Create Azure CI/CD Pipelines with YAML

The skill instructs the assistant to act as an Azure DevOps pipeline expert who builds reliable, efficient and easily maintained YAML CI/CD pipelines. That expertise spans build agents, deployment strategies, security practices and integration patterns. Pipelines should always be organized into clear stages with proper variable management and reusable components.

A base pipeline triggers on the main and develop branches while excluding docs/* and README.md. It pulls in a variable group named shared-variables, sets buildConfiguration to Release and vmImageName to ubuntu-latest, and runs a Build stage from templates/build-steps.yml. A Deploy stage depends on Build, runs only when the source branch is refs/heads/main, and uses a deployment job against the production environment with the runOnce strategy.

Multi-stage pipelines add a progression of environments with approvals and gates. A BuildAndTest stage runs DotNetCoreCLI@2 build and test with code coverage. DeployDev runs on the develop branch against the dev environment with the dev-variables group, downloads the drop artifact and publishes with AzureWebApp@1. DeployProd runs on main against production with a canary strategy and increments of 25, 50 and 100.

Reusable templates are parameterized. A Docker template takes imageName, dockerfilePath (default Dockerfile), buildContext (default .) and registryConnection, builds and pushes with Docker@2 using the build number and latest as tags, and exposes the tag as an output variable named imageTag.

Security and compliance stages run SonarCloud prepare, analyze and publish (polling timeout of 300 seconds), a SonarCloud quality gate check and a WhiteSource@21 scan. Infrastructure stages install Terraform 1.5.0, then run TerraformTaskV4@4 init against an azurerm backend (resource group, storage account, a tfstate container and an environment-specific state key), a plan with the environment's .tfvars file written to tfplan, and an apply of that plan.

Performance guidance runs tests with parallel: 4, caches NuGet packages with Cache@2 keyed on packages.lock.json, and runs dotnet test with a TRX logger, coverage and a filter that excludes the Integration category. Monitoring uses a PowerShell step with condition: always() that posts a failure message to a Teams webhook when the job status is Failed.

The skill's key recommendations:

• Use variable groups for environment-specific configuration and secrets.
• Implement branch-specific pipeline behavior.
• Use pipeline templates for consistency across projects.
• Include unit, integration and security testing stages.
• Use deployment jobs with environments for tracking and approvals.
• Implement blue-green or canary deployments for production safety.
• Cache dependencies to improve pipeline performance.
• Use conditions and dependencies for efficient flow control.
• Include infrastructure deployment in the pipeline where applicable.
• Set up monitoring and alerts for pipeline health and deployment success.

Use it when an assistant should write Azure DevOps YAML pipelines for build, test, deploy and infrastructure automation. The worked stages cover .NET builds and tests, Azure Web App deployment, Docker images and Terraform on azurerm, so the patterns are Azure DevOps-specific.

The output is Azure DevOps YAML, including stages, jobs, deployment jobs, step templates and parameterized templates. The base pipeline starts with this trigger section:

trigger:
  branches:
    include:
    - main
    - develop
  paths:
    exclude:
    - docs/*
    - README.md

The patterns use these Azure DevOps and third-party tasks: DotNetCoreCLI@2, AzureWebApp@1, Docker@2, SonarCloudPrepare@1, SonarCloudAnalyze@1, SonarCloudPublish@1, WhiteSource@21, TerraformInstaller@0, TerraformTaskV4@4, Cache@2 and PowerShell@2. Environments and variable groups carry per-stage configuration, and a Microsoft Teams webhook receives failure notifications.