That's fine except for the incorrect return values

my $grammar = q{
  list : '(' item(s) ')' { $item[2] }
  item : list <commit>   { $item[1] }
       | word
  word : /[^()]*/  
};
[download]

I don't like using q{...} for grammars. Here-docs provide the same functionality, but have fewer issues.

my $grammar = <<'__END_OF_GRAMMAR__';

  list : '(' item(s) ')' { $item[2] }
  item : list <commit>   { $item[1] }
       | word
  word : /[^()]*/  

__END_OF_GRAMMAR__
[download]

I strongly recommend the use of use strict and use warnings in your grammar. The ones in your main program won't affect the code in your grammar.

my $grammar = <<'__END_OF_GRAMMAR__';
  {
    use strict;
    use warnings;
  }

  list : '(' item(s) ')' { $item[2] }
  item : list <commit>   { $item[1] }
       | word
  word : /[^()]*/  

__END_OF_GRAMMAR__
[download]

Finally, I'm not sure if trailing <commit> directives are of any use. I'll have to look into that.

Update: If the outside parens are optional, you get:

  list : item(s)
  item : '(' <commit> list ')' { $item[3] }
       | word
  word : /[^()]*/
[download]

And if you don't want empty strings:

  list : item(s?)
  item : '(' <commit> list ')' { $item[3] }
       | word
  word : /[^()]+/
[download]