Mermaid Skill logo

Mermaid Skill

Community
Agents365-ai
mermaid-skill

Generate Mermaid diagrams (.mmd) and export to PNG/SVG/PDF using mmdc CLI or Kroki API. USE THIS SKILL when user mentions diagram, flowchart, sequence diagram, class diagram, ER diagram, state machine, architecture, visualize, git graph, 画图, 架构图, 流程图, 时序图, 类图, ER图, 甘特图, 状态机. PROACTIVELY USE when explaining ANY system with 3+ components, API flows, authentication sequences, class hierarchies, database schemas, or state machines. Supports 17+ diagram types with fully automatic layout.

Overview

PublisherAgents365-ai
Repository365-skills
Skill namemermaid-skill
Stars
99
Forks
16
Bundled files
10
Links
  • Markdown instructions

    A SKILL.md file the model loads on demand, so it only costs tokens when a request actually matches.

  • Works with any LLM

    AI skills are plain Markdown, not provider-specific code, so this works with GPT, Claude, Gemini, Grok, or a local model.

  • 10 bundled files

    Scripts, templates, and references the model can read while it works. Files are read-only and never executed.

  • Open source

    Published by Agents365-ai on GitHub. Read the source before you install it.

Installation

Install the Mermaid Skill AI skill in TypingMind to use it with any LLM, or drop it into another agent that reads SKILL.md.

1

Install in TypingMind

TypingMind installs a skill straight from its GitHub folder — it reads SKILL.md, bundles the resource files, and stores the result locally.

  1. Open the app and go to Plugins → Skills.
  2. Choose "Install from GitHub".
  3. Paste the skill folder URL below and confirm.
    https://github.com/Agents365-ai/365-skills/tree/main/plugins/mermaid/skills/mermaid-skill
  4. Enable the skill in any chat where you want it available.
Plugins → Skills → Add skill → From GitHub URL, then paste the folder URL and press Continue.
2

Install in another agent

Any agent that reads the Agent Skills format can use this skill — copy the folder into that agent's skills directory.

Claude Code — .claude/skills
git clone --depth 1 https://github.com/Agents365-ai/365-skills.git /tmp/365-skills
mkdir -p .claude/skills
cp -r /tmp/365-skills/plugins/mermaid/skills/mermaid-skill .claude/skills/mermaid-skill
Restart Claude Code after copying so it picks up the new skill.

Use it in TypingMind

Enable Mermaid Skill in any TypingMind chat and the model takes it from there. Its name and description sit in the system prompt, and the moment a request matches, the model loads the full instructions itself — you never invoke it by hand, and it costs no tokens until it is actually used.

The model loads Mermaid Skill on its own as soon as a request matches it.

Works with any AI model

AI skills are plain Markdown instructions rather than provider-specific code, so Mermaid Skill is not tied to the model it was written for. Install it once in TypingMind and use it with GPT-5, Claude, Gemini, Grok, DeepSeek, Mistral, Llama, or a local model you run yourself — all on your own API keys.

  • Loaded only when it is needed

    The system prompt carries just the name and description. The instructions are fetched on the first matching request, so an idle skill costs nothing.

  • Switch models mid-chat

    Because the skill is instructions rather than code, changing model does not break it — the next model reads the same SKILL.md.

Skill instructions

This is the SKILL.md content the model loads. Read it before installing — a skill is instructions your model will follow.

Mermaid Diagrams

Generate .mmd text files and export to PNG/SVG/PDF using mmdc (local) or Kroki API (no install).

Key advantage: Text-based syntax with fully automatic layout — no x/y coordinates needed.

When to use / when NOT to use

Use this skill for: diagrams-as-code with automatic layout (flowchart, sequence, class, state, ER, gantt, mindmap, architecture) — text source that lives in git and embeds in Markdown.

Do NOT use it — route elsewhere — for:

  • Pixel-precise placement, custom layout, branded icons, or heavy styling → drawio.
  • A hand-drawn / sketchy aesthetic → excalidraw or tldraw.
  • A freeform whiteboard or freehand strokes → tldraw.
  • Strict, conventional UML notation → plantuml.

