The next step was one that the papers from the Gold Order had seemed to have significant trouble describing. Magic flowing from one's soul to an effect in the world is the norm for a maintained spell, but transitioning it from that temporary stream to a more permanent bond is a difficult technique, not least of which because one has to rotate it around one's soul while doing so, as otherwise it will be located in the vertex of their soul that they habitually use for spellcasting. Retraining yourself to cast from a different edge of your soul is a long and tedious process. You also take care to make sure that its final resting place will be as far as possible from where Wolf's soul adjoins your own.
So there is a limit on the number of soul bonds a Wizard can maintain. And a stricter limit on the number they can have and still use magic for other things.
…I would bet money on the practical limit being 6 'slots', although it is potentially possible to fit multiple similar links in the same slot.
the structural integrity of your soul
It would be scientifically valuable to test whether that integrity is increased or decreased by arcane marks.
It would also be extremely dangerous and rather difficult to test that on yourself and ethically problematic to test on someone else.
that obedience will wear grooves in its being until it becomes its new nature.
Given what we know of local metaphysics this likely means that after sufficient time there may be naturally formed versions of summoned apparitions.
Directed is an utter waste as it means we are swapping one combatant for an interior combatant.
Instinctual just screams friendly fire risk.
Dual doesn't seem to have any significant use cases that Charge or Ambush couldn't also fill.
Ambush plays to the strength of the Rider in Red but isn't really what we want it for. Killing lone wizards is something Mathilde can do very well herself. Also doesn't fit the Dämmerlichtreiter theme particularly well.
Charge gives us a new tool as a fire and forget anti-unit weapon.
Bodyguard would be very useful as it makes Mathilde less likely to be overrun. However having it only able to attack when and where Mathilde does seems limiting.
All considerations must bow to AP hell. Particularly as we will want to codify the spell which costs its own action. One is all we can afford. Also sub-battle magic would be really nice.