The Agent Skills Directory

Dynamic Script Generation and Execution: The skill contains embedded PowerShell code blocks for a proxy and helper functions. At runtime, it extracts this content and writes it to the local workspace's .vscode/ directory as new .ps1 files. These generated scripts are then executed to handle JSON-RPC communication and Azure interactions. This pattern involves creating and running executable code that is not present as standalone files in the original skill distribution.
Remote Content Ingestion and Processing: The skill clones the microsoft/discovery repository to fetch agent definitions, YAML configurations, and markdown instructions. This introduces an attack surface for indirect prompt injection, as the agent's behavior is influenced by data retrieved from an external, public source. The skill performs some filtering (e.g., stripping preambles), but the processed instructions are directly used in agent provisioning.
System Command Execution: The skill makes extensive use of system-level tools, including the Azure CLI (az), git, and pwsh. It performs operations such as authenticating to Azure tenants, building container images via Azure Container Registry, and modifying local project files. These commands are used to manage cloud resources and local development configurations.
External Network Operations: The skill initiates network requests to GitHub and Azure services. It performs git clone operations, environment checks via Test-NetConnection, and interacts with Microsoft Discovery MCP endpoints via a local proxy. These operations are directed toward well-known technology providers and are consistent with the skill's stated purpose of deploying agents to cloud infrastructure.

add-example-agents