Microsoft recently released the Azure Skills Plugin - a package that gives AI coding agents like GitHub Copilot and Claude Code genuine Azure expertise instead of generic cloud advice. This article distills all announcments into a single cookbook you can bookmark and reference whenever you need it.

The Azure Skills Plugin bundles 21 curated skills, the Azure MCP Server (~200 tools across 40+ services), and the Foundry MCP Server in a single install. Skills provide the workflow logic; MCP provides the execution layer.

What the Plugin Actually Is

The plugin is not a prompt pack or a folder of clever one-liners. It ships three layers:

Layer	What It Does	Scale
Azure Skills	Reusable workflows, decision trees, and guardrails that teach the agent how Azure work gets done	24 skill files
Azure MCP Server	Structured tools for listing resources, checking prices, querying logs, provisioning infra, and driving deployments	~200 tools, 40+ services
Foundry MCP Server	Model deployment, agent management, and model catalog workflows for AI-powered apps	AI Foundry integration

Skills are plain-text markdown files (SKILL.md) that the agent reads on demand. They do not run code themselves - they instruct the agent on when, how, and in what order to call MCP tools. The result is that a single prompt like "Deploy my Python Flask API to Azure" triggers a coordinated chain of azure-prepare, azure-validate, and azure-deploy instead of producing a generic tutorial.

Prerequisites

Before installing, make sure you have:

• Node.js 18+ on your PATH (MCP servers run via npx)
• Azure CLI (az) installed and authenticated (az login)
• Azure Developer CLI (azd) installed and authenticated (azd auth login) for deployment workflows
• An Azure account (free tier works for exploring, real subscription for deployments)
• Git CLI (required for the VS Code extension)

Installation

GitHub Copilot CLI

# Add the marketplace source (first time only)
/plugin marketplace add microsoft/azure-skills

# Install the plugin
/plugin install azure@azure-skills

# IMPORTANT: reload MCP servers after install
/mcp reload

Verify with /mcp status - you should see azure and context7 servers listed and running.

VS Code

Install the Azure MCP extension from the VS Code Marketplace (Extension ID: ms-azuretools.vscode-azure-mcp-server). It auto-installs a companion extension containing the Azure skills. After install:

1. Open Copilot Chat (Ctrl+Shift+I / Cmd+Shift+I)
2. Switch to Agent mode (not Ask or Edit mode)
3. Command Palette (Ctrl+Shift+P) > search "MCP" > verify servers are running

Claude Code

# Add the marketplace source (first time only)
claude plugin marketplace add microsoft/azure-skills

# Install the plugin
claude plugin install azure@azure-skills

Smoke Test (All Hosts)

Try these two prompts to confirm everything works:

"List my Azure resource groups"

If you get live results from your subscription, MCP tools are working.

"What Azure services would I need to deploy a Python web API?"

If the agent references azure-prepare and gives structured guidance (not generic suggestions), skills are loaded.

Skills Quick Reference

Here is the complete list of all 21 skills, what they do, and a sample prompt for each.

Infrastructure Planning

Skill	Purpose	Try This
azure-enterprise-infra-planner	Architects enterprise-scale infrastructure - landing zones, hub-spoke networks, multi-region DR topologies, subscription-scope Bicep deployments	"Design a hub-spoke network with a firewall and private endpoints for my production workloads"

This skill is for infrastructure-first planning (landing zones, network topologies, multi-region designs). For app-centric workflows where you're deploying an application, use azure-prepare instead.

Deploy Lifecycle

Skill	Purpose	Try This
azure-prepare	Analyzes your project, recommends services, generates Bicep/Terraform, Dockerfile, azure.yaml	"Prepare my Node.js Express API for Azure Container Apps"
azure-validate	Pre-flight checks on infra code, auth, permissions	"Validate my Azure configuration before I deploy"
azure-deploy	Orchestrates deployment via azd, provisions infra, pushes images, returns live URL	"Deploy this project to Azure"

You rarely need to call these individually. A single prompt like "Create and deploy this Python Flask API to Azure Container Apps" chains all three automatically.

Cost and Quotas

Skill	Purpose	Try This
azure-cost-optimization	Finds orphaned resources, oversized VMs, wrong storage tiers, idle dev environments	"Analyze my Azure spending and find savings"
azure-quotas	Shows current usage vs. limits for compute in any region	"How many vCPUs do I have available in East US?"

Resource Discovery

Skill	Purpose	Try This
azure-resource-lookup	Queries Azure Resource Graph to find and filter resources	"Find resources tagged with environment=staging"
azure-resource-visualizer	Generates Mermaid architecture diagrams from a resource group	"Generate an architecture diagram of my resource group 'prod-rg'"

Compute

Skill	Purpose	Try This
azure-compute	Recommends VM sizes with real-time pricing from Azure Retail Prices API	"What VM size should I use for a Python ML training job with 64GB RAM and a GPU?"
azure-kubernetes	It covers AKS cluster planning, SKU selection, networking, and Day-0 configuration decisions	"Create a production-ready AKS cluster with private API server and Azure CNI Overlay"

Diagnostics

Skill	Purpose	Try This
azure-diagnostics	Runs KQL queries against Log Analytics, correlates events, identifies root cause	"My container app is returning 503 errors - what's wrong?"

