Postgres MCP Pro
Pass
Audited by Gen Agent Trust Hub on Aug 14, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSPROMPT_INJECTIONNO_CODE
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill documentation instructs users to install third-party software from external sources, including a Docker image (crystaldba/postgres-mcp) and a Python package (postgres-mcp). These software sources are not from recognized trusted organizations or the skill author's own domain.
- [PROMPT_INJECTION]: The skill handles data from external PostgreSQL databases, representing an indirect prompt injection surface where untrusted data in query results or schemas could influence agent behavior.
- Ingestion points: SQL query results, database schema metadata, and query execution plans.
- Boundary markers: No explicit delimiters or instructions to ignore embedded commands are present.
- Capability inventory: Database query execution, performance tuning recommendations, and system monitoring.
- Sanitization: No visible mechanisms to sanitize or validate content returned from the database.
- [NO_CODE]: The skill consists exclusively of documentation and metadata without providing any executable scripts or binary files within the package itself.
Audit Metadata