Before submitting
Area
Not sure
Problem or use case
Claude Code recently introduced "Auto Mode" where a safety model reviews recent messages and tool calls for safety, auto-approving those it deems safe and denying those it deems dangerous.
This is a nice middle-ground between yolo mode and approve everything by hand mode.
Proposed solution
Essentially recreate Claude Code's auto mode
Why this matters
It somewhat improves safety whilst keeping most of the usability benefits of not having to manually approve tool calls
Smallest useful scope
I think it's probably all or nothing if I'm honest
Alternatives considered
No response
Risks or tradeoffs
No response
Examples or references
No response
Contribution
Before submitting
Area
Not sure
Problem or use case
Claude Code recently introduced "Auto Mode" where a safety model reviews recent messages and tool calls for safety, auto-approving those it deems safe and denying those it deems dangerous.
This is a nice middle-ground between yolo mode and approve everything by hand mode.
Proposed solution
Essentially recreate Claude Code's auto mode
Why this matters
It somewhat improves safety whilst keeping most of the usability benefits of not having to manually approve tool calls
Smallest useful scope
I think it's probably all or nothing if I'm honest
Alternatives considered
No response
Risks or tradeoffs
No response
Examples or references
No response
Contribution