debugging-failed-data-quality-checks
Pass
Audited by Gen Agent Trust Hub on Oct 6, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill uses the
posthog:execute-sqltool to run HogQL queries that are retrieved dynamically from thecompiled_queryfield inposthog:data-quality-check-results. This is a standard functionality for replaying failing data checks. - [INDIRECT_PROMPT_INJECTION]: The skill ingests data from external sources, specifically the
compiled_queryanderrorfields from theposthog:data-quality-check-resultstool. This creates a potential surface for indirect injection if these fields were to contain malicious instructions. However, the agent's capabilities are restricted to the vendor's database tools, and the instructions include safety checks such as requiring user confirmation before changing check severities. - Ingestion points:
posthog:data-quality-check-resultstool output. - Boundary markers: None present.
- Capability inventory: Includes
posthog:execute-sql,posthog:data-quality-check-update, andview-runtools. - Sanitization: The instructions do not specify explicit sanitization, relying on the database engine's parsing of the HogQL query.
Audit Metadata