Tool

External sandboxes for coding agents

Page published

Earliest dated coverage: 17 Nov 2026 · First observed: 18 Nov 2026 · Latest dated coverage: 17 Nov 2026

Coverage timeline

17 Nov 2026simonwillison.net

Single-source research — one report is available.

Why it matters

External sandboxes give defenders a practical way to limit host damage, network access, and credential exposure when coding agents execute unsafe or prompt-injected instructions.

Prime Radiant released three external-sandbox implementations that isolate agent-executed code while keeping the coding agent outside the sandbox. The implementations provide a shared localhost API, optional HTTP/HTTPS destination allowlisting, and credential replacement for authenticated requests without exposing secrets to the agent.