Spamassassin doesn't do anything besides check mail for spam content and give it a score accordingly. What you can then set it to do is write its results [and score] in the header so the user can't see it but the email client and something like Procmail can see. You can set a score that email has to attain to be considered spam and get a "X-Spam-Flag: YES" to be added into the header. This score can be low or high, depending on how confident you are in it and how points are awarded can be altered, so email that is heavy HTML might score more points then other criteria does; it's all up to you or you can just leave the defaults as is.
What many do is then set rules in Procmail to do certain things with email marked "X-Spam-Flag: YES", but Spamassassin itself doesn't deal with this end of it.
I think this concept wasn't explained in a few tutorials I read to set it up and I assumed [wrongfully?] that Spamassassin actualy did a whole lot more. Once I had the concept right, then dealing with Spamassassin was much much easier.