reverse-etl-activation

Pass

Audited by Gen Agent Trust Hub on Jul 29, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [SAFE]: No malicious code, obfuscation, or unauthorized data access patterns were detected. The skill consists entirely of instructional markdown and follows best practices for system safety and data integrity.
  • [PROMPT_INJECTION]: The skill describes processing external data from warehouse models, which presents a surface for indirect prompt injection. This risk is managed through a sequence of validation steps, including identity resolution floors and strict field-level mapping guards.
  • Ingestion points: Data warehouse models (BigQuery, Snowflake, Redshift).
  • Boundary markers: None explicitly defined in the text; however, the sync is restricted to predefined mapped fields.
  • Capability inventory: The skill provides logic for agent decision-making and does not include any scripts or subprocess capabilities.
  • Sanitization: Enforces strict type coercion, date formatting, and picklist validation for every field ingested.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 29, 2026, 12:17 AM
Security Audit — agent-trust-hub — reverse-etl-activation