signals-scout-ai-observability
Pass
Audited by Gen Agent Trust Hub on Oct 17, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted telemetry data (LLM traces, generations, and evaluations) using tools like
execute-sqlandquery-llm-trace. Malicious content within these traces could potentially attempt to influence the agent's report generation or its persistent state in the scratchpad. - Ingestion points: Untrusted data enters via
execute-sql,query-llm-trace,inbox-reports-list, andscout-scratchpad-searchas described inSKILL.md. - Boundary markers: The instructions do not define specific delimiters or instructions for the agent to ignore embedded commands within the analyzed telemetry content.
- Capability inventory: The skill has the ability to author reports (
scout-emit-report), edit existing reports (scout-edit-report), and write to persistent memory (scout-scratchpad-remember). - Sanitization: No explicit sanitization or filtering logic is provided for the text content within the traces or evaluations before they are processed or cited in reports.
Audit Metadata