agentsclimarketplace

Generating gcp diagrams

Skill a5c-ai/babysitter/library/specializations/devops-sre-platform/skills/generating-gcp-diagrams

Babysitter enforces obedience on agentic workforces and enables them to manage extremely complex tasks and workflows through deterministic, hallucination-free self-orchestration

Install
npx -y skills add a5c-ai/babysitter --skill generating-gcp-diagrams

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

What its author says it does

Copied from the file, not written here

Generates DrawIO XML diagrams for Google Cloud Platform architectures from text descriptions or images. Analyzes existing .drawio files to extract GCP components. Use for GCP architecture diagrams, cloud infrastructure documentation, or when converting GCP diagram images to editable DrawIO format.

The file declares its own license as MIT. That is the author’s claim about this one file, and it is not the same thing as the license GitHub reports for the repository, which is listed with the other numbers below.

SKILL.md

14.6 KB, as published. Nobody here has run it

GCP DrawIO Diagram Generator

Generates professional DrawIO XML diagrams for Google Cloud Platform architectures.

Capabilities

  1. Extract - Analyze existing DrawIO XML files to identify GCP shapes, connections, and structure
  2. Identify - Recognize GCP service icons from architecture diagram images
  3. Generate - Create valid DrawIO XML from images or text descriptions
  4. Convert - Transform GCP architecture diagrams into editable DrawIO format

Quick Reference

GCP Shape Pattern

shape=mxgraph.gcp2.{service_name}

Note: GCP uses snake_case for shape names (e.g., cloud_run, cloud_sql, cloud_storage).

Icon Pattern: Unlike AWS, GCP uses a single icon pattern for all services — there is no service vs instance icon distinction. The same shape=mxgraph.gcp2.{name} is used whether labeling the service itself ("Cloud Run") or a specific instance ("Cloud Run (API Handler)"). Differentiate by label text only.

Common GCP Services

ServiceShape Code
Cloud Runmxgraph.gcp2.cloud_run
BigQuerymxgraph.gcp2.bigquery
Cloud Storagemxgraph.gcp2.cloud_storage
Vertex AImxgraph.gcp2.cloud_machine_learning
Cloud Schedulermxgraph.gcp2.cloud_scheduler
Apigeemxgraph.gcp2.apigee_api_platform
Pub/Submxgraph.gcp2.cloud_pubsub
Cloud SQLmxgraph.gcp2.cloud_sql
GKEmxgraph.gcp2.compute_engine
Cloud Functionsmxgraph.gcp2.cloud_functions

GCP Container Types

ContainerUse Case
gcp_projectMain project boundary (two-cell pattern)
gcp_vpc_scVPC Service Controls perimeter (green border)
gcp_regionRegional grouping
gcp_zoneZone grouping
logical_group_dashedLogical grouping with dashed border
logical_group_solidSolid border grouping
subnetSubnet boundary
firewall_rulesFirewall rules grouping
instance_groupInstance group container

Task 1: Analyze a DrawIO File

Use this workflow to extract and document all components from an existing DrawIO file.

Steps

  1. Read the file - Load the .drawio XML file
  2. Parse structure - Extract all <mxCell> elements
  3. Identify shapes - Find cells with vertex="1"
  4. Identify connections - Find cells with edge="1"
  5. Extract styles - Parse style strings for each element
  6. Map hierarchy - Build container/child relationships using parent attribute
  7. Generate report - Output findings in structured format

Input

  • Path to .drawio file

Output

Generate a Markdown report with:

# DrawIO Analysis Report

## Summary
- Total shapes: X
- Total connections: Y
- Containers: Z

## Shape Inventory

| ID | Label | Type | Position | Parent |
|----|-------|------|----------|--------|
| abc | Cloud Run | mxgraph.gcp2.cloud_run | (100,200) | vpc1 |

## Connection Matrix

| From | To | Label | Type |
|------|-----|-------|------|
| Cloud Run | BigQuery | API | solid |

## Container Hierarchy

- VPC-SC (vpc1)
  - Cloud Run (run1)
  - Cloud Run (run2)
  - BigQuery (bq1)

## Style Analysis

### Unique Shapes Found
- mxgraph.gcp2.cloud_run (4 instances)
- mxgraph.gcp2.bigquery (2 instances)

Task 2: Convert Image to DrawIO

Use this workflow to recreate a GCP architecture diagram from an image.

