sota-databases

Pass

Audited by Gen Agent Trust Hub on Nov 6, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill possesses a surface for indirect prompt injection as it is designed to ingest and audit untrusted user-provided database schemas and queries. Evidence includes: 1. Ingestion point: User intent and schema/queries processed in AUDIT mode (SKILL.md). 2. Boundary markers: Detailed audit checklists and finding formats are provided to structure agent behavior. 3. Capability inventory: The skill utilizes code analysis and reporting capabilities. 4. Sanitization: Explicit instructions to verify all claims against the actual code are included.
  • [SAFE]: The skill is educational and procedural, containing no executable malicious code, obfuscated scripts, or persistence mechanisms.
  • [SAFE]: All referenced external tools, such as Flyway, Alembic, and Atlas, and database engines like Postgres, Redis, and Qdrant, are well-known, trusted, and reputable sources within the developer community.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 6, 2026, 10:22 AM