signals-scout-mcp-tool-calls
Pass
Audited by Gen Agent Trust Hub on Nov 2, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes telemetry data (error messages, tool names, intents) originating from external clients, creating an indirect prompt injection surface.
- Ingestion points:
execute-sqlqueries inSKILL.mdandreferences/queries.mdfetch telemetry properties such as$mcp_error_messageand$mcp_intent. - Boundary markers: Instructions require quoting telemetry as "short untrusted snippets" and truncating content to prevent injection.
- Capability inventory: Capabilities include report generation via
scout-emit-reportandscout-edit-report. - Sanitization: Explicit guidelines in the "Untrusted data" section instruct the agent to treat telemetry strictly as data and ignore any embedded directives.
Audit Metadata