syncing-migrated-self-managed-tables
Pass
Audited by Gen Agent Trust Hub on Nov 6, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill instructions follow security best practices by explicitly prohibiting the agent from requesting or transmitting sensitive credentials (access keys/secrets) through the chat or tool interface. It also mandates a human-in-the-loop confirmation step before performing any actions that would initiate billing or change table configurations.
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from external tool outputs, such as table schema metadata and error messages retrieved from the PostHog API. While this creates a theoretical surface for indirect prompt injection if the source data were maliciously manipulated, the risk is mitigated by the requirement for explicit user consent before any state-changing operations.
- Ingestion points: Metadata and error messages from
external-data-schemas-retrieveandexternal-data-schemas-list. - Boundary markers: None defined for the interpolation of tool outputs.
- Capability inventory: Tool access for updating schema configurations (
external-data-schemas-partial-update) and triggering data reloads (external-data-schemas-reload). - Sanitization: No explicit sanitization or escaping of the ingested error strings is described.
Audit Metadata