agent-sandbox
Pass
Audited by Gen Agent Trust Hub on Oct 12, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill focus is entirely on security tooling and isolation. The provided code templates for the
AgentSandbox(filesystem-level) andDockerSandbox(container-level) implement defensive patterns such as path allowlisting/denylisting, command filtering, and resource usage caps. No suspicious network activity, credential harvesting, or prompt injection attempts were detected in the instructions or example implementations.
Audit Metadata