Actually, a 3055 works pretty well as an experimental device. Mine produced about 30kv for short periods of time. No capacitor is required in when using a transistor switch, because the switch cuts off the current sufficiently fast. I've measured 400V spikes on the primary, but the transistor seems to tolerate fast spikes even as they are well over the breakdown voltage. But using a higher rated transistor would be wise.
But, the 555 does not well tolerate spikes on the VCC rail, so supperssion should be used between the chip and the coil.
PS: Any regulated power supply will fry if powering this curcuit. Ask me how I know this!