The Agent Skills Directory

[Remote Code Execution]: The skill documents the installation of the Azure Developer CLI (azd) using a piped shell command (curl -fsSL https://aka.ms/install-azd.sh | bash). While this pattern involves executing remote code, the source is a well-known Microsoft domain (aka.ms), and the action is a standard installation procedure for the required tooling.
[Command Execution]: The skill is designed to execute a wide array of system commands including az, azd, terraform, and dotnet to manage Azure resources. To mitigate risk, the skill mandates global rules requiring explicit user confirmation (ask_user) for any destructive actions like deleting resource groups or purging Key Vaults, and for confirming the target subscription and location.
[Dynamic Script Execution]: The skill provides utility scripts (Shell and PowerShell) for applying Entity Framework migrations and granting SQL access during the postprovision phase of a deployment. These scripts are intended to be integrated into the project's deployment hooks and handle environment variables dynamically, but they include safety measures to avoid unsafe evaluation of variables.
[Indirect Prompt Injection]: The skill possesses an attack surface for indirect prompt injection as it processes untrusted project files such as azure.yaml, infrastructure templates (Bicep/Terraform), and a deployment plan (.azure/deployment-plan.md). Malicious instructions embedded in these files could theoretically influence the agent's behavior, although the skill emphasizes strict adherence to a validated plan and specific command sequences to minimize this risk.
[External Package Installation]: The migration scripts include logic to automatically install the dotnet-ef global tool if it is missing. This involves downloading software from the official .NET tool registry, which is a standard developer workflow for this application type.

azure-deploy