To really do this subject justice would take many, many pages, but for
now we will settle for a short overview.

Lima does not use add_action() at all; it uses verbs instead.
Absolutely nothing remotely similar to add_action() exists in the lib.
It is completely impossible for a room or object to add new commands
to the game.

The reason for this is consistency.  With the Lima lib, things work
the same everywhere, making things much easier to understand.  Of
course, many things may not do anything interesting; however they
should at least give a reasonable error message.  For example, if
there is anything in the MUD that can be twisted, it makes more sense
for EVERYTHING to be able to be twisted, and simply do nothing,
instead of each object which can be twisted having to completely
reimplementing the concept.

The add_action() way of doing things leads to lots of code
duplication, and in many cases poor parsing since the person writing
the command is more interested in getting it to work for him than
doing any sort of general parsing; in many cases the person in
question is an area coder with little mudlib experience anyway.  As
anyone who has worked extensively with add_action() knows, bringing
any two such objects into close proximity often results in complete
disasters, and rarely even succeeds in generating the correct error
message for most commands.

-Beek

P.S. if you really need it somewhere, there is always:

void create() {
    parse_init();
    parse_add_rule("foo", "STR");
}

mixed can_foo_str() {
    // or something appropriate to objects meant to be carried
    return environment(this_body()) == this_object();
}

Intelligent readers will note this proves that the LIMA parser
is capable of doing anything add_action() was capable of doing.
The reverse, however, is not true, nor is it anywhere close to being true.