Prerequisites

Option A: Local (mmdc) — also needs a headless Chrome (mmdc renders via Puppeteer)

bash
npm install -g @mermaid-js/mermaid-cli
npx puppeteer browsers install chrome-headless-shell   # required — mmdc has no bundled browser
mmdc --version

mmdc --version succeeds even with no Chrome installed, but every export then fails with Could not find Chrome. Install the browser above (or set PUPPETEER_EXECUTABLE_PATH to a system Chrome). If you can't, use Kroki (Option B) — it needs no browser.

CI / Docker: mmdc crashes with Running as root without --no-sandbox in containers. Pass the bundled puppeteer config (see scripts/puppeteer-config.json):

bash
mmdc -p scripts/puppeteer-config.json -i diagram.mmd -o diagram.png

Option B: Kroki API (no install)

bash
curl --version  # Just need curl

Workflow

  1. Check deps — validate via Kroki (needs only curl); check mmdc --version and a headless Chrome only when local export (PNG quality / PDF) is wanted
  2. Pick diagram type — choose from table below
  3. Generate — write .mmd file to disk
  4. Validate — Kroki-first (see Validation; REQUIRED before export)
  5. Export — use mmdc or Kroki API to produce PNG/SVG/PDF
  6. Self-check (vision) — read the exported PNG and fix readability/layout defects that automatic layout can't prevent (clipped labels, cramped density, wrong orientation), then re-validate + re-export. Max 2 rounds; skip if no vision. See Self-Check (vision) below.
  7. Review loop — show the image to the user, apply the minimal .mmd edit per request, re-export until approved (5-round safety valve). See Review Loop below.
  8. Report — tell user the output file paths

Validation (Required)

NEVER export a diagram without validating first.

Prefer Kroki for validation — it needs no browser and sidesteps the Could not find Chrome trap entirely. Use mmdc for validation only when offline.

bash
# Validate with Kroki (preferred — no browser needed)
curl -s -X POST -H "Content-Type: text/plain" --data-binary @diagram.mmd https://kroki.io/mermaid/svg -o /tmp/test.svg && echo "Valid" || echo "Invalid"

# Validate with mmdc (offline fallback — requires headless Chrome)
mmdc -i diagram.mmd -o /tmp/test.svg 2>&1

# If error, fix the .mmd file and validate again
# Only proceed to export after validation passes

Common validation errors:

  • Missing quotes around labels with special characters
  • Wrong arrow syntax (use ->> for sequence, --> for flowchart)
  • Undeclared participants in sequence diagrams

A Could not find Chrome (or puppeteer) error from mmdc is a setup problem, not a diagram error — the .mmd may be perfectly valid. Validate via Kroki instead of "fixing" correct syntax.

Self-Check (vision)

Zero-cost pre-check: before spending a vision call, read the PNG dimensions (sips -g pixelWidth -g pixelHeight diagram.png on macOS, file diagram.png elsewhere). An extreme aspect ratio (longer side > 8x the shorter) is a "wrong orientation" defect — fix it without vision.

Validation (above) only proves the syntax is legal — it says nothing about whether the rendered diagram is readable. After exporting, use the agent's vision capability to read the PNG and catch what automatic layout can't prevent. Mermaid positions everything itself, so the failures here are about content and readability, not overlaps:

CheckWhat to look forFix
Label truncationNode / edge text clipped or cut offShorten the label, or wrap it with <br/>
Cramped, unreadable densityToo many nodes crammed together; tangled linesFlip direction (TD↔LR), split into subgraphs, or reduce nodes
Wrong orientation / aspectDiagram far too wide or too tall to readChange flowchart TD↔LR (or set direction in class/state)
Edge spaghettiMany edges crossing, hard to followReorder node declarations so connected nodes sit adjacent; group with subgraph
Wrong diagram typeType doesn't suit the content (e.g. flowchart for a timeline)Switch type (gantt, sequenceDiagram, stateDiagram-v2, …)
Low contrastText blends into the node fillAdjust classDef / theme so text contrasts the fill
  • Max 2 self-check rounds — if issues remain after 2 fixes, show the user anyway.
  • Re-validate (syntax) and re-export after every fix.
  • If vision is unavailable, skip self-check and show the PNG directly.

