IMO you should explain exactly what happens, because your assumption that the process is "lenition" could be wrong. I will assume for the sake of discussion that some stop becomes more fricative-like in this context. An explanation then resides in a couple of phonetic factors. First, in word initial position, you don't have cues from before the stop that tell you anything about the segment, whereas if there is a vowel before the stop, you have formant transitions that tell you that all of the sudden you have a stop. All of the evidence for there being a stop come after the release. Then, with the following vowel [a], the following target has a maximally open vocal tract (whereas a stop has a maximally closed vocal tract). Speakers will tend to anticipate that vocal tract position, leading to subtle variants of the stop that are "more open" vs. "more closed", which can lead to releases that have more-fricative like properties. Since the release properties of initial stops are so important in manner-identification, this can lead to new allophones like affricates, which can then simplify to fricatives. This is especially applicable to back consonants like velars and uvulars. The prediction is that labials would be least-susceptible to this kind of anticipatory coarticulation leading to re-analysis. (OTOH if this is about Osage, there's a different explanation).
FYI at the other end of the scale, there is also a similar effect of high front vowels, owing to the constriction formed by [i], and resulting release turbulence being reinterpreted.