specific:

    Default to roughness: 1; use 0 only for dense technical regions and 2 only for a deliberately rough annotation.

    line 107
  15. Very specific:

    Default text to Excalifont (fontFamily: 5), which uses Xiaolai as the current Excalidraw CJK hand-drawn fallback.

    line 109
  16. Very specific:

    Use font sizes 36 / 28 / 20 / 16 / 14.

    line 112
  17. Distinctive:

    Keep titles to one line and bodies to three short lines where possible.

    line 113
  18. Distinctive:

    Use color plus labels and stroke style.

    line 114
  19. Distinctive:

    Add icons only when they identify a real platform or service faster than text.

    line 115
  20. Distinctive:

    Do not use gradients, shadows, decorative blobs, or ornamental arrows.

    line 116
  21. Distinctive:

    every important route has a direction and valid endpoints.

    line 125
  22. Distinctive:

    no text is clipped, overlapping, or too small at export size.

    line 126
  23. Distinctive:

    all nodes, bindings, IDs, and phase styles pass the validator.

    line 128

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 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.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Excalidraw Architecture

Create diagrams that explain a decision or route, not decorated inventories.

Choose the medium

Read tool-choice.md whenever the request could fit both Excalidraw and Archify.

  • Use Excalidraw when the artifact must help people think together: the structure is still moving, alternatives and questions matter, live editing is central, or hand-drawn language usefully avoids false precision.
  • Use Archify when the artifact must freeze an inspectable technical claim: typed topology, repository evidence, architecture deltas, multiple reading depths, route exploration, deterministic validation, or formal delivery.
  • Use both in sequence when discovery is genuinely needed: explore in Excalidraw, freeze the accepted facts and question, then author fresh typed Archify IR. Never convert the sketch geometry mechanically.

If a strict offline or zero-external-request artifact is required, prefer the local Excalidraw export unless the Archify delivery has been explicitly made offline-safe.

Choose the working mode

  • Workshop mode: expose competing routes, assumptions, questions, and decision points. Keep uncertainty visible and leave room for people to move the scene. Optimize for conversation, not final polish.
  • Delivery mode: copy the accepted structure into a clean scene, remove abandoned branches, freeze connector semantics, and run the full render and screenshot gate below.

Do not mix a messy workshop history and a final architecture claim in one canvas. Preserve the workshop source separately when its reasoning history is valuable.

Deliverables

Unless the user narrows the request, deliver:

  1. editable .excalidraw;
  2. reviewable .svg and .png;
  3. a one-paragraph reading guide;
  4. the final scene opened in Excalidraw when browser delivery is useful.

In workshop mode, the editable .excalidraw is mandatory; rendered outputs are optional unless the user needs an async review snapshot.

Use solid strokes for V1/current scope and dashed strokes for later scope. Add a visible legend. Do not reuse dashed strokes for uncertainty in the same architecture diagram.

Workflow

  1. Read before drawing. Inspect the source notes, repository, existing architecture, and audience. Write down the one question the diagram answers. In workshop mode, also name the decision that remains open.
  2. Freeze semantics. Separate current, later, external, derived, and authoritative components. Identify the main route, feedback route, trust boundary, and forbidden shortcuts.
  3. Choose one composition.
    • Product story: left to right, 3-5 stages, outcome loop below.
    • System architecture: top to bottom layers, primary runtime left to right.
    • Lifecycle: circular only when the final state truly feeds the first.
    • Topology: hub and spoke only when one authority actually coordinates peers.
  4. Sketch hierarchy. Use 5-9 primary groups. Give the primary route the shortest visual path. Put details inside labeled bands, not nested cards.
  5. Build the scene. Follow visual-system.md and json-contract.md. Mark scope-bearing elements with customData.phase: "v1" or "later".
  6. Render and validate.

    Run these commands from this skill directory:

    node scripts/render-excalidraw.mjs \
      path/to/diagram.excalidraw --png path/to/diagram.png
    node scripts/validate-excalidraw.mjs \
      path/to/diagram.excalidraw --png path/to/diagram.png

    SVG export requires only Node.js. PNG export also requires ImageMagick; macOS uses Quick Look when available and other systems render through ImageMagick directly.

  7. View the full-resolution PNG. Fix clipping, collisions, weak hierarchy, bad wrapping, crowded arrows, excess whitespace, and route ambiguity. Render again after every meaningful edit.
  8. Audit the meaning. Trace every important arrow. Confirm that readers cannot mistake a derived view for a source of truth or a proposal for an executed action.

Drawing Rules

  • Prefer one strong visual argument over a grid of equal boxes.
  • Show overview, section boundaries, and concrete nodes at distinct zoom levels.
  • Keep connector labels short; put explanations beside the route, not on it.
  • Route connectors behind nodes and into edges. Avoid crossing text or entering arbitrary box corners.
  • Reserve accent color for one decision, write boundary, or critical transition.
  • Default to roughness: 1; use 0 only for dense technical regions and 2 only for a deliberately rough annotation.
  • Default text to Excalifont (fontFamily: 5), which uses Xiaolai as the current Excalidraw CJK hand-drawn fallback. Use Nunito (6) when dense technical copy needs a quieter voice.
  • Use font sizes 36 / 28 / 20 / 16 / 14; never finish with body text below 14.
  • Keep titles to one line and bodies to three short lines where possible.
  • Use color plus labels and stroke style; never rely on color alone.
  • Add icons only when they identify a real platform or service faster than text.
  • Do not use gradients, shadows, decorative blobs, or ornamental arrows.

Completion Gate

Finish only when:

  • the audience and diagram question are obvious on first scan;
  • workshop unknowns cannot be mistaken for accepted facts;
  • V1 and later scope are distinguishable without color;
  • every important route has a direction and valid endpoints;
  • no text is clipped, overlapping, or too small at export size;
  • contrast is at least 4.5:1 for normal text;
  • all nodes, bindings, IDs, and phase styles pass the validator;
  • the PNG has been visually inspected, not merely generated;
  • the editable source and rendered outputs describe the same frozen scene.

Download now (ZIP)

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
What did you use it for? e.g. Reviewing pull requests in a Rails app

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.