It is inefficient to re-write your entire target file once for each "drop word". Luckily, there is a better algorithm; read your 'drop-words' into a hash, use the hash as a lookup table, then run through the words in your 'temp.txt' file one time. Every time you find that a word in the 'temp.txt' file exists within your hash, drop the line and move onto the next. Any line where you don't come across a drop-word, print the line to a new file.

use strict;
use warnings;
use autodie;
use List::MoreUtils qw( any );

my %drop_words;

open my $words_ifh, '<', 'words.txt';

while( <$words_ifh> ) {
    $drop_words{ ( split /\s+/, $_, 2 )[0] } = 1;
}

close $words_ifh;

open my $temp_ifh,   '<', 'temp.txt';
open my $result_ofh, '>', 'temp_mod.txt';

while( <$temp_ifh> ) {
    chomp;
    next if any { exists $drop_words{$_} } split /\s+/;
    print {$result_ofh} $_, "\n";
}

close $temp_ifh;
close $result_ofh;
[download]

If you're not interested in using the non-core module List::MoreUtils, you could achieve about the same goal by changing line 21 to look like this:

next if defined first { exists $drop_words{$_} } split /\s+/;
[download]

...and replacing line 4 with use List::Util qw(first); (a core module).