Your regex m/<a href="\s*(.+)\s*"/g is too greedy - the .* gobbles up as much of the string as possible, so for the line

<a href="link1">link 1</a>Some text<a href="link2">link 2</a>
[download]

$1 will be

link1">link 1</a>Some text<a href="link2"
[download]

instead of what you expected:

link1
[download]

What you want to do in the first step is to make the match non-greedy, or make it stop at the first double-quote it encounters:

m/<a href="\s*(.+?)\s*"/g # non-greedy due to .*?
m/<a href="\s*([^"]+)\s*"/g # stops at first "
[download]

Neither of these solutions will work with arbitrary HTML, so if you want to rewrite HTML pages from more than one source, you will likely be better off with a HTML parser like HTML::TokeParser::Simple.

The question of how to stuff the changes back into the string is best answered by printing out the tokens you get from HTML::TokeParser::Simple, but if you first want to stick with your regular expression approach, the s/// operator will do a search-and-replace operation:

    $html =~ s!(<a href="\s*)([^"]+)(\s*")!${1}foo$2$3!g;
[download]