backend-observability

Pass

Audited by Gen Agent Trust Hub on Oct 7, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: The skill is entirely instructional, providing a set of guidelines and best practices for the AI agent to follow when writing backend observability code. It does not contain any executable scripts, shell commands, or network-enabled operations.
  • [SAFE]: The instructions explicitly mandate PII and secret redaction at the infrastructure level (e.g., logger deny-lists and Sentry's beforeSend hooks), which is a positive security practice to prevent accidental data exposure.
  • [PROMPT_INJECTION]: No prompt injection or behavior override patterns were detected. The use of "IMPORTANT" or similar emphasis is confined to technical best practices like correlation and redaction.
  • [REMOTE_CODE_EXECUTION]: No remote code execution patterns, package installation commands, or dynamic code execution were found. References to standard observability libraries (Pino, Sentry, OpenTelemetry) are for instrumentation purposes only.
  • [DATA_EXFILTRATION]: No data exfiltration patterns or sensitive file access were detected. The skill correctly identifies sensitive data (tokens, PANs, auth headers) as items that must be redacted.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 7, 2026, 11:26 PM