well, blast. that limit on automanips is a major problem, although workarounds are still possible.
((and i knew Nyars would refuse, it just was the logical follow up so...))
plan 1) the hard way
look up various forms of immaterial containment and/or shielding fields developed. particularly look for methods of containing either heat or plasma effects that do not require surrounding the effect but instead curve their shielding around it, probably by having an array of devices clustered together, a few forming the shield itself, initially aimed mildly off-center, and the rest forming modifying fields that act similarly to prisms upon the shield effect, causing it to curve as it passes the target area, forming a hollow sphere.
attempt to either duplicate or develop such a device, activate it, then have a dedicated automanip cause there to be a specific heat level or plasma ball inside the shielded sphere.
create a program to adjust the angles, shield strength, and prism strengths, to automate the process of moving the sphere around so that all it needs is directional input.
test the maximum range the device can handle the sphere being from the emitters before the containment shields are too diffused to be functional, and make the program not allow moving the sphere that far from the emitters. also disallow suicidal fireball placement, such as on the user or on the emitters themselves. teach it that the initial fireball position must be a fixed direction and distance from the device array, to allow the automanip to function. treat this position as the default position for the fireball to go throughout design.
check size, price, and power constraints, and attempt to remove excess and specialize the device array to make more portable and cheaper to produce, reducing fireball size as needed as well, and updating the program as needed to compensate for this process.
do not stop downsizing until it either stops working or is both realistically man-portable and cheap enough in materials i think it would cost about 5 tokens or less. check how much this has reduced performance, and note exactly when it became man-portable, and when i think it would only cost about 5 tokens, if either point is reached during this downsizing and efficiency work.
((i'm assuming here that the field itself will keep the fire/plasma ball going through simple conservation of energy, since it doesn't have anywhere to go to dissipate itself. also since the automanip is set to cause there to be a specific amount of heat/plasma in the area, instead of adding a specific amount of heat/plasma to the area, it should be able to simply be triggered again to bring the fireball back to full strength))
refinements:
if the device takes significant battle damage, make it refuse to activate the automanip anymore for any reason, set the shields on a middling-strong setting, and turn off the prisms, reshaping the shields into an outward facing cone, and expelling the fireball away from the user. after the fireball has had three times the time normally needed to dissipate,the rest of the device turns off and stays that way until repaired.
attempting to trigger the automanip causes the shields to return to the default position, then triggers it. the automanip will not trigger if the shields are out of position.
give the device a physical switch on it, which is covered and latched to prevent accidental use, which will disable the safeties with the exception of the auto-shut-off listed above. it defaults to have the safeties on, is a toggle, and if damaged (without the rest of the device being damaged) locks itself with the safeties on.
make the device compatible with prosthetic limb inputs. (remember it only receives directional movement input, and the automanip trigger request, there is no other input to this device)
plan 2) stupidly expensive
first, test each of the various amps, manips, and automanips. find which ones can make the fireball, and which of those have the longest battery life when set to keep it going as long as possible.
next, go through the amps, manips, and automanips again, take the longest lasting result from the above test, and see which amps, manips, and automanips can move it around. for now ignore practicality (if necessary use 6 automanips (one for each dimension) to move it for example).
next, attempt to find a way to reduce the price tag of the system by specializing it to this one purpose, removing unnecessary versatility options, unneeded inputs, etc.