troubleshoot-postgresql

Pass

Audited by Gen Agent Trust Hub on Jul 10, 2026

Risk Level: SAFE
Full Analysis
  • [PROMPT_INJECTION]: No attempts to override agent behavior or bypass safety guidelines were found. The instructions are focused on database diagnostics and follow a structured troubleshooting logic.
  • [DATA_EXFILTRATION]: No evidence of hardcoded credentials or unauthorized data transmission. The skill uses standard MCP tools to query telemetry from a Netdata instance, which is its intended purpose. No access to sensitive local files (~/.ssh, etc.) is attempted.
  • [REMOTE_CODE_EXECUTION]: The skill does not download or execute external scripts, binary packages, or remote code. All operations are confined to querying the Netdata monitoring interface.
  • [COMMAND_EXECUTION]: No suspicious or high-risk shell commands are present. The skill relies on pre-defined MCP tools (list_metrics, query_metrics, find_anomalous_metrics) for metric retrieval rather than direct system-level command execution.
  • [SAFE]: The skill represents a legitimate administrative tool. It is well-documented, follows least-privilege principles by using specific monitoring contexts, and aligns with its stated purpose of PostgreSQL performance analysis.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 10, 2026, 11:23 AM
Security Audit — agent-trust-hub — troubleshoot-postgresql