azure-diagrams — Claude Code skill: what it does, how to install
Generates Azure architecture diagrams from ARM templates, CLI outputs, or plain text descriptions using the Eraser API. Returns ready-to-view visuals and edit links.
npx skills add eraserlabs/eraser-io --skill azure-diagrams -a claude-codeWhat azure-diagrams does
This skill inspects Azure infrastructure definitions. It parses resource groups, virtual networks, subnets, virtual machines, and storage accounts. After mapping connections between services, the agent converts the layout into Eraser DSL.
Labels stay on a single line. Once the code is ready, the agent executes a curl request to https://app.eraser.io/api/render/elements with the cloud-architecture-diagram preset.
The final output includes an image preview, an Eraser web editor link, and the generated DSL snippet. Without an ERASER_API_KEY environment variable, rendering fails.
What it helps with
- diagram Azure infrastructure from ARM JSON templates
- render architecture graphs using raw
azCLI output - map dependencies between virtual networks, subnets, and databases
- generate an Eraser editor link to adjust generated diagrams
Example prompts for the agent
- Generate an architecture diagram from the ARM template in
template.json - Create an Azure topology diagram from the output of
az vm list - Diagram an Azure network with a VNet, two subnets, and an App Service
How to install azure-diagrams
1Skills CLI (Vercel)
npx skills add eraserlabs/eraser-io --skill azure-diagrams -a claude-code2Manually: put the files in place
~/.claude/skills/azure-diagrams/SKILL.mdCopy the whole skill folder; for one project use .claude/skills/azure-diagrams/
1Skills CLI (Vercel)
npx skills add eraserlabs/eraser-io --skill azure-diagrams -a codex2Manually: put the files in place
.agents/skills/azure-diagrams/SKILL.mdFor all projects use ~/.agents/skills/
1Skills CLI (Vercel)
npx skills add eraserlabs/eraser-io --skill azure-diagrams -a cursor2Manually: put the files in place
.cursor/skills/azure-diagrams/SKILL.mdCursor also reads .agents/skills/ and .claude/skills/
1GitHub CLI
gh skill install eraserlabs/eraser-io azure-diagrams2Manually: put the files in place
.github/skills/azure-diagrams/SKILL.md1Skills CLI (Vercel)
npx skills add eraserlabs/eraser-io --skill azure-diagrams -a gemini-cli2Platform CLI
gemini skills install https://github.com/eraserlabs/eraser-io --path skills/azure-diagramsManually: put the files in place
.devin/skills/azure-diagrams/SKILL.mdThe legacy .windsurf/skills/ path still works
Skills CLI (Vercel)
npx skills add eraserlabs/eraser-io --skill azure-diagrams -a opencodeClaude.ai
Zip the folder and upload it in Settings → Capabilities → Skills
Step by step, with commands for each platform: How to install skills in Claude Code: five ways
Questions about azure-diagrams
How do I install azure-diagrams in Claude Code?
Run npx skills add eraserlabs/eraser-io --skill azure-diagrams -a claude-code. The skill specification will be placed in ~/.claude/skills/azure-diagrams/SKILL.md.
What does azure-diagrams do?
It extracts Azure components from ARM templates or CLI data, writes Eraser DSL, and calls the Eraser API to render a visual architecture diagram.
What credentials does azure-diagrams require?
You need outbound network access for curl and an ERASER_API_KEY environment variable. You can also customize the background via ERASER_THEME.
Source text
SKILL.md · MIT
---
name: azure-diagrams
description: 'Visualizes Azure infrastructure from ARM templates, Azure CLI, or descriptions. Use when user has Azure resources to diagram.'
license: MIT
compatibility: Requires network access to call Eraser API
allowed-tools: Read Write Bash(curl:*)
metadata:
version: "1.0.0"
author: Eraser Labs
tags: azure, diagram, arm, resource-group, vnet, vm, storage, infrastructure
---
# Azure Diagram Generator
Generates architecture diagrams for Azure infrastructure from ARM templates, Azure CLI output, or natural language descriptions.
