Bay 12 Games Forum

The arbitrary distinction of animals into normal, war and hunting variants predates the personality system and in my opinion it should be scrapped and replaced with simple, exaggerated personality traits and skills (already implemented) for animals. In the child stage, an animal trainer could impart "aggression", "stealth"' "labor" or tricks" in a child animal, with each training facet increasing relevant skills and personality traits (e.g. training for aggression increases Fighter, Discipline, Biter if the animal has an appropriate body part, etc.).  When the animal becomes an adult, its personality and skills are fixed and potentially some labors enabled (e.g. an animal highly trained in tricks will dance if its owner is also doing a performance task (or sing if it's a bird?), and an animal trained in labor will haul items for its owner (but not do more complex tasks like crafting). Higher animal trainer skill increases traits faster or decreases cooldown time between when  an animal can be trained. Animal species and caste could affect how much skills are trained by each session, so e.g. a monkey could train tricks very fast but protectiveness very slowly. Training aptitude could be handled by evaluating existing tokens and implementing a handful of new ones.

Well, it is a shame that idea didn't work, IndigoSnake. Got to say, this idea is relevant to my interests as well.

Personalities are mostly already present and moddable.

Syndromes can grant a creature the ability to speak, it's just that the code that determines if a player can open the talk menu doesn't account for it - you'll notice AI creatures affected by the syndrome can talk just fine . As for dogs that can learn combat skills, you could always try giving them [CAN_LEARN] or [SLOW_LEARNER] in their creature raws (I believe the game checks their caste to see if the can learn skills, rather than accounting for syndromes?), and then setting the SKILL_LEARN_RATES for non-combat skills to 0. You'd gain the additional problems that intelligence grants (like needing to eat, sleep, have personalities(?), not be butcherable), and it'd be up to the game whether it actually respects those rates when generating those creatures.

You could use the syndrome to grant an on_self interaction that triggers the interaction you really want.

Uninformed idea, Regarding your cow: Have you tried using it's tile number that you can see with dfhack with gui/gm-editor df.global.world.texture.page? - that page lists all graphics tilesheets, and I've gotten lua viewscreens to display graphics by using numbers from it (for me, they were in the thousands).

Barding  for war-trained animals is the logical conclusion, here.

Spoiler (click to show/hide)

Why would you need to do that?  The attributes are all you need, you don't even need minimum requirements at that, just have some kind of computation that makes it increasingly unfeasible to train a given animal if it's relevant scores are not high enough.  At one point you will hit the lifespan of the animal and you have effectively set an "impossible to train" flag on it. 

This way, each animal type only needs its average and standard deviation defined for every attribute (how sentient species are done already) and the natural variance included there gives you a controlled randomization on how many of said species are trainable and to what extent, etc, on average.