The Agent Skills Directory

[SAFE]: The skill documents the use of the sbx CLI tool for managing microVM-based sandboxes. While the documentation mentions sensitive file paths (e.g., ~/.ssh/id_rsa, ~/.aws/credentials) and high-privilege operations (root access within the VM), these are described in the context of isolation boundaries and secure credential management rather than attempts to exfiltrate data.
[SAFE]: Documentation regarding credential management (using sbx secret set) encourages secure practices, such as host-side proxy injection, which prevents real secrets from entering the sandbox environment.
[SAFE]: External references target official Docker documentation (docs.docker.com) and well-known services (Anthropic, Google, OpenAI), which are recognized as trusted sources.
[SAFE]: The skill mentions administrative bypass flags like --dangerously-skip-permissions and --yolo as part of the documented tool's command-line interface but does not use them to execute malicious commands.

running-sbx-sandboxes