design-azure-sql-schema
Pass
Audited by Gen Agent Trust Hub on Sep 28, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [Indirect Prompt Injection]: The skill facilitates the execution of SQL queries and shell commands to retrieve database schema information, which the agent then processes. This introduces a surface for indirect prompt injection where untrusted metadata could theoretically influence agent behavior.
- Ingestion points: Database schema metadata retrieved via sqlcmd and the provided audit SQL scripts.
- Boundary markers: The instructions do not define specific delimiters to isolate the untrusted output of database queries.
- Capability inventory: Includes the ability to execute shell commands such as bash and sqlcmd, along with T-SQL statements.
- Sanitization: There are no explicit sanitization steps defined for the data extracted from database system views.
Audit Metadata