semantic-model-consumption
Pass
Audited by Gen Agent Trust Hub on Jun 13, 2026
Risk Level: SAFE
Full Analysis
- Read-Only Operation Guardrails: The skill explicitly restricts its scope to metadata discovery and analytical queries. It includes clear 'AVOID' instructions for any model-change operations, which helps prevent accidental or unauthorized modification of semantic model definitions.
- Restricted Tool Invocations: The instructions mandate the use of specific MCP tools (e.g., ExecuteQuery) rather than generic shell commands or PowerShell scripts. This approach utilizes structured APIs that typically enforce their own authentication and authorization checks at the service level.
- Data Volume Management: The ExecuteQuery implementation includes a 'maxRows' parameter with a default of 250 and a hard limit of 1,000 rows. This serves as a security consideration to prevent large-scale data exfiltration and manage session performance.
- Progressive Metadata Discovery: The skill encourages a 'Must/Prefer' pattern of progressive discovery using INFO functions. By recommending filtered and projected queries (selecting only necessary columns), it minimizes the exposure of sensitive model structure information in the agent's working context.
- Dynamic Identity Resolution: The skill avoids hardcoded GUIDs or identifiers, requiring dynamic resolution of workspace and item IDs. This ensures that operations are performed against the intended resources and respects current user permissions.
Audit Metadata