Steps

  1. Analyze image - Identify all visual elements:

    • GCP service icons (shape, color, label)
    • Containers/boundaries (color, border style)
    • Connections (solid, dashed, arrows)
    • Labels and text
  2. Map to library - For each identified element:

    • Look up in assets/gcp-icons.json by visual signature or label
    • Match containers to assets/containers.json
    • Note any unrecognized elements
  3. Estimate layout - Determine positions:

    • Identify container boundaries first
    • Place icons within containers
    • Estimate x,y coordinates and dimensions
    • Standard icon size: 50x50 pixels
  4. Generate XML - Build the DrawIO structure:

    • Start with base template from assets/templates/drawio-base.xml
    • Add containers first (they become parents)
    • Add service icons with correct parent references
    • Add connections between shapes
  5. Create confidence report - Document accuracy:

    • List all identified components
    • Note any uncertain matches
    • Flag potential issues

Input

  • GCP architecture diagram image (PNG/JPG)

Output

  1. Valid .drawio XML file
  2. Confidence report (Markdown)

Confidence Report Format

# Conversion Confidence Report

## Overall Confidence: 85%

## Identified Components

### High Confidence (>90%)
- Cloud Run x4 - Clear icon match
- BigQuery x2 - Clear icon match
- VPC-SC container - Green border, correct label

### Medium Confidence (70-90%)
- Vertex AI Search - Icon similar, label confirms

### Low Confidence (<70%)
- Unknown icon at position (300, 400) - Mapped to generic service

## Connection Accuracy
- 12/14 connections clearly visible
- 2 connections inferred from layout

## Notes
- "Same Instance" dashed container identified
- Bidirectional arrows on 3 connections

Task 3: Create DrawIO from Description

Use this workflow to generate a new GCP diagram from text specifications.

Steps

  1. Parse requirements - Extract from description:

    • Required GCP services
    • Container/grouping needs
    • Connection requirements
    • Layout preferences
  2. Select components - From libraries:

    • Look up services in assets/gcp-icons.json
    • Choose containers from assets/containers.json
    • Select connection styles
  3. Plan layout - Design the arrangement:

    • Determine canvas size
    • Position containers first
    • Arrange services logically (left-to-right data flow, top-to-bottom hierarchy)
    • Standard spacing: 100px between 50x50 icons
  4. Generate XML - Build the diagram:

    • Use assets/templates/drawio-base.xml as starting point
    • Add elements in order: containers, services, connections
    • Assign unique IDs to all elements
  5. Validate - Check the output:

    • All requested components present
    • Connections reference valid IDs
    • Layout is logical and readable

Input

  • Text description of desired GCP architecture

Output

  • Valid .drawio XML file

Example Input

Create a GCP architecture with:
- VPC-SC container
- Cloud Scheduler triggering Cloud Run
- Cloud Run connecting to BigQuery and Cloud Storage
- Vertex AI Search connected to BigQuery

Example Output Structure

<mxfile ...>
  <diagram name="GCP Architecture">
    <mxGraphModel ...>
      <root>
        <mxCell id="0" />
        <mxCell id="1" parent="0" />
        <!-- VPC-SC Container -->
        <mxCell id="vpc" value="VPC-SC" style="..." vertex="1" parent="1">
          <mxGeometry x="50" y="50" width="700" height="400" />
        </mxCell>
        <!-- Cloud Scheduler -->
        <mxCell id="sched" value="Cloud Scheduler" style="...mxgraph.gcp2.cloud_scheduler" vertex="1" parent="vpc">
          <mxGeometry x="50" y="100" width="50" height="50" />
        </mxCell>
        <!-- More shapes... -->
        <!-- Connections -->
        <mxCell id="conn1" edge="1" source="sched" target="run" style="..." />
      </root>
    </mxGraphModel>
  </diagram>
</mxfile>

Shape Library Reference

Looking Up a GCP Service

  1. Open assets/gcp-icons.json
  2. Search by service_name or recognition_keywords
  3. Use the drawio_shape.full_style for complete styling
  4. Or construct style using shape=mxgraph.gcp2.{shape_name}

Service Coverage:

  • 46 GCP services across 11 categories
  • 40 exact matches, 6 fallback icons
  • Categories: compute, database, storage, networking, ai_ml, integration, operations, api_management, data_analytics, devops, security

Note: 6 services use fallback icons (Workflows, Eventarc, Artifact Registry, Cloud Deploy, Secret Manager, Identity Platform) as they're newer services not yet in DrawIO's mxgraph.gcp2 stencil. See references/ICON-COMPATIBILITY.md for complete validation details.

GCP Service Categories

CategoryServices
computeCloud Run, Compute Engine, GKE, Cloud Functions, App Engine
databaseBigQuery, Cloud SQL, Firestore, Spanner, Bigtable, Memorystore
storageCloud Storage, Filestore, Persistent Disk
networkingVPC, Load Balancing, CDN, DNS, Armor, Cloud NAT
ai_mlVertex AI, AI Platform, Vision, NLP, Speech-to-Text
integrationPub/Sub, Cloud Tasks, Workflows, Eventarc, Scheduler
operationsLogging, Monitoring, Trace, Error Reporting
api_managementApigee, API Gateway
data_analyticsDataflow, Dataproc, Cloud Composer
devopsCloud Build, Artifact Registry, Container Registry, Cloud Deploy
securityCloud KMS, Secret Manager, Identity Platform

Visual Best Practices (Summary)

For detailed visual design guidelines, see references/DIAGRAM-BEST-PRACTICES.md.

