sql-query

Pass

Audited by Gen Agent Trust Hub on Mar 30, 2026

Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [PROMPT_INJECTION]: The skill documents a common pattern where user-provided natural language is converted into SQL queries using an AI client. This implementation is susceptible to indirect prompt injection, where a crafted user prompt might manipulate the resulting SQL to retrieve unauthorized information.
  • Ingestion points: User-supplied questions processed by AIClient.text_to_sql in SKILL.md.
  • Capability inventory: SqlClient.get executes the generated SQL string directly against datasets.
  • Boundary markers: No delimiters or protective instructions are used in the prompt interpolation examples.
  • Sanitization: No validation or filtering of the AI-generated SQL is performed before it is executed.
  • [COMMAND_EXECUTION]: The skill enables the execution of raw SQL commands via the SqlClient class. This provides high-privilege access to database contents and should be used with caution when processing dynamic input.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 30, 2026, 08:38 PM
Security Audit — agent-trust-hub — sql-query