Claude Code Plugins

Community-maintained marketplace

Feedback

Practical guide for creating human-readable and agent-parseable diagrams using Mermaid. Includes conservative, renderer-compatible templates and when-to-use guidance.

Install Skill

1Download skill
2Enable skills in Claude

Open claude.ai/settings/capabilities and find the "Skills" section

3Upload to Claude

Click "Upload skill" and select the downloaded ZIP file

Note: Please verify skill by going through its instructions before using it.

SKILL.md

name mermaid-skill
description Practical guide for creating human-readable and agent-parseable diagrams using Mermaid. Includes conservative, renderer-compatible templates and when-to-use guidance.
license MIT

Mermaid Diagrams Skill

This skill provides:

  • A conservative set of Mermaid templates that render on older renderers (VS Code/Markdown previewers, Git platforms) and remain clear to humans.
  • Guidance on which diagram type to use for which situation.
  • Compatibility tips and fallbacks when advanced Mermaid types are unavailable.

IMPORTANT: Skills must be authored in English. User-facing answers may remain in Russian per project rules.

Compatibility Rules (Read First)

  • Prefer graph LR/graph TB for flowcharts; some renderers fail on flowchart keyword.
  • Quote labels containing spaces/special characters: A["Text (x|y) |"].
  • Do not use literal \n inside labels – Mermaid не интерпретирует такие переводы строки. Используйте <br/> для разрыва строки.
  • Advanced types like quadrantChart, sankey-beta, requirementDiagram, gitGraph may not be available. Use provided flowchart fallbacks.
  • Code fences must start at column 0 with language mermaid.

ASCII/Unicode Sidecar (Human-Readable Raw Markdown)

To optimize for quick human scanning in raw Markdown and robust parsing by agents, always ship an ASCII/Unicode sidecar immediately below each Mermaid block.

Agent Policy:

  • MUST include a monospace, text-only diagram right under the Mermaid block using fenced code with language text.
  • MUST keep Mermaid and sidecar in sync (same nodes/edges, same labels where feasible). If they diverge, treat Mermaid as the source of truth and update the sidecar.
  • SHOULD limit width to ~80 columns for readability in diffs and terminals.
  • SHOULD use simple line art characters (ASCII first; Unicode box-drawing optional when environment supports it).
  • MAY add a one-line caption above the pair: Diagram: <name> (<type>).

Recommended primitives:

  • Boxes: [Name], (Name), +-----+\n| N |\n+-----+
  • Flows: -->, decisions as {cond?} lines, lists with or -.
  • Sequence (text-based): Actor -> Actor: message with indented lifelines.

Example (Flowchart):

graph LR
  A["Start"] --> B{Auth?}
  B -->|Yes| C["Dashboard"]
  B -->|No|  D["Login"]
Diagram: Auth flow (flowchart)
  [Start] --> {Auth?}
      {Auth?} -- Yes --> [Dashboard]
      {Auth?} -- No  --> [Login]

Example (Text-based Sequence):

sequenceDiagram
  participant U as User
  participant W as WebApp
  U->>W: Open
  W-->>U: OK
Diagram: Happy path (sequence)
  User -> WebApp : Open
  WebApp -> User : OK

Working Templates (Renderer-Compatible)

Flowchart

graph LR
  A["Start"] --> B{Auth?}
  B -->|Yes| C["Dashboard"]
  B -->|No|  D["Login"]
  C --> E["Settings"]

Sequence

sequenceDiagram
  autonumber
  participant U as User
  participant W as WebApp
  participant API
  U->>W: Open
  W->>API: GET /status
  API-->>W: 200
  W-->>U: OK

Class

classDiagram
  class User {
    +String id
    +String name
    +login(): bool
  }
  class Order {
    +String id
    +Decimal total
    +submit()
  }
  User "1" o-- "*" Order

