A majority of Claude interruptions are asking whether it's okay to run some
command. This is why --dangerously-skip-permissions is so powerful: we can
achieve so much more per unit prompt.
But why is that the choice? Why must we choose between becoming a pedantic babysitter or giving Claude's actions dangerous free rein over our system?
Clash gives us the best of both worlds. It allows sandboxing what matters: the actions Claude is performing. Every tool, every command, independently.
The Clash plugin for Claude runs in two phases,
- matching and deciding what actions Claude can perform, and
- dynamically selecting the sandbox in which that action will be performed.
Unlike wrapping our agent in a container or VM, we are in control of which tool gets what access.
With Clash, we get the productivity of --dangerously-skip-permissions without
losing the control.