database-design-patterns
Pass
Audited by Gen Agent Trust Hub on Apr 2, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill serves as a purely instructional guide for database architecture. It contains no executable scripts, remote downloads, or hidden commands.
- [CREDENTIALS_UNSAFE]: While connection strings for Prisma and PgBouncer are mentioned in the instructional text, they correctly use generic placeholders (e.g., 'user:pass', 'host:6432') and do not expose real secrets.
- [PROMPT_INJECTION]: No evidence of jailbreak attempts, system prompt extraction, or instructions to ignore safety filters. The use of 'IMPORTANT' and 'Warning' keywords is strictly confined to technical database safety (e.g., avoiding locks or data drift).
- [DATA_EXFILTRATION]: No network-enabled tools are requested in the frontmatter, and no commands attempt to access sensitive system files like SSH keys or environment variables. All SQL examples focus on public schema metadata and standard tables.
- [REMOTE_CODE_EXECUTION]: The skill does not perform any dynamic package installations or fetch scripts from external sources. It relies entirely on its own documentation and the agent's pre-existing tools.
Audit Metadata