Is making the first part non-greedy not enough ...

my $list = "This is my list";
$list =~ m/^(.+?)   # The 'rest' (Everything before last word)
            (\w+)   # Last 'word' (string of contiguous word character
+s)
            \W*$    # Possible non-word characters at end of string
          /x;
[download]

... (would it fail on some other string)?

Much later. For English language, \w is not inclusive enough (lacks hyphen) and includes too much (includes underscore & digits). Short of a proper grammar based parser, I would rather use word regex which addresses that ...

$word_re = qr{ (?: & | -? [a-zA-Z]+ [a-zA-Z-]* ) }x;
[download]

... is still incomplete as it does not deal with accented characters; periods in a title; acronyms with spaces and/or periods, among other things.