It seems to me that what you want to do is differentiate between lines that contain a closing brace and those that do not; whitespace should be irrelevant. In that case, how about something like this modification:

script_body:
      '{' script_line(s?) /\s*/s '}'
script_line:
      <skip: ''> /\s*/s /[^}]*/
[download]

The idea is to only skip lines that have no closing brace within them. If the line does have a closing brace,  script_line will fail, and after eating whitesapce, the closing brace will match. (Apologies if this isn't correct syntax; I have never used Parse::RecDescent before.)

If script lines can contain closing braces, then you will need to do a bit more parsing (matching braces, etc.)

-Mark