auditing-warehouse-data-health
Pass
Audited by Gen Agent Trust Hub on Apr 27, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill uses read-only tools to perform audits and explicitly prohibits autonomous remediation, requiring user confirmation for any follow-up actions.
- [SAFE]: No prompt injection, obfuscation, or data exfiltration patterns were detected in the skill instructions or metadata.
- [SAFE]: Indirect Prompt Injection surface is assessed as safe: (1) Ingestion points: Data enters via platform-specific warehouse health and schema tools. (2) Boundary markers: Instructions specify human-readable tables for output. (3) Capability inventory: All tools are read-only; no subprocess, evaluation, or file-writing tools are available. (4) Sanitization: The agent's role is strictly limited to reporting and triage.
Audit Metadata