## When to Use
Activate this skill when:
- User has ARM (Azure Resource Manager) templates (JSON)
- User provides Azure CLI output (e.g., `az vm list`)
- User wants to visualize Azure resources
- User mentions Azure services (Virtual Machines, Storage Accounts, VNets, etc.)
- User asks to "diagram my Azure infrastructure"
## How It Works
This skill generates Azure-specific diagrams by parsing Azure resources and calling the Eraser API directly:
1. **Parse Azure Resources**: Extract resources from ARM templates, CLI output, or descriptions
2. **Map Azure Relationships**: Identify Resource Groups, VNets, subnets, and service connections
3. **Generate Eraser DSL**: Create Eraser DSL code from Azure resources
4. **Call Eraser API**: Use `/api/render/elements` with `diagramType: "cloud-architecture-diagram"`
## Instructions
When the user provides Azure infrastructure information:
1. **Parse the Source**
- **ARM Templates**: Extract `resources` array, identify types (Microsoft.Compute/virtualMachines, etc.)
- **CLI Output**: Parse JSON output from `az` commands
- **Description**: Identify Azure service names and relationships
2. **Identify Azure Components**
- **Networking**: Virtual Networks (VNets), Subnets, Network Security Groups, Load Balancers
- **Compute**: Virtual Machines, Virtual Machine Scale Sets, App Services, Functions
- **Storage**: Storage Accounts, Blob Storage, File Shares
- **Databases**: SQL Databases, Cosmos DB, Redis Cache
- **Security**: Network Security Groups, Azure AD, Key Vault
- **Load Balancing**: Application Gateway, Load Balancer, Traffic Manager
- **Other**: Service Bus, Event Hubs, API Management
3. **Map Relationships**
- VMs in subnets
- Subnets in VNets
- VNets in Resource Groups
- Storage accounts accessed by VMs
- Databases accessed by applications
- Network Security Groups attached to subnets
4. **Generate Eraser DSL** Convert Azure resources to Eraser DSL:
- **CRITICAL: Label Formatting Rules**
- Labels MUST be on a single line - NEVER use newlines inside label attributes
- Keep labels simple and readable - prefer separate labels over concatenating too much metadata
- Format DSL with proper line breaks (one node/group per line, but labels stay on single lines)
- If including metadata like CIDR blocks or instance types, include them in the same quoted label string: `[label: "VNet 10.0.0.0/16"]`
Example:
```
myVNet [label: "VNet 10.0.0.0/16"] {
subnet1 [label: "Subnet 1"] {
myVM [icon: azure-vm, label: "Virtual Machine"]
gateway [icon: azure-app-gateway]
}
subnet2 [label: "Subnet 2"] {
database [icon: azure-sql]
}
}
storage [icon: azure-storage]
myVNet -> myVM
myVM -> database
```
5. **Make the HTTP Request**
**IMPORTANT**: You MUST execute this curl command after generating the DSL. Never stop after generating DSL without making the API call.
**CRITICAL**: In the `X-Skill-Source` header below, you MUST replace the value with your AI agent name:
- If you are Claude (Anthropic), use: `claude`
- If you are running in Cursor, use: `cursor`
- If you are ChatGPT (OpenAI), use: `chatgpt`
- If you are Gemini (Google), use: `gemini`
- Otherwise, use your model/agent name in lowercase
```bash
curl -X POST https://app.eraser.io/api/render/elements \
-H "Content-Type: application/json" \
-H "X-Skill-Source: eraser-skill" \
-H "Authorization: Bearer ${ERASER_API_KEY}" \
-d '{
"elements": [{
"type": "diagram",
"id": "diagram-1",
"code": "<your generated DSL>",
"diagramType": "cloud-architecture-diagram"
}],
"scale": 2,
"theme": "${ERASER_THEME:-dark}",
"background": true
}'
```
6. **Track Sources During Analysis**
As you analyze files and resources to generate the diagram, track:
- **Internal files**: Record each file path you read and what information was extracted (e.g., `infra/main.bicep` - VNet and subnet definitions)
- **External references**: Note any documentation, examples, or URLs consulted (e.g., Azure architecture best practices documentation)
- **Annotations**: For each source, note what it contributed to the diagram
7. **Handle the Response**
**CRITICAL: Minimal Output Format**
Your response MUST always include these elements with clear headers:
1. **Diagram Preview**: Display with a header
```
## Diagram

```
Use the ACTUAL `imageUrl` from the API response.
