As with day 1, the first part of this is easy. I'm "parsing" the input data by using s:tokenize and then unpacking the two element array. Convert the numeric part to a number, then a simple set of s:cases to process things.

~~~:a:unpack [ ] a:for-each ; 
:parse    ASCII:SPACE s:tokenize a:unpack s:to-number swap ; 
 
:process 
  parse 
  'forward [ + ] s:case 
  'down    [ swap [ + ] dip ] s:case 
  'up      [ swap [ - ] dip ] s:case 
  drop-pair ; 
 
(depth,horizontal) 
#0 #0 'day2.input [ process ] file:for-each-line 
* n:put nl 
~~~

The second half is modestly more difficult. As I'm minimizing any use of variables, this is a little messier than I'd prefer, but it's still pretty straightforward.

I'm reusing the "parsing" part from the first half.

My stack values are: aim depth horizontal

~~~:process 
  parse 
  'forward [ [ + ] [ 'abcd 'acbda reorder * + swap ] bi ] s:case 
  'down    [ [ rot ] dip + rot rot ] s:case 
  'up      [ [ rot ] dip - rot rot ] s:case 
  drop-pair ; 
 
(aim,depth,horizontal) 
#0 #0 #0 'day2.input [ process ] file:for-each-line 
 
* n:put nl drop (discard_the_aim_field) 
~~~