skills/business-flow-diagrams/SKILL.md
Create professional, presentation-ready Excalidraw diagrams for business flows, system architectures, pipelines, and process maps using the official excalidraw/excalidraw-mcp server. Use this skill whenever the user asks to draw, diagram, visualize, map, or create any kind of flowchart, architecture, process flow, pipeline, or workflow. Default output is the interactive Excalidraw MCP result and a shareable/openable Excalidraw link when the server or client surfaces one.
npx skillsauth add hollaugo/prompt-circle-skills business-flow-diagramsInstall this skill globally with one command. Works with Claude Code, Cursor, and Windsurf.
3 of 9 scanners reported clean
Some scanners were skipped, did not run, or reported a non-clean status. Review each row below.
This skill uses the official excalidraw/excalidraw-mcp
server as the default path for diagram work.
Primary goal:
.excalidraw JSON.Read:
references/design-system.md for colors, spacing, grouping, and visual quality.references/prompt-patterns.md for prompt shapes that reliably produce good diagrams.Do not default to references/tool-reference.md. That file is for the older multi-tool canvas workflow and is now fallback-only.
Always prefer the official remote MCP server:
https://mcp.excalidraw.comhttps://github.com/excalidraw/excalidraw-mcpNormal workflow:
Default deliverable:
Do not:
Preferred outputs, in order:
.excalidraw or raw JSON only if the user explicitly asks, or if MCP output is unavailableIf the client/server exposes a direct Excalidraw link, include it plainly in the response.
If the MCP app renders inline but does not expose a portable link, say that directly and do not invent one.
The expected output is not merely "a diagram exists." It must be visually clean enough to present.
For a good final result, the diagram should:
When the system contains multiple independent flows:
For architecture diagrams with multiple automations, the preferred output shape is:
Do not force everything into one dense canvas if splitting the story creates a cleaner result.
When sending a request to the MCP app, structure the prompt with:
Use compact, explicit language. Example template:
Draw a clean Excalidraw architecture diagram for [NAME].
Type: [business flow | system architecture | pipeline | swimlane]
Flow direction: [left-to-right | top-to-bottom | swimlane]
Actors/systems:
- ...
Main steps:
1. ...
2. ...
3. ...
Shared infrastructure:
- ...
Style:
- presentation-ready
- clearly grouped containers
- distinct colors for each major flow
- labeled arrows
- separate shared-infrastructure section
Keep prompts specific enough to avoid missing components, but avoid over-specifying pixel-level coordinates unless the user asks for exact layout.
Apply the design language from references/design-system.md:
Layout preference order:
For architecture diagrams:
For business workflows:
For multi-flow automation systems:
The upstream README currently recommends the remote server:
https://mcp.excalidraw.comIf the client supports custom MCP connectors, configure:
{
"mcpServers": {
"excalidraw": {
"url": "https://mcp.excalidraw.com"
}
}
}
The upstream repo also documents a local build path, but remote is the default and preferred route for this skill.
Use fallback behavior only when necessary:
If the official MCP server is available:
If the official MCP server is unavailable but the user still needs a diagram:
.excalidraw file or JSON only if the user agrees or explicitly asked for itIf the user explicitly requests iterative canvas editing or low-level element control:
yctimlin/mcp_excalidraw canvas workflow as the refinement pathIf the one-shot diagram is structurally correct but visually messy:
When the diagram succeeds through the MCP app:
When the server does not provide a link:
When blocked:
Do not stop after the first diagram if it fails obvious visual checks.
Before treating the output as complete, verify:
If the first pass fails:
When browser or canvas inspection is available, use it to critique the actual rendered output rather than trusting the prompt alone.
Example architecture request:
Draw a clean Excalidraw architecture diagram for an automated tax preparation system. Type: system architecture Flow direction: left-to-right Main flows:
- Invoice ingestion and classification
- Stripe transaction sync
- On-demand tax report generation Shared infrastructure:
- Supabase
- OpenAI API
- Slack bot Style:
- top-to-bottom side-by-side columns for the 3 flows
- distinct colors per flow
- labeled containers
- minimal arrows
- separate shared infrastructure section
- presentation-ready
Example business workflow request:
Draw a swimlane Excalidraw diagram for invoice approval. Actors: employee, finance, approver, ERP Steps: submit invoice, validate, request approval, post to ERP, notify employee Style: clean, modern, readable, labeled handoffs
Example expected-output instruction:
If the overview becomes cluttered, produce one overview diagram and three separate flow diagrams. Prioritize clarity over squeezing everything into one image.
Official source used for this skill:
As of the current upstream README, the recommended installation path is the remote server at:
https://mcp.excalidraw.comtools
Create, recreate, redesign, publish, and operate websites managed from Notion, including blogs, CMS-driven sections, widgets, filtering/search interactions, SEO/AEO/GEO improvements, and lightweight deployment workflows. Use when a user wants one skill that can both build and manage a website over time, with OpenClaw-friendly automation but no hard dependency on OpenClaw-specific tooling.
tools
Connect OpenClaw to the Shopify Admin API and run agent-driven store workflows. Covers authentication setup (client credentials grant), token management, and eight ready-to-run use cases: PDP rewrite, weekly store digest, competitive intelligence, abandoned cart recovery, flash sale planner, low stock alert, customer VIP report, and product launch prep. Use this skill when the user asks about connecting OpenClaw to Shopify, automating any Shopify store task, setting up Shopify API credentials, or running any of the eight use cases.
testing
Deploy, harden, and operate OpenClaw across local and hosted environments (Fly.io, Render, Railway, Hetzner, GCP) with secure defaults, channel setup guidance, integration onboarding, and troubleshooting workflows grounded in official OpenClaw documentation. Use when users need install/deploy help, migration support, runtime hardening, memory/agent operations tuning, or incident response.
tools
Build new MCP Apps (MCP servers with React UI output) using @modelcontextprotocol/ext-apps and the MCP SDK. Use when asked to scaffold or implement MCP App servers, add UI-rendering tools/resources, or migrate a standard MCP server to an MCP App with Vite single-file UI bundles.