The new version of Claude Code adds a new "More Dangerous" control item. After turning it on, you can turn off the confirmation prompt for key directory deletion operations.

#ARTIFICIAL INTELLIGENCE Claude Code The new version adds a "more dangerous" option, which can be executed even if the generated command may delete the user's home directory. For example, commands such as rm and rmdir targeting the user's home directory require manual approval or direct refusal to execute in normal mode. With the new mode enabled, dangerous commands like these can be executed without human approval. The new option is primarily intended for developers who want their Claude Code to run continuously in an unattended environment, such as automated tasks or temporary testing environments. After closing the confirmation prompt, the agent can continue to execute when it encounters related deletion commands, but the user also needs to bear the consequences of accidental deletion. View the full text: https://ourl.co/127838