excalidraw-diagrams
Skill6 starsChanged 23 days agoMITView on GitHub
Use when creating, editing, or updating any Excalidraw scene, diagram, flowchart, architecture/system diagram, mind map, wireframe, timeline, or slide deck through the Excalidraw+ MCP — whenever the user says "excalidraw", "draw a diagram", "make a flowchart/architecture diagram", "put this on a canvas", or wants recognizable tech/logo icons (Postgres, React, Docker, AWS…) in a diagram. Covers the MCP workflow, element format, layout/centering/alignment, real library icons, and the MCP quirks that silently break diagrams (invisible text, z-order, bindings). Use even for a "simple" diagram — the gotchas bite there too.
7 other projects use a version of this.
Adopted by
Most recently taken up by
- elad-cohen-dev/claude-skillstheir own version, first seen Oct 13, 2026
- sammcj/agentic-codingtheir own version, first seen Oct 8, 2026
- adtn0810/truestacktheir own version, first seen Sep 12, 2026
- shreyas-makes/agent-skillstheir own version, first seen Jul 15, 2026
- alemdarlabs/sage-unreal-mcptheir own version, first seen Jun 4, 2026
- rbrown101010/codex-marketing-skillstheir own version, first seen May 18, 2026
Used with
Projects with it also use code-review.
Other files people use alongside this one.
Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
GoogleSkillcode-review
Adversarial systems, concurrency, memory-model, and architecture code review protocol. Use this skill whenever asked to review code, PRs, diffs, or branches in Filament.
Also in 3 projects that use this
Other versions7 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 7
- sammcj/agentic-coding★ 172same name
- rbrown101010/codex-marketing-skills★ 95same name
- rbrown101010/Rileys-Top-Skills★ 9same name
- adtn0810/truestack★ 2same name
- alemdarlabs/sage-unreal-mcp★ 2same name
- shreyas-makes/agent-skills★ 2same name
- elad-cohen-dev/claude-skills★ 0same name
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository29 other files it gives its agents.
- AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- skills/prompt-engineer/references/claude.mdCLAUDE.md
- brand-assetsSkill
- brand-systemSkill
- codex-reviewSkill
- deck-to-videoSkill
- elevenlabs-sttSkill
- elevenlabs-ttsSkill
- gdoc-syncSkill
- generating-video-clipsSkill
- gsheetsSkill
- gslides-syncSkill
- image-generationSkill
- namecheap-domainsSkill
- narrating-pptxSkill
- office-renderSkill
- pair-intelligenceSkill
- presentation-generatorSkill
- producing-animated-episodesSkill
- prompt-engineerSkill
- proton-passSkill
- reviving-sessionsSkill
- self-presenting-decksSkill
- tavily-extractSkill
- tech-design-docSkill
- viewing-videosSkill
- writing-plainlySkill
- writing-project-instructionsSkill
The file
We split this skill into the rules it gives an agent
2 rules. 0 are common, 1 is distinctive, and 1 is very specific to this project.
0
Common
Most of the 6 edited versions we read keep these.
1
Distinctive
This author's own choices. Worth a look.
1
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 20
Call
read_excalidraw_formatonce: before your first write this session. - Distinctive: line 101
Read the file that matches the task.
“Common” is measured: kept by at least 6 in 10 of the 6 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Excalidraw diagrams (via the Excalidraw+ MCP)
Build scenes that read as a deliberate visual argument, not "boxes with text,"
and get them right the first time by knowing the MCP's non-obvious behaviours up
front. The element JSON here is standard Excalidraw, so it also applies if you ever
generate .excalidraw files directly — but this skill targets the hosted
Excalidraw+ MCP (shareable scenes, collections), which is the usual case.
The workflow
- Plan the structure first (see
reference/layout.md). Pick the family (workflow / architecture / timeline / hierarchy / comparison / cycle / slides), the reading direction, and the regions — before placing anything. The test: if you deleted every label, would the layout alone still convey the idea? - Call
read_excalidraw_formatonce before your first write this session. It's the MCP's own normative element guide (occasionally updated) and the server expects it.reference/elements.mdis your fast working subset. - Create or target a scene. New:
create_scene(needs acollectionId— get it fromlist_collections). Existing: find it vialist_scenes/get_workspace; to edit an earlier one, get real element IDs withsearch_scene_content. The shareable link ishttps://app.excalidraw.com/s/<workspace_id>/<scene_id>— both IDs come back in thecreate_scenemetadata (workspaceandid); give this link to the user. - Add content with
edit_scene_content— one coherent diagram per call. Build in placement order: regions/frames → containers → primary nodes → secondary nodes → labels → arrows (bound, once nodes are stable) → accents. - Add icons where they help (
reference/library-icons.md+ the bundled script) — real Postgres/React/Docker/AWS icons, inserted via the MCP. - Verify — this is mandatory, not optional (see "Verifying" below). Screenshot
for layout,
search_scene_contentfor text/geometry, fix, repeat.
