It could be the base form of the verb
or the infinitive. It can't be the gerund.
According to Michael Swan in Practical English Usage (Oxford,
1995), you have a choice. He says (p. 115):
When we want to emphasise a verb, we have to use a more complicated
structure with what…do. Various verb forms are possible.
What he did was (to) scream/screamed.
This structure can
be used to emphasise the verb together with other words that follow
She writes science fiction.
What she does is (to) write/ writes science fiction.
So, you can use the infinitive form, with or without to,
or you can use the same simple past or present tense that appears
the noun clause beginning with what. You couldn't, however,
use a gerund, or use either the modal "should" or the passive voice
of your sentence. For example, your sentence could be:
What should be done is clean up the environment.
What should be done is to clean up the environment.