Review Loop

After self-check, show the exported image and collect feedback. Apply the minimal .mmd edit for each request, then re-validate and re-export:

User requestEdit action
Change a labelEdit the node / edge text in the .mmd
Add / remove a node or edgeAdd or delete the matching line
Change a colorAdd / adjust a classDef and class <node> <className>
Change layout directionSwap TD↔LR (flowchart) or set direction (class / state)
Restructure / groupWrap related nodes in a subgraph, or regenerate
  • Overwrite the same diagram.mmd / diagram.png each round — don't create v1, v2, …
  • Safety valve: after 5 rounds, generate a mermaid.live handoff link (see next section) so the user fine-tunes the current diagram in the browser.

Mermaid.live Handoff

When the review loop ends (approved or safety-valve), generate a one-click link that opens the current .mmd in the mermaid.live editor for interactive fine-tuning:

bash
python3 scripts/mermaid_live_link.py diagram.mmd
# -> https://mermaid.live/edit#pako:...

Pure zlib + base64 (raw-deflate + URL-safe base64, same encoding mermaid.live uses), works offline. The script self-verifies with a round-trip decode before printing.

Batch Mode (Repo Survey)

When the user asks to diagram a whole repo ("survey this codebase", "map the architecture"):

  1. Scan — locate schemas (SQL / ORM models), API routes, and state machines / lifecycle enums in the codebase
  2. Propose — one diagram per finding: ER for the schema, sequence for a main API flow, state for a lifecycle, flowchart for overall service layout; confirm the set with the user before generating
  3. Generate + export — run the standard workflow per diagram into one output folder
  4. Overview page — inline every exported SVG into a single self-contained index.html (title + one section per diagram, no CDN), so the user opens one file

Same validation and self-check rules apply to every diagram in the batch.

Diagram Types

TypeKeywordUse for
Flowchartflowchart TD/LRprocesses, pipelines, decisions
SequencesequenceDiagramAPI calls, message passing
ClassclassDiagramOOP models, data structures
ERerDiagramdatabase schemas
StatestateDiagram-v2state machines, lifecycle
Ganttganttproject timelines
Piepieproportions
Git GraphgitGraphbranch strategies
C4 ContextC4Contexthigh-level system context
Architecturearchitecture-betacloud / CI/CD service layouts
Mind Mapmindmaptopic breakdowns
User Journeyjourneyuser-experience flows
Use Caseusecase-betaactor–system interactions (UML)
Cynefincynefin-betasense-making / complexity domains
Event Modelingeventmodelingevent-driven system timelines
Tree ViewtreeView-betafile / directory hierarchies
Wardley Mapswardley-betabusiness strategy / value chains

Syntax Reference

Flowchart: See reference/FLOWCHART.md Sequence: See reference/SEQUENCE.md Class & ER: See reference/CLASS-ER.md Architecture: See reference/ARCHITECTURE.md Use Case: See reference/USECASE.md Other types: See reference/OTHER-TYPES.md Themes & styling: See reference/THEMES.md

Examples

See reference/EXAMPLES.md for four worked examples (JWT auth sequence, microservices architecture, order state machine, cloud architecture) — each with the user prompt, the generated .mmd, and the output files.

Export Commands

Option 1: Local Export (mmdc)

Requires mmdc installed locally. Best for offline use.

bash
# PNG (recommended: 2048px wide, white background)
mmdc -i diagram.mmd -o diagram.png -w 2048 --backgroundColor white

# PNG with theme — valid -t values: default | dark | neutral | forest
# (`base` is NOT a valid -t value; it only works inside a %%{init: {'theme':'base'}}%% directive)
mmdc -i diagram.mmd -o diagram.png -w 2048 --backgroundColor white --theme neutral

# SVG
mmdc -i diagram.mmd -o diagram.svg

# PDF
mmdc -i diagram.mmd -o diagram.pdf

Option 2: Kroki API (No Install Required)

Use Kroki when mmdc is not available. No local dependencies needed.

bash
# SVG via Kroki
curl -X POST -H "Content-Type: text/plain" --data-binary @diagram.mmd https://kroki.io/mermaid/svg -o diagram.svg

# PNG via Kroki
curl -X POST -H "Content-Type: text/plain" --data-binary @diagram.mmd https://kroki.io/mermaid/png -o diagram.png

# PDF is NOT supported by Kroki for Mermaid — POSTing to /mermaid/pdf returns
# HTTP 400 ("Unsupported output format: pdf for mermaid. Must be one of png or svg").
# For PDF, use the local mmdc path instead:  mmdc -i diagram.mmd -o diagram.pdf

Kroki advantages:

  • No local installation required
  • Works on any system with curl
  • Supports 20+ diagram types (PlantUML, GraphViz, D2, etc.)

