We present the first quantitative spectroscopic modeling of an early-time supernova that interacts with its progenitor wind. Using the radiative transfer code CMFGEN, we investigate the recently-reported 15.5 h post-explosion spectrum of the type IIb SN 2013cu. For the first time, we are able to directly measure the chemical abundances of a SN progenitor and find a relatively H-rich wind, with H and He abundances (by mass) of X=0.46 +- 0.2 and Y=0.52 +- 0.2, respectively. The wind is enhanced in N and depleted in C relative to solar values (mass fractions of 8.2e-3 and 1e-5). We obtain that a dense wind/circumstellar medium, with a mass-loss rate of Mdot= 3e-3 Msun/yr and wind velocity vwind=100 km/s, surrounds the star at the pre-SN stage. These values are lower than previous analytical estimates, although we find Mdot/vinf consistent with previous work. We also compute a CMFGEN model to constrain the progenitor spectral type and find that the high Mdot and low vwind imply that the star had an effective temperature of ~8000 K immediately before the SN explosion. Our models suggest that the progenitor was either an unstable luminous blue variable or a yellow hypergiant undergoing an eruptive phase, and rule out a WR star. We classify the post-explosion spectra at 15.5 h as XWN5(h) and advocate for the use of the prefix `X (eXplosion) to avoid confusion between post-explosion, non-stellar spectra with those of massive stars. We show that the progenitor spectral type is significantly different than the early post-explosion spectral type owing to the huge differences in the ionization structure before and after the SN event. We find the following temporal evolution: LBV/YHG -> XWN5(h) -> SN IIb. Future early-time spectroscopy in the UV will give access to additional spectroscopic diagnostics and further constrain the properties of SN precursors, such as their metallicities.