I agree with Steve. This is a really dangerous project. If you need to ask for help, you probably are not experienced enough to do it safely.
If you chose to do it with the level of knowledge you are displaying, and without doing extensive testing. You are putting yourself at extreme risk of death.
100kv will give you your 3 inch sparks. I play with 50kv quite a bit, and it can arc over an inch.
250kv will give you over a foot of arcing, and treat most things as conductors, including the surface of dielectrics if there is an opposing charge on the other side. Like your body (pretty much everything is an opposing charge to 250kv) So as Steve pointed out, the arc would likely travel up the glove and around to your body, likely causing instant death. Or go right through a microscopic pinhole that was undetectable. A well grounded Faraday's suit would be a must for such an experiment!
I would suggest making a 9v powered
555 stun gun, and working up from there. After a few months of playing with HV I am still caught out from time to time, and get zapped. I have made sure my HV driver wont kill me. I wont be upgrading till I get used to what I have first. Things like insulated leads, can make great arc points with HV. Things in the area can store quite a charge via surface capacitance too. I have been zapped quite hard. from a 90L plastic storage crate, a few min after turning off my flyback driver. Turns out the crate can hold a few nano farads worth of charge on its surface... At 50kv thats not quite lethal, it will put me on the floor though! At 250kv it not going to be very forgiving.
Another point, The device you intend to make, could be used as a weapon. To harm or even kill people. Have you considered the legal ramifications, should you, or someone else harm another person either accidentally or intentionally? As well as how you would feel if your party trick turned lethal?
If you must proceed. Watch this
video to see how to make a test rig. There is no way you want to trust an online calculator with your life. I'd be doing a few hundred repeat tests on anything I was actually going to use too. Then take worst case results and add some buffer room.