What probably matters to youSection of the release
Unclear The finding does not say what specifically counts as an 'isolation setting' as opposed to a 'sandbox setting'.
What
Claude Code's sandbox configuration can restrict what a session is allowed to do, and settings changes made by an organization admin can be shown to the user for approval before they take effect. This release gives sandbox isolation its own tracked category throughout that machinery:
The settings-diff and permission-approval summary code now tracks isolationSettings/isolationRows as its own bucket, alongside the existing shell, environment variable, and sandbox settings buckets, both in the diff view and in prompts like "3 sandbox settings x3".
Settings validation now treats isolation the same way it already treated managedMcpServers: if it appears outside of managed (admin-controlled) settings, it's stripped out with a warning that it's "only honored from managed settings and was ignored here."
The managed-settings approval dialog now reads an isolation block with required/egress keys and shows an elided isolation count alongside existing elided command/sandbox counts.
Why
This makes sandbox isolation settings visible and auditable in the same way other managed settings already were, so admins and users can see when isolation-related policy is being changed, enforced only from managed settings, or partially hidden (elided) in an approval prompt.
