Enhancement: Whitelisting

Once Claude Code gets going it is easy to let it run by itself for a few minutes... unless I just began the session and need to allow it to cd ls cat grep etc., etc., over and over.

If we could whitelist a set of commands, and/or directories on a per-project basis it would go a long way.

Examples:
Whitelist low-security risk bash commands in /app
Whitelist cat and grep in /var/log/

Related: Claude Code seems to go back and forth trying to use Read (what appears to be the internal function) vs tail cat and grep. It will do the same thing writing files, sometimes using the builtin function and sometimes trying to echo a file into existence. Consistent use of the relevant functions would cut down on the number of times users need to OK a command.