blender-host-sandbox-boundary

Pass

Audited by Gen Agent Trust Hub on Jul 22, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: The skill defines a robust security boundary between a sandbox environment (Hermes) and a host system running Blender. It explicitly instructs the agent to respect these boundaries, providing mechanisms for classified operations (MCP-only, Sandbox-only, Hybrid) and structured handoffs using TRANSFER_REQUIRED receipts. It actively discourages common bypass techniques such as scanning the sandbox for host resources or fabricating results. No malicious patterns, data exfiltration, or unauthorized execution methods were found. All referenced vendor resources (NVIDIA) and architectural patterns are consistent with secure multi-environment automation.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 22, 2026, 03:07 PM
Security Audit — agent-trust-hub — blender-host-sandbox-boundary