You don't seem to understand what I mean.
If the system is designed to be cut in the event that it would be unaligned with its on-ground receptors, through opto-mechanical interrupters, then you cannot use it for an aggression, unless you move the on-ground receptors all the way to your target, but that would be very flagrant and difficult to hide, and every worker at the receptor facility would see it, as well as any other civilian far or near the project.
The only way that such satellite could go wrong is if it actually does not integrate such failsafe mechanisms, and if its real design is hidden, and ill-known to anybody except a handful of military-secret-bound engineers.
If I were to follow your reasoning, then by the same rule any energy plant or even any technological artifact would be incredibly unsafe. But the fact is that we integrate security mechanisms in those, and are mostly satisfied with them, such that we accept to use these technologies.
Of course accidents happen, but the security mechanisms evolve too. Some are so good that they rarely fail anymore, like for example the master cylinder in car brakes (it is an actuator that enhances your effort on the brakes, but if it fails then you end up just pushing raw on the brakes and thus it works anyway, just a little harder).
Of course, this kind of Dyson sphere project will require quite a bit of popular education to convince and explain why its mechanisms make it safe.
This reminds me of the company who developed RNA vaccines because they don't risk to change cells' DNA and they are safer globally, and then they have to explain endlessly to people that no, it's not an evil design, and, no, it does not even include nano-5G-antennas....