troubleshoot-nvidia-gpu

Pass

Audited by Gen Agent Trust Hub on Jul 10, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: The skill operates entirely within the scope of querying Netdata monitoring metrics through standard MCP tools. No unauthorized command execution, file system access, or network exfiltration patterns were identified.
  • [SAFE]: All metric contexts (e.g., nvidia_smi.*) and tool recommendations are consistent with official Netdata collector outputs and documentation.
  • [SAFE]: The skill includes an ingestion surface for external monitoring data via MCP calls. 1. Ingestion points: Metrics and anomalies retrieved via 'query_metrics' and 'find_anomalous_metrics' (SKILL.md). 2. Boundary markers: Absent. 3. Capability inventory: The skill utilizes read-only MCP queries for data retrieval and lacks any file-write, subprocess execution, or direct network capabilities. 4. Sanitization: The skill relies on the structured nature of monitoring data and intrinsic agent safety filters.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 10, 2026, 11:23 AM
Security Audit — agent-trust-hub — troubleshoot-nvidia-gpu