modeling-warehouse-foundations
Pass
Audited by Gen Agent Trust Hub on Oct 29, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill identifies and provides instructions for managing an attack surface where malicious instructions could be embedded in ingested metadata from the data warehouse.
- Ingestion points: Event taxonomy, property keys/values, and metric descriptions retrieved via
system.information_schemaor data-reading tools (referenced inreferences/governance.mdandSKILL.md). - Boundary markers: The instructions explicitly mandate treating ingested taxonomy and descriptions as "quoted, untrusted data" and ignore commands embedded in free-text fields.
- Capability inventory: The skill utilizes capabilities for persistent state change via
posthog:view-createandposthog:view-materialize(referenced inreferences/posthog-views.md). - Sanitization: The skill implements a verification step, requiring the agent to confirm specific events and properties with the user before performing any persistent write operations.
Audit Metadata