Security and Identity

Skill	Purpose	Try This
azure-rbac	Recommends least-privilege roles, generates az role assignment commands and Bicep	"What's the least-privilege role for reading blobs from a storage account?"
azure-compliance	Audits resources against Azure best practices, checks Key Vault expiration	"Run a compliance scan on my subscription"
entra-app-registration	Full app registration flow: create, configure redirect URIs, set API permissions, generate MSAL code	"Create a new Entra ID app registration for my web API"

AI and Foundry

Skill	Purpose	Try This
azure-ai	SDK setup and working code for Azure AI Search, Speech, OpenAI, Document Intelligence	"Set up Azure AI Search with vector search for my documents"
microsoft-foundry	Deploys models, creates agents, runs evaluations, manages Foundry projects	"Deploy a GPT-4o model through Microsoft Foundry"
azure-aigateway	Configures Azure API Management with semantic caching, token limits, content safety	"Set up Azure API Management as an AI Gateway in front of my Azure OpenAI models"

Migration

Skill	Purpose	Try This
azure-cloud-migrate	Maps AWS/GCP services to Azure equivalents, converts infra code, generates migration assessments	"Assess my AWS Lambda functions for migration to Azure Functions"
azure-upgrade	That section is currently about cross-cloud migration; this skill handles intra-Azure upgrades between plans, tiers, and SKUs (e.g. Consumption to Flex Consumption, App Service to Container Apps)	"Upgrade my function app from Consumption to Flex Consumption"

azure-cloud-migrate handles migration from other cloud providers (AWS, GCP) to Azure. azure-upgrade handles upgrades within Azure - changing plans, tiers, or SKUs (e.g. moving from Consumption to Flex Consumption, or App Service to Container Apps).

Observability and Data

Skill	Purpose	Try This
appinsights-instrumentation	Generates SDK setup and telemetry patterns for Application Insights	"Add Application Insights telemetry to my Express.js API"
azure-kusto	Generates and runs KQL queries against Azure Data Explorer	"Write a KQL query to find the top 10 slowest API requests in the last hour"

Storage and Messaging

Skill	Purpose	Try This
azure-storage	Working code for Blob, File, Queue, Table operations; lifecycle policies and access tiers	"Set up lifecycle management to move old blobs to cool storage"
azure-messaging	Troubleshoots Event Hubs/Service Bus SDK issues; generates producer/consumer code	"Set up a Service Bus queue with dead-letter handling in Python"

Copilot SDK

Skill	Purpose	Try This
azure-hosted-copilot-sdk	Scaffolds a Copilot SDK app with session management, model config, and Azure deployment	"Build a GitHub Copilot SDK app and deploy it to Azure"

Prompting Tips

Be Specific About Your Stack

Vague	Better
"Deploy my app"	"Deploy my Python FastAPI app to Azure Container Apps"
"Fix my Azure errors"	"My container app checkout-api is returning 503s since 6 AM"
"Set up a database"	"Create a PostgreSQL Flexible Server with Entra ID auth for my Node.js API"

Name Your Resources

Include resource names so the agent can query them directly without asking follow-up questions:

"Show me errors from the container app 'payments-api' in the last 4 hours"

Chain Prompts for Complex Workflows

You can build up a full workflow incrementally:

1. "What Azure services would I need for this project?"
2. "Prepare it for Azure Container Apps with a PostgreSQL backend"
3. "Validate the configuration"
4. "Deploy it"
5. "Set up Application Insights monitoring"
6. "Run a cost optimization scan"

Ask About Trade-offs

Skills can explain Azure decisions, not just execute them:

"Should I use Container Apps or App Service for this API?"
"What's the trade-off between Bicep and Terraform for this project?"

Troubleshooting

Skills Not Activating

• Copilot CLI: Run /plugin list to confirm the plugin is installed. Check budget with /context.
• VS Code: Confirm you are in Agent mode. Reload window (Ctrl+Shift+P > "Reload Window").
• Claude Code: Run /skills to confirm. Check budget with /context.

MCP Tools Not Showing

• Copilot CLI: Run /mcp reload (required after install), then /mcp show azure.
• VS Code: Check .mcp.json in workspace root. Try Command Palette > "MCP: Restart Servers".
• All hosts: Confirm node --version returns 18+.

Auth Errors

• Refresh with az login and azd auth login.
• Set your subscription: az account set --subscription <your-sub-id>.

Skill Budget Limits

Each host limits how much skill content can be loaded per request:

Host	Budget	Override
Copilot CLI	15,000 tokens	Environment variable
Claude Code	2% of context window (floor: 16,000 chars)	SLASH_COMMAND_TOOL_CHAR_BUDGET env var

If you have many plugins/MCP servers loaded, some skills may be silently excluded. Disable unused skills or reduce total tool count to free up space.

What Gets Installed

The plugin payload lands under .copilot/installed-plugins/azure-skills/azure/ and contains:

• skills/ - 21 SKILL.md files
• .mcp.json - MCP server definitions (the azure server and context7 for up-to-date docs)

The .mcp.json configures two servers:

{
  "servers": {
    "azure": {
      "command": "npx",
      "args": ["-y", "@azure/mcp@latest", "server", "start"]
    },
    "context7": {
      "command": "npx",
      "args": ["-y", "@upstash/context7-mcp@latest"]
    }
  }
}

If you already had the Azure MCP Server installed separately, the plugin simply adds the skills layer on top. Your existing MCP tools continue to work.

Links

• Install the plugin: aka.ms/azure-plugin
• Plugin repo: aka.ms/azure-skills
• Troubleshooting guide: GitHub - Troubleshooting.md
• Announcement series: Part 1 - Announcement | Part 2 - Installation | Part 3 - Prompt Cookbook