If you just want wave *form*, and can do without amplitude, just hold a normal 10x probe somewhere in the vicinity. It'll AC-couple in just fine. Or put a single turn sense winding on the transformer.
Making a wideband probe is tricky. If you use a chain of resistors, the capacitances from each resistor to free space (and to each other, and to whatever kind of shielding you have nearby) screws with compensation in a complicated manner. You can get a step response that's more-or-less correct, but has odd risey-dippy behavior in the 1-100kHz range. It's only give or take a dB or so, but it really sucks for a probe.
You can attempt to dominate parasitic capacitances with a bulk capacitance (for this kind of ratio, really just having a high voltage wire near the divider node -- a gimmick capacitor), but you can quickly run into unfavorable ratios: say, needing 10nF to ground to compensate it -- and no way to adjust it, because you can't just buy 10nF trimmer caps...cheaply, anyway.
The pros do it by keeping things small -- under oil if possible, like the sealed Tek (6026??!?) probe, which uses a small resistor and a guard ring / shield / capacitance hat to help with compensation.
I've done it before for lower voltages (~2kV) by stacking chip resistors and capacitors, say 200V each. That's about as much as you'd want to risk with them. Still needs a big compensation cap, and you can't vary AC gain because of it. You'd want a better approach for 10kV.
Tim