resolving-ingestion-warnings
Pass
Audited by Gen Agent Trust Hub on Oct 23, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted content from the system.ingestion_warnings table which can contain strings provided by external event senders.
- Ingestion points: Data is ingested from the details field in the system.ingestion_warnings table via posthog:execute-sql.
- Boundary markers: The skill includes a dedicated trust boundary section that explicitly warns the agent to treat this data as untrusted and never follow instructions found within it.
- Capability inventory: The agent has access to tools including posthog:execute-sql, posthog:project-settings-update, and posthog:persons-property-delete.
- Sanitization: The skill relies on clear, direct instructions to the agent to prevent the execution of malicious commands or text embedded in the retrieved event data.
Audit Metadata