You'll noticeTier: how much it should matter to you
2Useful: my rating, 1 to 5
2Signal: worth watching, 1 to 5
PermissionsArea: what it touches
ImprovementsKind: in v2.1.281,
ImprovementsSection of the release
What
When Claude Code asks your permission before doing something, you can usually choose to always allow it, which saves a rule so you are not asked again. For prompts caused by a safety check that is marked as not approvable by the automatic classifier (classifierApprovable false), that saved rule is now dropped. You can still allow the action once, but no "always allow" rule is kept.
Why
These prompts will keep appearing each time the same safety check is triggered, so every such action gets your explicit approval.
