I would personally manage the data internally, but you could always look at perltie and write an appropriate tie class that did it for you. Here is an example class:

package Tie::Input::Insertable;

sub PRINT {
  my $self = shift;
  $self->{buffer} = join '', @_, $self->{buffer};
}

sub TIEHANDLE {
  my ($class, $fh) = @_;
  bless {fh => $fh, buffer => ""}, $class;
}

sub READLINE {
  my $self = shift;
  return undef if $self->{eof};
  while (-1 == index($self->{buffer}, $/)) {
    my $fh = $self->{fh};
    my $data = <$fh>;
    if (length($data)) {
      $self->{buffer} .= $data;
    }
    else {
      $self->{eof} = 1;
      return delete $self->{buffer};
    }
  }
  my $pos = index($self->{buffer}, $/) + length($/);
  return substr($self->{buffer}, 0, $pos, "");
}

sub EOF {
  my $self = shift;
  $self->{eof} ||= not length($self->{buffer}) or $self->{fh}->eof();
}

1;
[download]

Save that in Tie/Input/Insertable.pm and then you can do the following:

use Tie::Input::Insertable;
tie *FOO, 'TIE::Input::Insertable', *STDIN;

while (<FOO>) {
  print FOO "Hello, world\n" if /foo/;
  print $_;
}
[download]

Note that I made the tied handle not be the one that I was tying it to. If you do that, then you are playing on the edges of infinite recursion. Just make the handles different and you avoid lots of possible confusion.