Like this?

#! perl -slw
use strict;

my $data = do{ local $/; <DATA> };

$data =~ s[(if.*?{)]{
    ( my $subscope = $1 ) =~ s[\n][]g;
    $subscope;
}smge;

print $data;

__END__
if (@Agent eq 'XX')
{
  @Agent='{XX}'
}

if (
    @Agent
    eq
    'XX'
) {
  @Agent='{XX}'
}
[download]

Prints

C:\test>junk7
if (@Agent eq 'XX'){
  @Agent='{XX}'
}

if (    @Agent    eq    'XX') {
  @Agent='{XX}'
}
[download]

Looks good, but some of the syntax is a bit alien to me (in particular the use of [] and {} in '$x=~ s.{.}smge' Is there a good text/url on this kind of syntax?

Those are alternate delimeters. See the "substitution operator" in perlop. Also read perlre. There is a wealth of information on perldoc.perl.org.

You can use any pair of non-alphanumeric, non-whitespace charaters as delimiters of a regex pattern. So these regexes are equivalent:

s/a/b/;
s"a"b";
s^a^b^;
s{a}{b};
s(a)[b];
[download]

Only ' is an exception, this character prevents variable interpolation if used as delimiter

Advantage is that you can use / inside them without escaping the character. Also it could be argued that {} [] or () as delimiters just look better, especially for complex patterns. Edit: Deleted one erraneous example found by ikegami

{
    local $/;  # go to slurp mode for input
    open( I, "<", "my_source_code" ) or die "my_source_code: $!";
    $_ = <I>;
    close I;

    s/(if.*?)\s+{/$1 {/g;    # fix spacing/line-breaks for "if .... {"

    open( O, ">", "my_new_source_code" ) or die "my_new_source_code: $
+!";
    print O;
    close O;
}
[download]

if ( "blah { blah } blah" )

{
...
[download]

     s/(if.*?\))[ \r\t]*\n\s+{/$1 {/;  # fix spacing/line-breaks for "
+if (...) {"
[download]

(updated second s/// example to include "\r")

Thanks for the response. Though correct, I am trying to get around multiple passes and extra complexity. Ideally I am looking for something to first identify a series of scopes via REGEX and then do a regular search and replace on these scopes ignoring anything outside those scopes. This is a common problem to solve in config management. I am thinking there might be an easy way to do this since it is built in vi which came long before perl. There isn't anything in the new features additions in regex with 5.10 are there?

There isn't anything in the new features additions in regex with 5.10 are there?

I don't know, but you can see what's new for 5.10 at perldelta.

Maybe. I am looking to parse Netcool rules files which is basically a glorified if-then-else/switch-case-default stmts with assignments @x=,%x=,$x= and a handful of other functions. Also there are includes that need to be followed. I'm thinking the scope is small enough and the variances large enough that a tidy project would be larger than brute force coding.

Well that certainly explains where @Agent eq 'XX' is coming from. And I just assumed that you didn't the difference between an array and a scalar. :-)