signals-scout-product-analytics
Pass
Audited by Gen Agent Trust Hub on Nov 2, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted metadata and event data from the analytics project, such as insight names, event names, and query definitions. This content is interpolated into reports and scratchpad memory, creating a surface for indirect prompt injection where malicious project data could influence the agent's generated summaries.\n
- Ingestion points: Event data from
top_events, insight definitions fromsystem.insights, and custom query results fromexecute-sql(found inSKILL.md).\n - Boundary markers: Absent; untrusted strings are used directly in prompts and report fields.\n
- Capability inventory: The agent has the ability to emit/edit reports (
emit_report,edit_report) and execute database queries (execute-sql).\n - Sanitization: No explicit sanitization or filtering of event/insight names is documented before they are included in report evidence.\n- [COMMAND_EXECUTION]: The skill utilizes the
execute-sqltool to perform ad-hoc queries against the product analytics database. This is used for orientation, volume validation, and data exploration, which is an expected capability for a product analytics scout.
Audit Metadata