In general that is not really a job for a regex, it is a job for a parser. The issue is that generally it is messy to handle balanced text with regexen (not impossible with Perl regexes, but messy). If the commands were simple and it was not possible to have [, ], { or } in the 'commands etc' then the job is fairly trivial, but that is not the case.

Your simplest option may be to roll your own simple recursive decent parser for the job. Consider:

use strict;
use warnings;

while (defined (my $line = <DATA>)) {
    next unless $line =~ /^(catch\s*)\[(.*)/s;

    my $prefix = $1;
    my ($body, $tail) = parse ($2);

    $line = "${prefix}{$body}$tail";
} continue {
    print $line;
}


sub parse {
    my ($tail) = @_;
    my $outStr = '';

    while (defined $tail) {
        if ($tail =~ s/^([^[]*)\[(.*)//s) {
            my $prefix = $1;
            my $body;

            ($body, $tail) = parse ($2);
            $outStr .= "$prefix [$body]";
            next if length $tail;
        } elsif ($tail =~ s/^([^\]]*)](.*)//s) {
            return "$outStr$1", $2;
        }

        $outStr .= $tail;
        $tail = <DATA>;
    }

    return $outStr, '';
}


__DATA__
catch [ this; that; the; other; ]

some random stuff that isn't a catch

catch [
this[];
that + 1;
the other;
]
[download]

Prints:

catch { this; that; the; other; }

some random stuff that isn't a catch

catch {
this [];
that + 1;
the other;
}
[download]

although there are likely to be many special cases (quoted or commented unbalanced brackets for example) that mean that a heavier duty parser like Parse::RecDecent may be what you really need.