Disclosed weakness · no CVE recorded

OpenAI Agents SDK workspace checks could execute shell startup files

OpenAI · OpenAI Agents SDK for Python · sandbox helpers

First documented disclosure · Review checkpoint

Summary

Internal file checks and snapshot creation could execute shell startup files from the workspace.

Severity and exploitation evidence

No CVSS score recorded in this review.

Exploitation not established: The reviewed sources describe a correction or controlled reproduction; malicious exploitation against users is not established.

Exploitation evidence source

Technical severity is not a measure of observed attacks or of risk to your deployment. Attempts, demonstrations, and successful compromise retain the qualifications in the source.

How the weakness works

Helper operations invoked login shells, which loaded startup files before doing the intended internal operation.

Conditions required

An untrusted party or prior process can influence workspace shell startup files used by a sandbox helper.

Potential impact

Startup commands could run during internal file or snapshot operations. This source does not establish a host sandbox escape.

Affected versions and configurations

The reviewed upstream report does not establish a complete affected-release range.

Fix and mitigation status

0.25.0 contains the correction; the earliest fixed prerelease is unverified.

Practical next steps

Upgrade openai-agents to 0.25.0 and review the release’s separate sandbox-provider migration requirements.

Operating lesson

Internal housekeeping should avoid executing workspace-controlled initialization code.

Evidence and limits

The October 21 PR is the earliest reviewed public disclosure. Tagged verification: file checks and snapshot helper commands use non-login shells instead of loading workspace startup files.

Sources