fossflow-diagrams（福斯流程图）

FossFLOW Isometric Diagrams

Overview

Generate isometric architecture diagrams using the FossFLOW/Isoflow compact JSON format. This format uses minified keys for token efficiency while producing valid importable diagrams.

Iron Rule: The compact format structure defined in llm-generation-guide.md (in this skill directory) is the ONLY valid output format. No variations, no "improved" formats, no object-based alternatives. The exact structure must be respected at all costs.

Canonical reference: https://raw.githubusercontent.com/stan-smith/FossFLOW/refs/heads/master/LLM-GENERATION-GUIDE.md

When to Use

• User asks for an architecture diagram, infrastructure diagram, or system topology
• User mentions FossFLOW, Isoflow, or isometric diagrams
• User wants to visualize cloud infrastructure (AWS, Azure, GCP)
• User wants to visualize Kubernetes deployments
• User wants a visual representation of microservices, data pipelines, or any system architecture

When NOT to use: For UML diagrams, sequence diagrams, flowcharts, or non-isometric visualizations (use Mermaid or PlantUML instead).

The Compact Format - MANDATORY Structure

{
  "t": "Title (max 40 chars)",
  "i": [
    ["Item Name (max 30)", "icon_id", "Description (max 100)"]
  ],
  "v": [
    [
      [[itemIndex, x, y], ...],
      [[fromIndex, toIndex], ...]
    ]
  ],
  "_": { "f": "compact", "v": "1.0" }
}

Keys Explained

Items (i) - Each item is a 3-element array

["Display Name", "icon_id", "Brief description"]
     ^max 30        ^valid      ^max 100

Views (v) - Each view is a 2-element array

[
  [[0, x0, y0], [1, x1, y1], ...],   // positions: [itemIndex, x, y]
  [[0, 1], [1, 2], ...]               // connections: [fromIndex, toIndex]
]

Red Flags - STOP and Fix

If you catch yourself doing ANY of these, delete and start over:

• Using object syntax for items ({id:..., type:...}) instead of arrays
• Using string references in connections ("from": "lb-1") instead of integer indices
• Adding keys not in the spec (version, components, connections at root)
• Putting positions inside item definitions instead of in the views array
• Using 3D coordinates [x, y, z] - positions are [itemIndex, x, y]
• Inventing icon names not in the valid icons list
• Omitting the _ metadata field
• Using "v": "1.0" without the enclosing "_": {"f": "compact",...}

Valid Icons - Quick Reference

Generic (Isoflow collection)

storage server user cloud network security api queue cache function mobile web email analytics backup load-balancer cdn vpn firewall monitor

AWS (prefix: aws-)

aws-ec2 aws-s3 aws-rds aws-lambda aws-api-gateway aws-cloudfront aws-route-53 aws-vpc aws-elb aws-iam aws-cloudwatch aws-sns aws-sqs aws-dynamodb aws-elasticsearch aws-redshift aws-kinesis aws-eks aws-fargate aws-cognito

Azure (prefix: azure-)

azure-virtual-machine azure-storage-account azure-sql-database azure-app-service azure-function-app azure-api-management azure-cdn azure-dns azure-load-balancer azure-active-directory azure-monitor azure-service-bus azure-cosmos-db azure-redis-cache azure-kubernetes-service azure-container-instances azure-logic-apps azure-data-factory azure-key-vault azure-cognitive-services

GCP (prefix: gcp-)

gcp-compute-engine gcp-cloud-storage gcp-cloud-sql gcp-app-engine gcp-cloud-functions gcp-api-gateway gcp-cloud-cdn gcp-cloud-dns gcp-cloud-load-balancing gcp-identity-access-management gcp-cloud-monitoring gcp-cloud-pub-sub gcp-cloud-firestore gcp-memorystore gcp-kubernetes-engine gcp-cloud-run gcp-cloud-workflows gcp-cloud-dataflow gcp-secret-manager gcp-ai-platform

Kubernetes (prefix: k8s-)

k8s-pod k8s-service k8s-deployment k8s-configmap k8s-secret k8s-ingress k8s-namespace k8s-node k8s-persistent-volume k8s-daemonset k8s-statefulset k8s-job k8s-cronjob k8s-hpa k8s-rbac

For the complete icon list (320 AWS, 369 Azure, 280 GCP, 56 K8s icons), see llm-generation-guide.md in this skill directory.

Positioning Guidelines

• Grid coordinates: X (horizontal), Y (vertical)
• Negative X = left, positive X = right
• Negative Y = up, positive Y = down
• Range: -20 to +20 for both axes
• Spacing: 3-5 units between components
• Start main components near (0, 0)
• Arrange in logical flow (left-to-right or top-to-bottom)

Connection Patterns

Generation Workflow

1. Identify components from the user's description
2. Map each to a valid icon (use cloud-specific icons when platform is specified)
3. Build items array with concise names and descriptions
4. Calculate positions using grid layout with 3-5 unit spacing
5. Define connections based on data/control flow
6. Validate against the checklist below
7. Output raw JSON (no markdown code fences unless explicitly asked)

Validation Checklist (Run Before Output)

• Root keys are exactly t, i, v, _ (no extras)
• _ is exactly {"f": "compact", "v": "1.0"}
• Each item in i is a 3-element array [name, icon, description]
• All icon names are from the valid icons list
• Item names <= 30 chars, descriptions <= 100 chars, title <= 40 chars
• v contains at least one view: [[positions], [connections]]
• Each position is [itemIndex, x, y] with valid item index
• Each connection is [fromIndex, toIndex] with valid item indices
• Coordinates are within -20 to +20
• Number of positions matches number of items
• JSON is valid (no trailing commas, proper quoting)

Common Mistakes

Complete Example

Microservices with load balancer, 3 services, message queue, and database:

{
  "t": "Microservices Architecture",
  "i": [
    ["Load Balancer", "load-balancer", "Distributes incoming traffic"],
    ["User Service", "server", "Handles user management"],
    ["Order Service", "server", "Processes orders"],
    ["Payment Service", "server", "Payment processing"],
    ["Message Queue", "queue", "Async communication between services"],
    ["Shared Database", "storage", "PostgreSQL shared data store"]
  ],
  "v": [
    [
      [[0, 0, -6], [1, -6, 0], [2, 0, 0], [3, 6, 0], [4, 0, 6], [5, 0, 10]],
      [[0, 1], [0, 2], [0, 3], [1, 4], [2, 4], [3, 4], [1, 5], [2, 5], [3, 5]]
    ]
  ],
  "_": { "f": "compact", "v": "1.0" }
}