Tool

Inside the Codex sandbox

Page published

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

Coverage timeline

18 Nov 2026openai.comprimary

Single-source analysis — one report is available.

Why it matters

The Codex sandbox limits agent command access, but readable secrets and writable dependencies remain important considerations when configuring permissions and assessing downstream execution risks.

OpenAI explains the Codex sandbox, which restricts agent-run commands through filesystem and network permission profiles. The default workspace profile blocks network access and protects selected directories while allowing broad filesystem reads and workspace writes. OpenAI highlights residual risks, including changes to local dependencies that could later execute outside the sandbox through editor tooling.