azure

name: "azure" description: 'Build scalable, secure, and reliable applications on Microsoft Azure cloud services. Use when deploying to Azure, configuring Azure compute/storage/database services, setting up Azure networking, implementing Azure security, or managing Azure costs and monitoring.' metadata: author: "AgentX" version: "1.0.0" created: "2025-01-15" updated: "2026-02-11" compatibility: providers: ["azure"] platforms: ["windows", "linux"]

Azure Cloud Development

Purpose: Best practices for developing and deploying applications on Microsoft Azure, including compute, storage, security, and DevOps.

When to Use This Skill

• Deploying applications to Microsoft Azure
• Configuring Azure compute, storage, or database services
• Setting up Azure networking and security
• Implementing CI/CD with Azure DevOps
• Managing Azure costs and monitoring

Decision Tree

Azure Service Selection
+-- Hosting a web app or API?
|   +-- Containers? -> Azure Container Apps
|   +-- Serverless functions? -> Azure Functions
|   +-- Full PaaS? -> Azure App Service
|   +-- Kubernetes? -> AKS
+-- Storing data?
|   +-- Relational? -> Azure SQL / PostgreSQL Flexible Server
|   +-- Document/NoSQL? -> Cosmos DB
|   +-- Blob/files? -> Azure Storage
+-- Messaging or events?
|   +-- Queue-based? -> Service Bus
|   +-- Event streaming? -> Event Hubs
+-- AI/ML workloads?
|   +-- LLM hosting? -> Azure OpenAI Service
|   +-- Search + RAG? -> Azure AI Search
+-- Not sure? -> Start with App Service + Azure SQL

Prerequisites

• Azure subscription
• Azure CLI installed
• Basic cloud computing concepts

Table of Contents

1. Azure Fundamentals
2. Compute Services
3. Storage Services
4. Database Services
5. Networking
6. Security
7. DevOps and CI/CD
8. Monitoring and Logging
9. Cost Management
10. AI Services
11. Best Practices

Azure Fundamentals

Resource Hierarchy

Management Group
-- Subscription
 -- Resource Group
 -- Resources (VMs, Storage, etc.)

Naming Conventions

{resource-type}-{workload}-{environment}-{region}-{instance}

Examples:
vm-web-prod-eastus-001
st-data-dev-westus-001
sql-orders-staging-northeu-001

Resource Groups

• Group resources by lifecycle
• One resource group per environment per application
• Apply tags for cost allocation and management

# Create resource group
az group create --name rg-myapp-prod --location eastus --tags Environment=Production Team=Engineering

AI Services

Azure OpenAI Service

• Purpose: LLM hosting (GPT-4o, GPT-3.5-Turbo).
• Best Practice: Use Managed Identity for authentication. Disable keys.
• Networking: Deploy in Virtual Network with Private Endpoint.

# Create OpenAI Resource
az cognitiveservices account create \
 --name "ai-agentx-dev" \
 --resource-group "rg-agentx-dev-eastus-001" \
 --location "eastus2" \
 --kind "OpenAI" \
 --sku "S0"

Azure AI Search

• Purpose: Vector database and retrieval engine for RAG.
• Configuration: Standard Tier required for Semantic Search.
• RBAC: Assign Search Index Data Contributor to the Agent Identity.

Azure Cosmos DB (NoSQL)

• Purpose: Conversation history and agent state.
• Feature: Use Vector Indexing for memory retrieval.
• Consistency: Session consistency is usually sufficient for chat.

Core Rules

[PASS] DO

• Use Managed Identity instead of connection strings
• Enable Azure Defender for security
• Use availability zones for high availability
• Implement proper tagging strategy
• Use Infrastructure as Code (Bicep/Terraform)
• Enable diagnostic logging
• Use Private Endpoints for PaaS services
• Implement proper backup and DR strategy

[FAIL] DON'T

• Store secrets in code or config files
• Use overly permissive network rules
• Ignore Azure Advisor recommendations
• Skip resource locks on production resources
• Use public endpoints for databases
• Forget to set up cost alerts

Anti-Patterns

• Secrets in App Settings: Storing passwords or keys in plain app config -> Use Key Vault references with managed identity
• Oversized SKUs: Choosing premium tiers for dev/test environments -> Use dev-appropriate SKUs, scale up for production only
• No Resource Locks: Production resources unprotected from accidental deletion -> Apply CanNotDelete locks on production resource groups
• Public Database Endpoints: Databases accessible from the internet -> Use Private Endpoints and VNet integration
• Manual Deployments: Deploying via portal clicks without repeatable process -> Use Infrastructure as Code (Bicep/Terraform) with CI/CD pipelines
• Single Region No DR: All resources in one region with no failover plan -> Deploy across availability zones and plan geo-redundancy

References

• Azure Architecture Center
• Azure Well-Architected Framework
• Azure CLI Documentation
• Azure SDK for .NET

Version: 1.1 Last Updated: February 11, 2026

Troubleshooting