@User @User I meant to reply to your conversation the other day but got distracted. I have played around with each of the tools your mentioned.
1. Xyce python hooks work well and I would recommend.
2. ADMS is really not suitable for much beyond device modelling and has very poor to no support for the digital side of Verilog-AMS like cross and transition functions that really are needed for behavioural stimulus. There's been a fair bit of discussion about this on the Xyce mailing list for the interested reader.
3. Finally I couldn't get the isotel arrangement to work with one of my blocks. But I maybe need to spend more time with it because in theory it should work and I also really want it to work. The Xyce approach is more co-simulation rather than digital models in the analogue simulation like the isotel approach. For things like a PLL digital logic this is much preferred.
So I'm hoping you can make some headway. I also noticed (after my experiments) that there is some xspice models for sky130 in the PDK which could be very useful