A velomobile would probably be an ideal candidate for a bit of regenerative braking.
I'd be quite happy to see regenerative braking, and a motor. As long as the battery starts at 0 volts, the criteria for being powered solely by human effort is satisfied. A redesigned hub dynamo would work well. It would work especially well for timid descenders.
Suitable gearless hub motors
[1] are readily available, and the controllers that drive them can do regenerative braking simply by altering the phase of the switching (it is, in effect, a simple software modification), so most of them support it. The real-world benefit is negligible
[2] (you typically extend the battery range by a couple of percent), as on a bicycle there's so little kinetic energy to recover, and a limit to how much current you can dump into a small battery.
A velomobile might do better, as it's slightly heavier and loses less power to drag, but most multi-track recumbent people are leaning towards mid-drive motors for boring practical reasons, which are lighter, but upstream of the freewheel and can't regenerate.
Starting from 0V isn't practical with a lithium-ion battery (though you could of course integrate the current in and out to determine the net consumption), but might be with supercapacitors (which handily tolerate higher peak currents, too). That'd be a pretty niche application - basically for competing in events with a "no stored energy" rule that are enlightened enough not to dismiss electrikery out of hand. (I believe the HPV racing community - which would permit this sort of thing - have experimented with flywheels in the past, to conclude they aren't worth the weight.)
[1] Which are basically a bigger, three-phase version of a hub dynamo.
[2] Other than as an anti-lock braking system for mountain bikes.