The Agent Skills Directory

Command Execution: The skill executes local commands using az and git to manage pipeline operations.
Evidence: azure-pipeline.ts uses child_process.spawn to run az commands for queuing, checking status, and canceling builds.
Context: Input parameters like branch names and build IDs are validated against strict regex patterns to prevent command injection.
External Data Ingestion: The skill retrieves and processes external content from Azure DevOps, which represents a potential surface for indirect prompt injection.
Ingestion points: Build logs retrieved via the --download-log flag and build status data from the Azure DevOps API.
Boundary markers: The skill does not currently implement specific boundary markers or warnings to the agent context when processing these logs.
Capability inventory: The script can write to the local filesystem (fs.writeFileSync), execute authenticated Azure CLI commands, and perform network requests.
Sanitization: While command-line arguments are sanitized, the content of the downloaded logs or artifacts is not filtered for embedded instructions.
Remote Resource Access: The skill references and downloads tools from external sources to facilitate environment setup.
Evidence: SKILL.md provides instructions to download the Azure CLI installer from https://aka.ms/InstallAzureCLIDeb.
Context: This resource is provided by Microsoft to facilitate the installation of necessary prerequisites for the skill's functionality.

azure-pipelines