NYC

monitoring-observability

Fail

Audited by Gen Agent Trust Hub on Feb 15, 2026

Risk Level: HIGH
Full Analysis
  • [Prompt Injection] (SAFE): No patterns attempting to override system prompts or bypass safety filters were found.
  • [Data Exposure & Exfiltration] (LOW): The skill correctly implements security best practices by including a 'MUST NOT' constraint against logging sensitive information like passwords or API keys. It logs standard metadata (IP, User-Agent) which is standard for observability.
  • [Indirect Prompt Injection] (LOW): The skill processes external data, creating a potential surface for log injection.
  • Ingestion points: 'req.path' and 'user-agent' headers are ingested in the Winston middleware (SKILL.md).
  • Boundary markers: Absent.
  • Capability inventory: Local file writing (logs) and console output.
  • Sanitization: Not explicitly shown in the snippets.
  • [External Downloads] (SAFE): All external links point to trusted documentation for Prometheus, Grafana, and Google SRE resources.
Recommendations
  • Contains 1 malicious URL(s) - DO NOT USE
Audit Metadata
Risk Level
HIGH
Analyzed
Feb 15, 2026, 10:42 PM