postgresql-admin

Pass

Audited by Gen Agent Trust Hub on Feb 17, 2026

Risk Level: LOW
Full Analysis
  • SAFE (SAFE): Comprehensive analysis of the instructions, documentation, and validation scripts revealed no security issues.
  • PROMPT_INJECTION (SAFE): Instructions are strictly focused on database administration. No override/bypass markers or jailbreak patterns were detected.
  • DATA_EXFILTRATION (SAFE): No unauthorized network calls or sensitive file access patterns. SQL examples provided are restricted to administrative tasks and use placeholders.
  • INDIRECT_PROMPT_INJECTION (LOW): The skill has an attack surface for handling database identifiers; however, this is mitigated by strict parameter validation (^[a-z][a-z0-9_]*$) and enum-based operation selection.
  • DYNAMIC_EXECUTION (SAFE): The provided Python script (scripts/validate.py) uses yaml.safe_load() and performs standard file existence checks without executing arbitrary strings or untrusted code.
Audit Metadata
Risk Level
LOW
Analyzed
Feb 17, 2026, 12:22 PM