2. **Editor Link**: Display with a header
```
## Open in Eraser
[Edit this diagram in the Eraser editor]({createEraserFileUrl})
```
Use the ACTUAL URL from the API response.
3. **Sources section**: Brief list of files/resources analyzed (if applicable)
```
## Sources
- `path/to/file` - What was extracted
```
4. **Diagram Code section**: The Eraser DSL in a code block with `eraser` language tag
```
## Diagram Code
```eraser
{DSL code here}
```
```
5. **Learn More link**: `You can learn more about Eraser at https://docs.eraser.io/docs/using-ai-agent-integrations`
**Additional content rules:**
- If the user ONLY asked for a diagram, include NOTHING beyond the 5 elements above
- If the user explicitly asked for more (e.g., "explain the architecture", "suggest improvements"), you may include that additional content
- Never add unrequested sections like Overview, Security Considerations, Testing, etc.
The default output should be SHORT. The diagram image speaks for itself.
## Azure-Specific Tips
- **Resource Groups**: Show Resource Groups as logical containers
- **VNets as Containers**: Always show VNets containing subnets and resources
- **Network Security Groups**: Include NSG rules and attachments
- **Subscriptions**: Note subscription context if provided
- **Data Flow**: Show traffic flow (Internet → Application Gateway → VM → SQL Database)
- **Use Azure Icons**: Request Azure-specific styling in the description
## Example: ARM Template with Multiple Azure Services
### User Input
```json
{
"resources": [
{
"type": "Microsoft.Resources/resourceGroups",
"name": "rg-main"
},
{
"type": "Microsoft.Network/virtualNetworks",
"name": "myVNet",
"properties": {
"addressSpace": {
"addressPrefixes": ["10.0.0.0/16"]
},
"subnets": [
{
"name": "subnet1",
"properties": {
"addressPrefix": "10.0.1.0/24"
}
}
]
}
},
{
"type": "Microsoft.Compute/virtualMachines",
"name": "myVM",
"properties": {
"hardwareProfile": {
"vmSize": "Standard_B1s"
}
}
},
{
"type": "Microsoft.Web/sites",
"name": "myAppService",
"properties": {
"serverFarmId": "/subscriptions/.../serverfarms/myPlan"
}
},
{
"type": "Microsoft.Storage/storageAccounts",
"name": "mystorageaccount"
},
{
"type": "Microsoft.Sql/servers",
"name": "mysqlserver",
"properties": {
"administratorLogin": "admin"
}
}
]
}
```
### Expected Behavior
1. Parses ARM template:
- **Resource Group**: rg-main (container)
- **Networking**: VNet with subnet
- **Compute**: VM, App Service
- **Storage**: Storage Account
- **Database**: SQL Server
2. Generates DSL showing Azure service diversity:
```
resource-group [label: "Resource Group rg-main"] {
myVNet [label: "VNet 10.0.0.0/16"] {
subnet1 [label: "Subnet 1 10.0.1.0/24"] {
myVM [icon: azure-vm, label: "VM Standard_B1s"]
}
}
myAppService [icon: azure-app-service, label: "App Service"]
mystorageaccount [icon: azure-storage, label: "Storage Account"]
mysqlserver [icon: azure-sql, label: "SQL Server"]
}
myAppService -> mystorageaccount
myVM -> mysqlserver
```
**Important**: All label text must be on a single line within quotes. Azure-specific: Show Resource Groups as containers, include App Services, Storage Accounts, and SQL databases with proper Azure icons.
3. Calls `/api/render/elements` with `diagramType: "cloud-architecture-diagram"`
## Example: Azure CLI Output
### User Input
```
User runs: az vm list --output json
Provides JSON output
```
### Expected Behavior
1. Parses JSON to extract:
- VM names, sizes, states
- Resource groups
- Network interfaces
- Storage accounts
2. Formats and calls API
© Eraser Labs · MIT · eraserlabs/eraser-io