maf-online-endpoint
Pass
Audited by Gen Agent Trust Hub on Aug 19, 2026
Risk Level: SAFECOMMAND_EXECUTIONCREDENTIALS_UNSAFE
Full Analysis
- Command Execution: The skill generates and executes scripts (e.g.,
deploy.shanddeploy_sdk.py) that utilize the Azure CLI (az ml) and the Azure AI ML Python SDK to provision managed online endpoints. These operations are essential for the deployment process and are executed within the user's authenticated environment. - Credential and Secret Management: The skill prompts for and processes sensitive information, such as Azure Subscription IDs and service API keys. To mitigate the risk of accidental exposure, the skill automatically generates a
.gitignoreentry to exclude rendered configuration files that might contain plaintext credentials from being tracked in version control. - Access Control Management: The deployment process includes automated Role-Based Access Control (RBAC) assignments for Azure Managed Identities. This pattern facilitates a more secure authentication model by using system-assigned identities instead of hardcoded API keys for service-to-service communication.
- Dynamic Code Loading: The provided scoring script (
score.py) dynamically imports the user's specific workflow logic from aworkflow.pyfile. This is a standard architectural pattern for Azure ML deployments, enabling a generic container to serve custom agent logic.
Audit Metadata