State (v2)

stateDiagram-v2
  [*] --> Idle
  Idle --> Loading : fetch
  Loading --> Ready : ok
  Loading --> Error : fail
  state Ready {
    [*] --> Viewing
    Viewing --> Editing : edit
    Editing --> Viewing : save
  }
  Error --> Idle : retry

ER (Entity-Relationship)

erDiagram
  USER ||--o{ ORDER : places
  ORDER ||--|{ ORDER_LINE : contains
  PRODUCT ||--o{ ORDER_LINE : referenced
  USER {
    string id
    string email
  }
  PRODUCT {
    string id
    string name
    float price
  }

Journey (User Journey)

journey
  title Checkout UX
  section Browse
    "See product": 5: User
    "Add to cart": 4: User
  section Payment
    "Enter card": 2: User
    "3DS confirm": 2: User
  section Result
    "Success page": 5: User

Gantt

gantt
  title Release Plan
  dateFormat  YYYY-MM-DD
  section Dev
  Spec  :done,   des1, 2025-10-01,2025-10-05
  Impl  :active, des2, 2025-10-06,2025-10-20
  Tests :        des3, 2025-10-21, 7d
  section Release
  Freeze :milestone, m1, 2025-10-28, 0d
  Deploy :crit,    des4, 2025-10-29, 1d

Pie (compatible syntax)

pie
  title Traffic by Source
  "Direct"  : 35
  "Organic" : 45
  "Ads"     : 20

Quadrant — flowchart fallback

graph TB
  Q1["Quick Wins<br/>High Impact - Low Effort<br/><br/>• Improve UX"]
  Q2["Major Projects<br/>High Impact - High Effort<br/><br/>• Rewrite Core"]
  Q3["Fill-ins<br/>Low Impact - Low Effort<br/><br/>• Docs polish"]
  Q4["Thankless<br/>Low Impact - High Effort<br/><br/>• Legacy migration"]

  Q1 --> Q2
  Q1 --> Q3
  Q2 --> Q4
  Q3 --> Q4

Requirement — flowchart fallback

graph LR
  R1["Requirement: PCI-DSS compliant"]
  T1["Test: PCI checklist"]
  SVC["Service"]

  SVC -- satisfies --> R1
  T1  -- verifies  --> R1

Sankey — flowchart fallback (weights on edges)

graph LR
  Checkout["Checkout"] -->|100| PSP["PSP"]
  PSP -->|60|  Settled["Settled"]
  PSP -->|40|  Declined["Declined"]

Git graph — flowchart fallback (simple DAG)

graph LR
  A["init"] --> B["feat-A"]
  A --> C["fix-1"]
  B --> D["merge"]
  C --> D

When to Use Which Diagram

  • Flowchart: General flows, decisions, and data movement in specs and PRDs.
  • Sequence: Interactions over time between actors/services (APIs, requests, responses).
  • Class: Domain models and static structure; useful for entity attributes and relations.
  • State: Lifecycle of an entity/component (idle → loading → ready/error, nested states).
  • ER: Database/logical data model with cardinalities.
  • Journey: User experience across steps/sections (great for PRD acceptance flows).
  • Gantt: Scheduling, releases, and dependencies by dates.
  • Pie: Simple composition/ratios; prefer tables when precision matters.
  • Quadrant (fallback): Prioritization matrix (Impact/Effort) without experimental chart support.
  • Requirement (fallback): Traceability between requirements, tests, and system elements.
  • Sankey (fallback): Convey relative volumes along a path when sankey is unavailable.
  • Git graph (fallback): Small branch/merge DAGs when gitGraph is unavailable.

Troubleshooting

  • If a diagram fails to render, try:
    1. Replace flowchart with graph and simplify shapes.
    2. Quote node texts.
    3. Test in https://mermaid.live to isolate environment issues.
    4. Fall back to the templates above for maximum compatibility.

Changelog

  • 2025-10-29: Initial version with conservative, renderer-compatible templates.