Various thoughts:
The dice pool system encourages players to use a variety of different stats/approaches. A lot of games end up producing a "players persuade the GM to let them use their highest stat" effect, if there's an equivalent here it would be "players persuade the GM to let them use the stat with the most remaining dice." (This also makes group checks very powerful, since they let players benefit from very low stats by putting them together). The first dynamic comes back in combat, there's not much reason to use your lower stats for attacking or defending at all. Maybe if they had more tightly defined outcomes?
Relationship tokens are cool, although the "spend to recharge dice" function feels much stronger than the +1/-1 to a roll. I guess the flat modifier gets stronger if it can be applied after seeing the result and you can spend multiple of them... but even one die restored is a +1 at least to the roll it's used on, and possibly up to a +6
Conditions produce a death spiral effect (as in, they're a failure consequence that makes failure more likely), which isn't a bad thing but is good to be aware of. I'm a little more cautious about it because a failed roll is already a pretty significant resource expenditure, since you're spending dice on it (EDIT: Just realized this doesn't apply because you gain injuries in combat, where dice don't deplete). On the other hand, minor conditions feel like their effects are too minor, a -1 isn't going to affect high-dice rolls very much at all.
Is the intent for the Sore roll that the player get a chapter "on screen" as their cat recovers, or they just get knocked out of play for a scene while the others keep going?
Currently, the chapter system is a pretty straightforward presentation of the recurring "scene" concept that provides a few extra prompts and helps ground the type of play the game expects. I wonder if there's a way to tie them more closely to XP advancement? currently XP triggers are marked or cleared on a per-session basis, but tying them to chapters would help keep pacing more consistent and at the very least would make it easier to adapt to play-by-post.
Overall I'm feeling positive about this, I think the project has potential!