sqldb-consumption-cli

Pass

Audited by Gen Agent Trust Hub on Jul 10, 2026

Risk Level: SAFECOMMAND_EXECUTION
Full Analysis
  • Read-Only Enforcement: The skill explicitly instructs the agent to refuse any data or schema modification commands, such as INSERT, UPDATE, or DROP. This acts as a primary safety layer to ensure that the agent remains in a consumption-only mode regardless of underlying database permissions.
  • Standard Tool Integration: It utilizes well-known, vendor-provided tools including the Azure CLI (az) and sqlcmd for connectivity and querying. These tools are used in accordance with best practices for Microsoft Entra ID authentication and do not involve untrusted third-party binaries.
  • Dynamic Query Generation: A utility script for text searching uses dynamic SQL (sp_executesql) to iterate through database columns. While this involves runtime code construction, it is a standard administrative pattern for schema discovery and is limited to string comparison operations within the database engine.
  • Data Ingestion Surface: Because the skill reads and displays data from external databases, there is an inherent surface for indirect prompt injection if the database content contains adversarial instructions. This is a common consideration for data-centric tools, and the skill addresses it by maintaining a strict read-only scope and focusing on structured metadata discovery.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 10, 2026, 02:44 PM
Security Audit — agent-trust-hub — sqldb-consumption-cli