When to use Kroki:

  • mmdc installation fails
  • Quick one-off diagrams
  • CI/CD pipelines without Node.js

Common Mistakes

MistakeFix
mmdc not foundnpm install -g @mermaid-js/mermaid-cli
mmdc error Could not find ChromeInstall the headless browser: npx puppeteer browsers install chrome-headless-shell (or use Kroki)
Kroki PDF fails with HTTP 400Kroki does PNG/SVG only for Mermaid; use local mmdc for PDF
Valid diagram reported "invalid" by mmdcThe error is a Chrome/puppeteer setup failure, not a syntax error — don't rewrite correct .mmd; fix the browser or validate via Kroki
Wrong arrow in sequenceUse ->> for request, -->> for response
Special chars in labelWrap in quotes: A["Label: value"]
Blank/small outputAdd -w 2048 flag
Participant order wrongDeclare participant explicitly at top
Subgraph name with spacesWrap in quotes: subgraph "My Layer"

Bundled files

The model reads these on demand while the skill is loaded. They are exposed as readable files and are never executed.

Frequently asked questions

What does the Mermaid Skill AI skill do?

Generate Mermaid diagrams (.mmd) and export to PNG/SVG/PDF using mmdc CLI or Kroki API. USE THIS SKILL when user mentions diagram, flowchart, sequence diagram, class diagram, ER diagram, state machine, architecture, visualize, git graph, 画图, 架构图, 流程图, 时序图, 类图, ER图, 甘特图, 状态机. PROACTIVELY USE when explaining ANY system with 3+ components, API flows, authentication sequences, class hierarchies, database schemas, or state machines. Supports 17+ diagram types with fully automatic layout.

Why use Mermaid Skill on TypingMind?

Because you install it once and use it with any model. Mermaid Skill is plain Markdown rather than provider-specific code, so the same skill runs on GPT-5, Claude, Gemini, Grok, or a local model — and you can switch model mid-chat without it breaking. TypingMind runs on your own API keys, so you pay providers directly instead of a per-seat subscription, and your skills and chats stay in your own storage.

How do I install Mermaid Skill in TypingMind?

Open Plugins → Skills → Install from GitHub in TypingMind and paste https://github.com/Agents365-ai/365-skills/tree/main/plugins/mermaid/skills/mermaid-skill. TypingMind reads its SKILL.md and bundles its files and installs it as a skill you can enable per chat.

Which AI models can use Mermaid Skill?

Any model you connect in TypingMind. AI skills are plain Markdown instructions rather than provider-specific code, so GPT, Claude, Gemini, Grok, and local models can all load this skill when a request matches it.

How many AI models can I use with Mermaid Skill?

As many as you like. As long as a model supports skills, you can use Mermaid Skill with it — GPT, Claude, Gemini, Grok, DeepSeek, Mistral, Llama and more — all on TypingMind with your own API keys.

Is the Mermaid Skill AI skill free?

It is published on GitHub by Agents365-ai. Check the repository for licensing terms. You only pay your own AI provider for the tokens you use.

What are AI skills?

An AI skill is a reusable instruction bundle that teaches an AI model how to do one specific task. It follows the open Agent Skills format: a SKILL.md file with a name and description, plus any scripts, templates or reference files the model may need. The model reads the instructions only when your request matches the skill, so an installed skill costs nothing until it is used.

How are AI skills different from plugins or MCP servers?

A plugin or MCP server gives a model new tools to call — code that runs somewhere and returns a result. An AI skill gives the model knowledge and process instead: how to approach a task, which steps to follow, what good output looks like. Skills are plain Markdown, so they need no server, no API key and no runtime, and they work with any model.

View all

Set up your own AI workspace now

Get notified about new features and future giveaways by subscribing to our newsletter 👇