azuresql-db-rag
Pass
Audited by Gen Agent Trust Hub on Sep 28, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- Command Execution for Database Management: The skill utilizes shell commands to automate the provisioning, lifecycle management, and schema configuration of the database container environment.
- External Container Image Usage: It references a specific database container image hosted on an external registry. This is a common practice for containerized local development and preview environments.
- Indirect Prompt Injection Surface: The skill implements a Retrieval-Augmented Generation (RAG) workflow where document content is ingested (SKILL.md, Step 4) and retrieved (SKILL.md, Step 5). This creates a surface where external data enters the system context. The capability inventory includes database access via pyodbc and network access for embeddings. While data values are parameterized, the skill lacks explicit boundary markers or content sanitization before the retrieved data is used in potential downstream prompts.
- SQL String Construction Pattern: The Python implementation uses string interpolation for the vector dimension in SQL statements. While used with a hardcoded constant in this context, the pattern requires careful validation if applied to dynamic inputs to prevent potential syntax issues.
Audit Metadata