Sandboxd Code Run

execute-code and execute-command are mostly distinct: one runs language-specific code, the other runs arbitrary shell commands. However, overlap exists because shell commands can also run code (e.g., python -c), creating slight ambiguity for some tasks.

Both tools follow a consistent execute_<noun> pattern (execute-code, execute-command) using verb_noun with hyphens. The naming is predictable and easy to parse.

With only two tools, the set feels thin for a sandbox execution service. While both tools earn their place, the count is borderline (1-2 tools) and could benefit from additional utilities like environment management or artifact retrieval.

The core operations for one-shot sandbox execution are covered: running code in multiple languages and running shell commands, both returning stdout, stderr, exit codes, and artifacts. Minor gaps include no tools for managing dependencies or inspecting the sandbox environment, which agents might need for complex workflows.