Non-negotiables (each one silently ruins a diagram)
These are the traps that make a diagram look broken or fall apart on the next edit.
Full explanations in reference/mcp-gotchas.md — internalize these five:
- Screenshots don't render text. A screenshot that looks "empty" usually has all
its labels. Never judge text presence/content from a screenshot — use
search_scene_content. - New elements can land under existing fills (the server assigns indices
non-monotonically, even on a fresh delete-free build). Icons placed on cards, or
cards placed in a box, can render invisibly under the opaque fill — often
inconsistently (some show, some don't). Simplest robust fix: give grouping
containers AND cards
backgroundColor:"transparent"(keep the coloured border) — a transparent fill can never occlude, so z-order stops mattering for content inside a box, and a white card on a white canvas looked identical anyway. Only fall back to an explicit highindex(read the current max viasearch_scene_content, don't trust a literal like"b01") when you need one opaque element atop another. Seereference/mcp-gotchas.md. textAlign:"center"makes thexyou pass the CENTER of a standalone label (not the left edge). To center under a node at center X, passx: X.- Bound labels are always vertically centered in their shape. For an
icon-above-label block, use a standalone label positioned yourself, not a bound
label. - Arrows pointing at shapes must be bound (
startBinding/endBinding, mode"inside"), or they detach on the next edit. A bound arrow label (label:{}) is created only ataddtime — adding one viaupdatesilently no-ops (added:0), so re-add an arrow to (re)label it, and read the label as the arrow'slabelText(it's invisible to atypes:["text"]search). You can't delete a shape and its bound arrow in the same call — delete the arrow first.
Verifying (do not claim done without this)
A green edit_scene_content result is not verification. After building:
take_screenshot→ check layout, spacing, overlap, arrow routing, colour, icon shapes. (Not text — it won't render.)search_scene_content→ confirm every label exists with the righttext, and that positions/alignment are what you intended (e.g. equal top/bottom margins for a centered block; sharedyacross a row).- Fix and re-verify until both pass. Typical: 2–4 iterations.
Centering, alignment & icons — the fiddly parts
The bundled script does the error-prone math and the icon extraction. Run it from the skill directory:
# centering: where to put an icon + standalone label as one centered block in a box
python scripts/excalidraw_tools.py place --box X Y W H --icon-h H --font F
# find an icon pack, list its items, extract one positioned icon as add-JSON
python scripts/excalidraw_tools.py catalog <keyword>
python scripts/excalidraw_tools.py list <author/name.excalidrawlib>
python scripts/excalidraw_tools.py icon <lib> --item <idx|name> --at <cx> <top_y> --height <h> [--swap OLD=NEW]place and icon both take --at as center-x, top-y (not the box's top-left
x,y) — feed icon --at the exact center_x/top_y that place printed.
Row/column alignment is just shared y+height (row) or x+width (column) with
a constant step — keeping those exact is what separates intentional from
hand-nudged. Details and the reasoning in reference/layout.md.
Reference map
Read the file that matches the task; don't load them all up front.
reference/mcp-gotchas.md— the MCP quirks above, in full. Read before any real scene work — these cause the most wasted time.reference/elements.md— element fields, fonts, semantic color palette, arrow bindings, frames/slides, charts. Your working reference while building.reference/layout.md— planning, spacing, centering/alignment math, and which structure to choose for a given idea (with a decision table).reference/library-icons.md— finding and inserting real community icons via the MCP; identifying unnamed icons; recommended packs.scripts/excalidraw_tools.py—catalog/list/icon/place. Run--helpfor any subcommand.
Professional vs hand-drawn
Excalidraw defaults to a hand-drawn look (roughness:1, Excalifont). For clean,
technical diagrams set roughness:0, use fontFamily:8 (mono) or 7 (titles),
and lean on gray/blue with one accent. For approachable/whiteboard diagrams, keep
the defaults. Match the register to the audience.
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.

