diagnosing-failed-warehouse-syncs
Pass
Audited by Gen Agent Trust Hub on Oct 25, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill relies on analyzing
latest_errormessages retrieved from external data sources and schemas to guide its troubleshooting workflow. - Ingestion points: Diagnostic data is ingested from
external-data-sources-retrieveandexternal-data-schemas-retrieveAPI calls, which return strings potentially controlled by external systems. - Boundary markers: The instructions lack specific guidance on how to delimit or sanitize these external error strings within the agent's prompt context.
- Capability inventory: The skill has access to high-impact tools, including
external-data-schemas-resync(destructive data wipe),external-data-schemas-delete-data(table drop), andexternal-data-sources-partial-update(credential modification). - Sanitization: No explicit sanitization logic is described for the external content before it influences the agent's logic for selecting a recovery action.
- Mitigation: The skill includes a mandatory confirmation step (Step 5) requiring explicit user approval before performing any destructive operations, which serves as a human-in-the-loop safeguard.
Audit Metadata