eventhouse-authoring-cli
Pass
Audited by Gen Agent Trust Hub on Jul 10, 2026
Risk Level: SAFECOMMAND_EXECUTIONDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
- Command Execution via CLI: The skill uses the
az restutility to execute KQL management commands. This is a core feature that allows for the creation and modification of database objects through an authenticated shell environment. - Indirect Command Processing: Patterns within the skill involve reading commands from local files or user prompts. This creates a surface for indirect prompt injection where commands stored in external files could influence agent behavior. The skill addresses this by instructing the agent to seek explicit user confirmation for ambiguous or impactful requests.
- Interaction with Official Service Endpoints: The skill communicates with well-known Microsoft services such as Fabric and Kusto. These operations are performed using the user's active CLI session credentials, ensuring actions are within the scope of established permissions.
- Data Ingestion from External Sources: The skill includes functionality to ingest data from locations like Azure Blob Storage and OneLake. Users should verify the source and integrity of external data before initiating ingestion tasks.
Audit Metadata