D2 logo

D2

Community
diegomarino
d2

Internal routing hub for d2 diagram skills

Overview

Publisherdiegomarino
Repositoryclaude-toolshed
Skill named2
Stars
58
Forks
5
Bundled files
11
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.

  • 11 bundled files

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

  • Open source

    Published by diegomarino on GitHub. Read the source before you install it.

Installation

Install the D2 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/diegomarino/claude-toolshed/tree/main/plugins/d2/skills/d2
  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/diegomarino/claude-toolshed.git /tmp/claude-toolshed
mkdir -p .claude/skills
cp -r /tmp/claude-toolshed/plugins/d2/skills/d2 .claude/skills/d2
Restart Claude Code after copying so it picks up the new skill.

Use it in TypingMind

Enable D2 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 D2 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 D2 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.

D2 — Diagram Composition Policies

Global rules for all D2 diagram generation. Every specialist and generation flow MUST enforce these. They are not optional.

CRITICAL — Every Diagram Must

  1. Use snake_case IDs — api_gw: "API Gateway", NOT "API Gateway" as ID. Emojis go in labels only.
  2. Set direction explicitly — direction: right for flows, direction: down for hierarchies. If unsure, use direction: right. Never omit.
  3. Stay at one abstraction level — don't mix context actors with component internals.
  4. Stay under visual budget — max 12 top-level nodes, 7 per group, 3 nesting levels, 4 outgoing edges.

Abstraction Levels

Every diagram operates at exactly ONE of these levels:

LevelShowsDoes NOT show
ContextUsers, external systems, the system as a black boxInternal services, databases, queues
ContainerApplications, services, databases, message brokersInternal classes, functions, SQL columns
ComponentModules, layers, handlers within a single containerOther containers, infrastructure
DeploymentRuntime nodes, pods, VMs, networks, regionsBusiness logic, data models

Rule: Do not mix levels. If the user's request implies multiple levels, normalize to the dominant level or generate separate diagrams.

Grouping Discipline

Every diagram uses exactly ONE grouping criterion:

CriterionGroups byTypical use
Layerfrontend / backend / data / infraContainer diagrams, layered architectures
Domainauth / orders / payments / usersMicroservices, bounded contexts
Ownershipteam-a / team-b / platformOrg-level views
Environmentdev / staging / prodDeployment diagrams

Rule: Choose one. If the user's description mixes criteria, ask or default to the most prominent one.

Visual Budget

Hard limits:

MetricLimitAction if exceeded
Visible top-level elements12Simplify or split into overview + details
Nodes per group7Split the group or elevate children
Nesting depth3 levelsFlatten or extract to separate diagram
Outgoing edges from one node4Introduce hub / gateway / bus
Groups with exactly 1 node0Remove the group wrapper
Label length~24 charsShorten, abbreviate, or move detail to tooltip

When the budget is exceeded:

  1. Simplify (remove low-value nodes)
  2. If all nodes are essential: split into overview + detail diagrams

Orientation Heuristics

PatternDirection
Request flow, API integration, pipeline, data flowdirection: right
Hierarchy, org chart, tree, dependency graphdirection: down (default)
Deployment with environmentsOne column per environment
Sequence diagramsN/A (fixed layout)

Rule: Always set direction explicitly. Never rely on the engine default.

Layout Engine Selection

ProfileEngineReason
< 8 nodes, simple flow, pipelinedagreFast (~78ms), clean for simple graphs
>= 8 nodes, nested containers, dense graphselkBetter spacing (~586ms), handles density
Sequence diagramsdagreLayout engine is ignored for sequences
ER with > 6 tableselkBetter table spacing

Node IDs and Labels

  • IDs: short, semantic, snake_case — api_gw, user_svc, orders_db
  • Labels: descriptive but compact — API Gateway, User Service, Orders DB
  • Emojis: in labels only, never in IDs
  • Format: api_gw: "API Gateway" — not "API Gateway" as both ID and label

Compositional Patterns by Type

Defaults — override if the flow demands otherwise.

