skills/openai/plugins/sandbox-sdk/Gen Agent Trust Hub

sandbox-sdk

Pass

Audited by Gen Agent Trust Hub on Jul 14, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [Command Execution]: The skill documents the use of sandbox.exec() and sandbox.runCode() methods. These are core functionalities of the Sandbox SDK designed to execute code within isolated environments, which is the primary purpose of the skill. Contextual examples show these methods being used for tasks like running shell commands or interpreting Python code.
  • [External Resource Integration]: The skill references official documentation and package registries. It instructs users to install the @cloudflare/sandbox package from the public NPM registry and refers to the official Cloudflare Developers site for documentation and API references. These are standard practices for developers using these services.
  • [Tooling and Configuration]: The instructions include standard configuration for Cloudflare Workers (using wrangler.jsonc) and Docker for local development. These configurations define the boundaries and properties of the execution containers (e.g., max_instances, instance_type), ensuring the sandbox environment is properly scoped.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 14, 2026, 11:51 PM
Security Audit — agent-trust-hub — sandbox-sdk