bigquery

Pass

Audited by Gen Agent Trust Hub on Jul 13, 2026

Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [PROMPT_INJECTION]: The skill processes user-supplied queries and external API data from providers like Amplitude, PostHog, and Mixpanel, creating a surface for indirect prompt injection.\n
  • Ingestion points: User requests for warehouse SQL, BigQuery tables, or specific metric definitions (SKILL.md).\n
  • Boundary markers: The skill does not provide explicit instructions or delimiters to isolate user-supplied data from the agent's internal reasoning.\n
  • Capability inventory: The skill utilizes bigquery --sql for warehouse interaction and run-code for complex data aggregation.\n
  • Sanitization: No explicit instructions for sanitizing or validating data retrieved from external sources before use in subsequent actions.\n- [COMMAND_EXECUTION]: The skill references a run-code tool used to aggregate data for corpus-wide event analysis. This capability involves the execution of dynamically generated code within the agent's environment.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 13, 2026, 04:51 AM
Security Audit — agent-trust-hub — bigquery