AWS Diagram Generator

When to Use

• User has AWS CloudFormation templates (YAML/JSON)
• User provides AWS CLI output (e.g.,

  aws ec2 describe-instances

  )
• User wants to visualize AWS resources
• User mentions AWS services (EC2, S3, RDS, Lambda, VPC, etc.)
• User asks to "diagram my AWS infrastructure"

How It Works

1. Parse AWS Resources: Extract resources from CloudFormation, CLI output, or descriptions
2. Map AWS Relationships: Identify VPCs, subnets, security groups, IAM roles
3. Generate Eraser DSL: Create Eraser DSL code from AWS resources
4. Call Eraser API: Use

   /api/render/elements

   with

   diagramType: "cloud-architecture-diagram"

Instructions

1. Parse the Source
   • CloudFormation: Extract

     Resources

     section, identify types (AWS::EC2::Instance, etc.)
   • CLI Output: Parse JSON output from

     aws

     commands
   • Description: Identify AWS service names and relationships
2. Identify AWS Components
   • Networking: VPCs, Subnets, Internet Gateways, NAT Gateways, Route Tables
   • Compute: EC2 Instances, Auto Scaling Groups, Lambda Functions, ECS Services
   • Storage: S3 Buckets, EBS Volumes, EFS File Systems
   • Databases: RDS Instances, DynamoDB Tables, ElastiCache Clusters
   • Security: Security Groups, IAM Roles, IAM Policies, NACLs
   • Load Balancing: ALB, NLB, CLB
   • Other: SQS Queues, SNS Topics, API Gateway, CloudFront
3. Map Relationships
   • EC2 instances in subnets
   • Subnets in VPCs
   • Security groups attached to instances
   • IAM roles attached to services
   • Load balancers targeting instances
   • Databases accessed by applications
4. Generate Eraser DSL Convert AWS 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: "VPC 10.0.0.0/16"]

   Example:

   main-vpc [label: "VPC 10.0.0.0/16"] {
     public-subnet [label: "Public Subnet"] {
       web-server [icon: aws-ec2, label: "Web Server"]
       load-balancer [icon: aws-elb]
     }
     private-subnet [label: "Private Subnet"] {
       database [icon: aws-rds]
       cache [icon: aws-elasticache]
     }
   }
   data-bucket [icon: aws-s3]
   function [icon: aws-lambda]
   load-balancer -> web-server
   web-server -> 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.tf

     - VPC and subnet definitions)
   • External references: Note any documentation, examples, or URLs consulted (e.g., AWS VPC 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
      ![{Title}]({imageUrl})

      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}

      undefined

   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.

AWS-Specific Tips

• Show Regions and AZs: Include availability zones for multi-AZ deployments
• VPC as Container: Always show VPCs containing subnets and resources
• Security Groups: Include security group rules and attachments
• IAM Roles: Show IAM roles attached to services
• Data Flow: Show traffic flow (Internet → ALB → EC2 → RDS)
• Use AWS Icons: Request AWS-specific styling in the description

Example: CloudFormation with Multiple AWS Services

User Input

Resources:
  MyVPC:
    Type: AWS::EC2::VPC
    Properties:
      CidrBlock: 10.0.0.0/16

  PublicSubnet:
    Type: AWS::EC2::Subnet
    Properties:
      VpcId: !Ref MyVPC
      CidrBlock: 10.0.1.0/24

  WebServer:
    Type: AWS::EC2::Instance
    Properties:
      InstanceType: t3.micro
      SubnetId: !Ref PublicSubnet

  MyBucket:
    Type: AWS::S3::Bucket
    Properties:
      BucketName: my-app-bucket

  MyFunction:
    Type: AWS::Lambda::Function
    Properties:
      Runtime: python3.9
      Handler: index.handler

  MyDatabase:
    Type: AWS::RDS::DBInstance
    Properties:
      Engine: postgres
      DBInstanceClass: db.t3.micro

Expected Behavior

1. Parses CloudFormation:
   • Networking: VPC, Subnet
   • Compute: EC2 instance, Lambda function
   • Storage: S3 bucket
   • Database: RDS PostgreSQL instance
2. Generates DSL showing AWS service diversity:

   MyVPC [label: "VPC 10.0.0.0/16"] {
     PublicSubnet [label: "Public Subnet 10.0.1.0/24"] {
       WebServer [icon: aws-ec2, label: "EC2 t3.micro"]
     }
   }
   
   MyBucket [icon: aws-s3, label: "S3 my-app-bucket"]
   MyFunction [icon: aws-lambda, label: "Lambda python3.9"]
   MyDatabase [icon: aws-rds, label: "RDS PostgreSQL db.t3.micro"]
   
   WebServer -> MyBucket
   MyFunction -> MyDatabase
   WebServer -> MyDatabase

   Important: All label text must be on a single line within quotes. AWS-specific: Include service icons, show data flows between services, group by VPC when applicable.
3. Calls

   /api/render/elements

   with

   diagramType: "cloud-architecture-diagram"

Example: AWS CLI Output

User Input

User runs: aws ec2 describe-instances
Provides JSON output

Expected Behavior

1. Parses JSON to extract:
   • Instance IDs, types, states
   • Subnet IDs, VPC IDs
   • Security groups
   • Tags
2. Formats and calls API