text-to-sql

Pass

Audited by Gen Agent Trust Hub on Mar 3, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The skill ingests natural language input to generate executable SQL code, creating a surface for indirect prompt injection. This is mitigated by 'Iron Laws' that require parameterized queries and validation of all identifiers against a provided schema.
  • [EXTERNAL_DOWNLOADS]: The documentation suggests the use of 'promptfoo' for evaluation. Promptfoo is a widely recognized and trusted open-source tool for testing LLM outputs.
  • [SAFE]: The skill incorporates safety best practices such as mandatory row limits on SELECT queries and explicit confirmation requirements for DELETE or DROP statements.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 3, 2026, 01:18 PM