System Context

  • Left: users, actors
  • Center: system under design (single box, no internals)
  • Right or below: external dependencies
  • Max: 5-8 nodes

Container

  • Left: entry points (users, gateways)
  • Center: application tier (services, APIs)
  • Right: data tier (databases, caches, queues)
  • Below: async/supporting (workers, monitoring)
  • Group by: layer (default) or domain

Component

  • Scope: one container only
  • Group by: layer or module (not both)
  • No deployment concerns, no external systems
  • Max: 8-10 nodes

Deployment

  • Group by: environment or network boundary
  • Separate: ingress / compute / data
  • No business actors unless showing access paths

Sequence

  • Max: 5-6 actors
  • Use groups for happy path / error path
  • Avoid self-messages — use notes instead
  • Labels under 30 chars

ER

  • Core entities: center
  • Lookup/reference tables: periphery
  • Crow's foot notation for cardinality
  • Max: 10 tables per diagram (split by domain)

Class

  • Interfaces/abstracts: top
  • Implementations: below
  • One package/module per diagram

Repair Pass

Before delivering ANY diagram, verify:

  1. Level mixing: context-level actors mixed with component internals? Fix.
  2. Grouping consistency: same criterion throughout? Fix.
  3. Budget compliance: counts within limits? Simplify or split.
  4. Labels: any label > 24 chars? Shorten.
  5. Orphan nodes: any node with zero connections? Remove or connect.
  6. Single-child groups: any group with 1 node? Dissolve.
  7. Fanout: any node with > 4 outgoing edges? Add intermediary.
  8. Self-loops (sequences): can they become notes? Convert.
  9. Direction set? Is direction explicitly declared? Add it.
  10. Engine appropriate? Complexity matches engine? Switch if needed.

Split Strategy

If the structural plan exceeds the visual budget:

  1. Generate an overview diagram at one level higher of abstraction
  2. Generate detail diagrams for each major group or subsystem
  3. Naming: {type}-{scope}-overview-{date}.d2, {type}-{scope}-detail-{subsystem}-{date}.d2

Plugin Path Resolution

Run once before any file reads or script calls:

bash
PLUGIN_DIR=$(find "$HOME/.claude/plugins/cache" -type d -name "d2" -path "*/skills/d2" 2>/dev/null | head -1)

If empty (dev/repo usage):

bash
PLUGIN_DIR=$(find "$HOME" -maxdepth 8 -type d -name "d2" -path "*/skills/d2" 2>/dev/null | head -1)

If still empty: stop — "Plugin not found. Is it installed? Run /plugin install d2@claude-toolshed."

Config Reading

Read .claude/d2.json as a first step. Supported keys:

  • theme_id (default: 0) — D2 theme ID
  • layout (default: "dagre") — layout engine: dagre, elk, tala
  • sketch (default: false) — hand-drawn style
  • output_directory (default: "./diagrams")
  • auto_validate (default: true)
  • auto_render (default: false)
  • output_format (default: "svg")

Embedded Config

Every generated .d2 file includes a vars block at the top:

d2
vars: {
  d2-config: {
    theme-id: 0
    layout-engine: dagre
    sketch: false
  }
}

Common Mistakes

MistakeFix
Using Mermaid arrow syntax (-->)D2 uses -> for directed connections
Sequence diagram outside containerUse shape: sequence_diagram at root level
vars block not at top of fileMust be the first statement
Layout engine tala not workingRequires separate tala binary: brew install tala
PNG rendering failsRequires Playwright: see troubleshooting.md

Entry Points

Entry pointWhen to use
/d2-diagramGenerate one diagram from a description
/d2-convertConvert a Mermaid diagram to D2 format
/d2-validateBatch-check existing .d2 or .md files
/d2-renderRender an existing .d2 file to SVG or PNG
/d2-configConfigure defaults and check dependencies
/d2-architectAnalyze a codebase path and generate a diagram suite
diagram-architect agentClaude detects diagram need during active development

Diagram Type Routing

User wantsSpecialist
API flow, service-to-service calls, request/responsespecialists/d2-sequence.md
System design, microservices, infrastructure, CI/CD, workflowsspecialists/d2-architecture.md
Database schema, SQL tables, entities, relationshipsspecialists/d2-er.md
Class hierarchy, OOP design, data models, interfacesspecialists/d2-class.md

Reference Guides

NeedLoad
Syntax errors, rendering failuresreferences/guides/troubleshooting.md
Learning D2, preventing errorsreferences/guides/common-mistakes.md
Themes, layout engines, sketch modereferences/guides/styling-guide.md
D2 layout features (grids, padding, direction syntax)references/guides/layout-guide.md
Quick routing from symptom to actionreferences/guides/quick-decision-matrix.md

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 D2 AI skill do?

Internal routing hub for d2 diagram skills

Why use D2 on TypingMind?

Because you install it once and use it with any model. D2 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 D2 in TypingMind?

Open Plugins → Skills → Install from GitHub in TypingMind and paste https://github.com/diegomarino/claude-toolshed/tree/main/plugins/d2/skills/d2. 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 D2?

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 D2?

As many as you like. As long as a model supports skills, you can use D2 with it — GPT, Claude, Gemini, Grok, DeepSeek, Mistral, Llama and more — all on TypingMind with your own API keys.

Is the D2 AI skill free?

It is published on GitHub by diegomarino. 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 👇