azuresql-db-auth
Pass
Audited by Gen Agent Trust Hub on Sep 28, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- Template Interpolation (Indirect Prompt Injection): The skill instructs the agent to populate SQL templates with the
AZURE_SQL_APP_PASSWORDenvironment variable to provision database users. This pattern creates a surface where malformed input could potentially influence the resulting SQL command. - Ingestion points: The
AZURE_SQL_APP_PASSWORDenvironment variable is the primary source of external data for template rendering, as described inSKILL.mdandreferences/auth-and-secrets.md. - Boundary markers: The skill explicitly labels the SQL snippets as "non-executable templates" that must be rendered before execution, providing a clear distinction between the template and the final command.
- Capability inventory: The skill is designed for an environment where the agent can execute SQL statements to provision databases, logins, users, and role memberships.
- Sanitization: The documentation in
references/auth-and-secrets.mdprovides explicit sanitization instructions, directing the agent to escape single quotes by replacing them with two single quotes before substitution. - Database Provisioning (Command Execution): The skill guides the agent in executing administrative SQL commands (e.g.,
CREATE LOGIN,CREATE USER,ALTER ROLE) to configure database security. This is consistent with the skill's stated purpose of establishing a least-privilege environment. - External Resource Reference: The skill references a specific Azure Container Registry image (
sqldbpreview-dpgaeqhmgphzd4bk.azurecr.io/azure-sql/db-dev:latest) and links to official Microsoft documentation. These references target well-known vendor infrastructure and represent standard deployment practices.
Audit Metadata