Now when I played the game, I more or less typed this: "push rock", "hit lion with club", "strangle lion", "cut lion", "hit king" - and thus missed most of the point. The ideal play-through is (and you might want to try this out if you haven't yet done so) "cut lion", "drink".
So what is happening here? Nothing qualitatively different from what happens in many other interactive fiction games. When "eat apple" automatically leads to the implicit action "take apple", the parser is filling in the details for us in order to make the play experience more smooth. In the same way, the parser fills in the details for us here--but of course, these are a lot of details, and they might be considered more important and interesting than taking the apple.
Most games only automate actions that are (a) obviously necessary for doing whatever it is the player typed, and (b) the kind of oft-repeated actions that players have type again and again in all the games they play. Taking stuff, opening doors, using keys to unlock things - you've done it countless times before, and you're not missing out on anything if you don't get to do them now.
However, as Emily also points out, there are other uses of automated actions. One could implement a difficult machine that the player can fiddle with if so inclined, or just activate with a single high-level command if not. You can either type "play don giovanni", or "take don giovanni / turn on stereo set / press open / put don giovanni in tray / press close / press play / press next track / g / g / g / g / g / sing". When would this be useful? You might be tempted to say that it's always useful, because it allows players who enjoy fiddling with stuff to fiddle, and players who do not enjoy fiddling to get on with the game.
But anyone who knows the first things about interactive fiction understands that things don't work that way. The player who enjoys fiddling is almost guaranteed to type "play don giovanni" and never to find out that he could also have fiddled. The same would happen to the other player, mutatis mutandis. As Aaron Reed once said about Blue Lacuna: every player appears to get the story he least enjoys. (This is not an exact quote.) The problem here is giving the right signals: interactive fiction has a powerful convention that if you can take actions on a certain level of abstraction, then that is the level of abstraction on which you have to do things.
This convention is incredibly useful, because the main problem of interactive fiction is exactly this: to make the player understand the space of possible commands and their effects. The player will have agency in the game exactly to the extent that he can guess which inputs will work (and very few of the incredibly many possible inputs will) and what they will do. In order to give the player this agency, interactive fiction games typically have a set of standard commands with associated standard behaviours. But a game could get away with other commands and behaviours, as long as it introduced the player to them and then kept applying them consistently. Among other things, this would normally mean introducing a single level of abstraction and sticking to it. There is nothing confusing about a game that answers to "cut lion" by describing a whole scene that ends with cutting the lion; but it would be very hard indeed to keep confusion out of a game that treats "cut lion" and "hit lion" as belonging to widely different levels of abstraction.
So I assume that creating multiple levels of abstraction is generally not a good idea, because you will end up confusing the player and thereby taking away his agency, his whole idea of being in charge of the protagonist.
It might be claimed that Nemean Lion is not really about levels of abstraction (cutting a lion is not more abstract than pushing a rock), but about chains of necessary conditions, strings of cause and effect. But the same point applies: unless you understand the chains in advance, confusion must result.
Okay, so I have been quite negative about multiple levels of actions, but Emily is positive about it. She writes:
this kind of multi-level implementation can produce the sense of an experienced protagonist moving easily through a world that is nonetheless deep enough to allow experimentation.I am not convinced. If your game allows experimentation, but experimentation is not rewarding, you would have been better of not implementing it. If, on the other hand, experimentation is rewarding, why would you write your game in such a way that it discourages experimentation, or at the very least does not encourage it? It sounds like you are trying to keep two different types of gamers happy at the same time, while it is generally a better idea to just choose one type of experience you wish to create and write all of your game with that type in mind. Otherwise, you fall into the trap that so many pen & paper RPGs have fallen into - trying to please all GNS-types at the same time, and ending up by being an incoherent mess.
This is not to say that it couldn't work; but you would have to do some very clever signalling. Don't read my post as a warning. Read it as a challenge.
Of course, you could also accept the confusion that results from multiple levels and use it to your advantage - there is nothing a shrewd artist cannot make use of, as Cadre has illustrated for us with Nemean Lion.
P. S. Why is it Heracles? I can understand Hercules, from the Latin, and Herakles, from the Greek, but where does this hybrid come from?