you can either use combinator (2 words) and combinator3 (3 words): https://hashcat.net/wiki/doku.php?id=has...ombinator3 one after the other (first 2 words, then 3 words)
for 2 words (only 2 words !) you could use a faster built-in alternative (-a 1) : https://hashcat.net/wiki/doku.php?id=combinator_attack
... but if you are too lazy to switch from 2 to 3 words you could also use princeprocessor.
for combinator, combinator2 and princeprocessor you should use stdin (-a 0 with pipes) which could be a little bit slower. for -a 1 you do not need to use any pipes but as said it will be only working for combining 2 word lists (but one list could contain the combination of 2 words already)
The good thing is also that combining 3 words is essentially combining the already combined list of 2 words again: (1+1) + 1 = 3 too , so you could combine a list, say the resulting list is dict2.txt with combinator or -a 1 --stdout, and do the 3 word combination again with -a 1 dict2.txt dict1.txt
There is one additional thing to consider, some people want to try really, really hard to avoid duplicate words in there combination, like YellowBlueYellow or YellowYellowPurple. Depending on your hash type, hardware and size of the list it could make sense, but often the speed difference (filtering vs -a 1 speedup) makes the "few additional wrong combinations" negligible and therefore it could be faster to just NOT filter out duplicate words (it depends !).
To sum up, princeprocessor should/could be the easiest and most straightforward technique, but -a 1 (depending on hash type and hardware etc) might sometimes be the fastest one, with the disadvantages that you need to run at least 2 commands (switching from 2 combinations to 3 combinations) and maybe saving a temporary list to disk (which also could be huge depending on the number of original words, disk I/O is always bad in some way !)
for 2 words (only 2 words !) you could use a faster built-in alternative (-a 1) : https://hashcat.net/wiki/doku.php?id=combinator_attack
... but if you are too lazy to switch from 2 to 3 words you could also use princeprocessor.
for combinator, combinator2 and princeprocessor you should use stdin (-a 0 with pipes) which could be a little bit slower. for -a 1 you do not need to use any pipes but as said it will be only working for combining 2 word lists (but one list could contain the combination of 2 words already)
The good thing is also that combining 3 words is essentially combining the already combined list of 2 words again: (1+1) + 1 = 3 too , so you could combine a list, say the resulting list is dict2.txt with combinator or -a 1 --stdout, and do the 3 word combination again with -a 1 dict2.txt dict1.txt
There is one additional thing to consider, some people want to try really, really hard to avoid duplicate words in there combination, like YellowBlueYellow or YellowYellowPurple. Depending on your hash type, hardware and size of the list it could make sense, but often the speed difference (filtering vs -a 1 speedup) makes the "few additional wrong combinations" negligible and therefore it could be faster to just NOT filter out duplicate words (it depends !).
To sum up, princeprocessor should/could be the easiest and most straightforward technique, but -a 1 (depending on hash type and hardware etc) might sometimes be the fastest one, with the disadvantages that you need to run at least 2 commands (switching from 2 combinations to 3 combinations) and maybe saving a temporary list to disk (which also could be huge depending on the number of original words, disk I/O is always bad in some way !)