I remember seeing a stem similar to *gʰrebʰ- somewhere in Proto-Semitic. As I've explained to Phoenix, I treat Nostratic works such as the sources of Allan Bomhard and John Kerns as helpful leads to potential IE-Semitic contact. Bare with me, since while I'm a kind of sympathist to the theory of Nostratic, I hardly believe that Bomhard and Kerns are completely on the ball with it and there are many series which they consider cognate but which I think are more likely more recent loans, particularly so when they list a supposed cognate series with only PIE and Proto-Semitic data to substantiate their reconstructions. Very suspicious indeed.
So, it's not surprising that I managed to find a lead in one of Bomhard and Kern's books (see Bomhard/Kerns, The Nostratic Macrofamily (1994), p.386) which compares PIE *gʰrebʰ- "to scratch" with a list of Semitic data which suggest a triliteral skeleton *grb. It just so happens that precisely this skeleton is explicitly listed in Murtonen, Hebrew in Its West Semitic Setting (1989), p.140 with accompanying attested forms in various Semitic languages and with semantics revolving around "itching".
This then leads me to wonder if the following hypothesis has meat:
Theoretically, the stemfinal *bʰ would devoice to *p in original *skerbʰ- once speakers of Indo-European no longer were consciously aware of the historical connection with *gʰrebʰ-, and this would especially occur after *s- came to be irregularly omitted and phonotactic "stop voicing harmony" pressures took over. Once the sibilant disappeared, it would be all too easy for even a native speaker to get confused between a historical phonetic [k] (an allophone of voiced *gʰ following a sibilant) and the homophonous phoneme *k. The confusion would be made more likely if this "optional" s-mobile as a fossilized foreign morpheme, as I continue to maintain, had never ever become a productive phoneme in Pre-IE or PIE as some PIEists still believe.
PSem *garābu "to scratch" ~ *šagāribu "made to scratch; itchy"
MIE *ɢaréba- "to scratch (tr.)" ~ *saɢérba- "to scratch (intr.)"
> early Late IE *ɢreb- ~ *sqerb-
> PIE *gʰrebʰ- ~ *(s)kerp-