You may be looking for the /g option to s///:

my $str="abc t[poop1] def [doop2] [woop3] xyz";
$str =~ s/\[.*?\]/~~/g;
print "$str\n";
__END__
abc t~~ def ~~ ~~ xyz
[download]

There was nothing about newlines being disallowed in the brackets. The following is faster and more likely to match correctly:

s/\[.*?\]/~~/sg
[download]

But that's risky, especially if you use it a larger pattern. The /.*?/ could match a "[" or a "]". That's obviously undesirable. The following pattern avoids that problem:

s/\[[^\[\]]*\]/~~/g;
[download]

... or, one of several possible alternates, on the off chance (ymmv) you find this will be easier to read someday, down the road:

my $regex = qr'(?:\[.*?\])'; compile regex; use non-capture grouping
$str =~ s/$regex/~~/g;
[download]

You are advocating building small patterns so that they can be combined into larger patterns, right? If so, your code is buggy (and I don't mean just the superfluous (?:)).

>perl -E"$sqr=qr/\[.*?\]/; $_='ab[cd]ef[gh]'; s/$sqr$//; say"
ab
[download]

Fixed:

>perl -E"$sqr=qr/\[[^\[\]]*\]/; $_='ab[cd]ef[gh]'; s/$sqr$//; say"
ab[cd]ef
[download]

 qr// automatically wraps its expression in a non-capturing group:

>perl -wMstrict -le
"my $regex = qr'(?:\[.*?\])';
 print $regex;
"
(?-xism:(?:\[.*?\]))
[download]