signals-scout-web-vitals
Pass
Audited by Gen Agent Trust Hub on Oct 13, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes telemetry data including $host, $pathname, $current_url, and interactionTarget, which are identified as attacker-controllable ingestion points. To mitigate the risk of indirect prompt injection, the instructions mandate the use of SQL-layer sanitization (regex-based character stripping and length capping) to ensure raw external strings do not enter the agent's context. The agent is further instructed to treat these fields as opaque data and to disregard any instructions or prose embedded within them. Capabilities used include execute-sql for data analysis, scout-emit-report for reporting, and scout-lighthouse-audit for automated site checks.
- [SAFE]: The skill demonstrates proactive security awareness and follows best practices for handling untrusted external data. No evidence of malicious intent, unauthorized data exfiltration, or obfuscation was found during the analysis. Primary tool usage and data flows are consistent with the skill's stated purpose of monitoring and reporting on web performance metrics.
Audit Metadata