The Agent Skills Directory

[Command Execution]: The skill utilizes the sandbox.exec and sandbox.terminal APIs to interact with isolated Linux environments. This is a core feature of the SDK designed to facilitate command execution and process management within a controlled sandbox.
[External Downloads]: The instructions reference official Cloudflare documentation and GitHub repositories for lifecycle management and code examples. These resources are from the established vendor of the tool and provide necessary technical context for the agent.
[Indirect Prompt Injection Surface]: The skill instructs the agent to inspect the user's application files (such as dependency manifests and container configurations). While this involves ingesting external data, the skill includes explicit checks to ensure that the environment configuration is consistent and safe for deployment.
Ingestion points: Inspection of local project files (e.g., package.json, Dockerfiles).
Boundary markers: None explicitly defined in the instruction text.
Capability inventory: Execution of shell commands via sandbox.exec, file system manipulation, and network tunnel creation.
Sanitization: Relies on the underlying Sandbox SDK's isolation for safety.

sandbox-stable