GCP Project Zone (Two-Cell Pattern)

The GCP Project container uses two cells, not one:

  1. Container cell: Plain rectangle with fillColor=#F6F6F6;strokeColor=none; and HTML value <b>Google </b>Cloud Platform
  2. Logo child cell: shape=mxgraph.gcp2.google_cloud_platform at 23x20px with relative=1 geometry

See assets/templates/node-template.xml for the exact template.

Font Colors

  • Service icon labels: fontColor=#424242 (dark gray)
  • GCP Project zone text: fontColor=#717171
  • VPC-SC container title: fontColor=#2E7D32 (dark green)
  • Dashed group labels: fontColor=#424242

Icon Spacing

  • Icon size: 50x50 pixels
  • Standard spacing: 100px between icons
  • Container padding: 20-30px

Connection Labels

Always add these properties to labeled connections:

labelBackgroundColor=#FFFFFF;fontSize=10;fontColor=#333333;

Connection Best Practices

  • Standard width: strokeWidth=1 for most connections
  • Thick width: strokeWidth=2 only for primary data paths (use sparingly - max 1-3 per diagram)
  • Orthogonal routing: edgeStyle=orthogonalEdgeStyle for professional appearance
  • When 3+ connections cross same corridor: consolidate into single connection

XML Structure Quick Reference

For complete XML examples and detailed structure, see references/xml-examples.md.

The key building blocks:

  • Shape: <mxCell id="..." value="Label" style="..." vertex="1" parent="1"> with <mxGeometry>
  • Connection: <mxCell id="..." edge="1" source="..." target="..." style="...">
  • Container: Shape with container=1 in style; children set parent to container ID
  • Root cells: Every diagram needs <mxCell id="0"/> and <mxCell id="1" parent="0"/>

For XML parsing and extraction techniques, see references/xml-parser-guide.md.


Troubleshooting

Icon Not Displaying

  • Verify shape name matches exactly: mxgraph.gcp2.cloud_run (underscore, not hyphen)
  • Check references/ICON-COMPATIBILITY.md for correct shape names
  • Ensure vertex="1" is present
  • Check that <mxGeometry> has valid width/height (50x50)
  • For Workflows/Eventarc, note these don't have exact icon matches in DrawIO's library

Connection Not Rendering

  • Verify source and target IDs exist
  • Ensure edge="1" is set
  • Check that source/target shapes are vertices

Shapes Not Inside Container

  • Set child's parent attribute to container's ID
  • Ensure container has container=1 in style
  • Position child coordinates relative to container (not absolute)

Label Not Showing

  • Check value attribute is set
  • Verify fontSize is reasonable (11-14)
  • Ensure fontColor=#424242 is set

File Won't Open in DrawIO

  • Validate XML structure (proper closing tags)
  • Ensure id="0" and id="1" root cells exist
  • Check for special characters in labels (use &#xa; for newlines)

Desktop Integration

After generating a .drawio file, you can validate and preview it:

  1. Validate: python scripts/validate-drawio.py output.drawio --verbose
  2. Analyze: python scripts/analyze-existing.py output.drawio --markdown
  3. Validate Icons: python scripts/validate-gcp-icons.py
  4. Export PNG: ./scripts/export-diagram.sh output.drawio png
  5. Open in DrawIO: ./scripts/open-diagram.sh output.drawio

Requires DrawIO Desktop. Install on macOS: brew install drawio


Files in This Skill

FilePurpose
SKILL.mdThis file - main instructions
Assets
assets/gcp-icons.jsonGCP service icon database (46 services)
assets/containers.jsonGCP container and connection styles
assets/templates/drawio-base.xmlBase XML template
assets/templates/node-template.xmlShape insertion template
assets/templates/connection-template.xmlConnection template
References
references/ICON-COMPATIBILITY.mdIcon validation reference
references/DIAGRAM-BEST-PRACTICES.mdVisual design and layout guidelines
references/xml-parser-guide.mdDetailed XML parsing reference
references/xml-examples.mdCopy-paste XML examples
references/coordinate-system.mdPositioning and layout guide
references/style-guide.mdStyle string reference
Scripts
scripts/validate-drawio.pyValidate .drawio XML structure
scripts/validate-gcp-icons.pyValidate GCP icon compatibility
scripts/fix-gcp-icons.pyAuto-fix icon shape names
scripts/fix-drawio-icons.pyBulk fix icon references in .drawio files
scripts/extract-shape-names.pyExtract available shapes from DrawIO stencil
scripts/analyze-existing.pyExtract shapes/connections from .drawio files
scripts/export-diagram.shExport to PNG/PDF via DrawIO Desktop CLI
scripts/open-diagram.shOpen .drawio file in DrawIO Desktop

Keep looking

Skills are one crate of 328,083. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.