The Agent Skills Directory

Standard Dependency Installation: The skill ensures its execution environment is prepared by installing well-known Python packages if they are missing.
The deployment script uses python -m pip install to fetch pyyaml and requests. These are standard libraries used for configuration parsing and API communication.
Azure Resource Management: The skill leverages the Azure CLI (az) and Docker for cloud infrastructure operations.
It uses az account get-access-token for authentication and az rest for managing Discovery tool resources.
It supports az acr build for remote container builds and docker build for local image generation, which are routine operations for containerized deployments.
Dynamic Task Execution: To optimize deployment speed, the skill generates and executes temporary Python scripts for parallel processing.
Scripts like deploy_tools_parallel.py and deploy_agents_parallel.py are written to the run directory to handle multiple resource updates simultaneously.
These scripts use standard Python modules such as concurrent.futures and subprocess to manage deployment tasks.
Indirect Data Processing Surface: The skill reads and processes local configuration files, which is a common pattern for deployment automation.
Ingestion points: The skill parses kit.json, agent.yaml, and tool.yaml files from the repository's starter-kits/ and agents/ directories.
Boundary markers: The skill performs regex validation on identifiers like knowledgeBaseId and toolResourceId before including them in deployment payloads.
Capability inventory: The skill utilizes pwsh, python, az, and docker to perform file system operations, network requests to Azure endpoints, and process execution.
Sanitization: Payload generation involves structured JSON/YAML transformation rather than raw string interpolation, which helps maintain data integrity during the deployment process.

discovery-services-starter-kit-deployer