gcp-bigquery

Pass

Audited by Gen Agent Trust Hub on Sep 8, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill documents the use of BigQuery AI functions like AI.GENERATE, AI.GENERATE_BOOL, and AI.SIMILARITY. These functions process data from table columns (e.g., review_text, description) directly through LLMs (Gemini). If these columns contain attacker-controlled content, they represent a surface for indirect prompt injection. The severity is LOW as this is an inherent risk of the documented capability rather than a flaw in the skill itself.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 8, 2026, 05:54 PM
Security Audit — agent-trust-hub — gcp-bigquery