The purpose of these capacitors is to provide a relatively low impedance above a designated frequency. The impedance at this designated breakpoint frequency is key to its function. But only the designer knows for sure, so unless you can figure out the criteria for selection, it will remain a mystery.
Starting with Cap "C" we see from my nomograph below that 0.05uF = 10 ohms the value in series with "C" occurs @ ~40kHz so we know it is not to provide an audio filter but rather to provide an AC coupled dummy load of 10 Ohms oscillations when the speaker impedance rises sharply due to inductance. This technique is common to most power amps because there is some instability from lack of phase margin at the higher frequency at unity gain. This dummy load improves that phase margin enough to prevent oscillations which can thermally damage the power output stage.
Cap "B" Must be large enough to pass bass frequencies into speakers. It appears to have a high pass breakpoint around 50 Hz into 8 Ohms. It could be made larger at the expense of size or perhaps more likely it was intended for 60 Ohm headphones then the cutoff is a lower frequency ~ 10 Hz.
Cap "A" bypasses and internal emitter resistor of 1.25k to boost gain which appears to be above 10Hz which is the impedance of 10uF.
There may be slightly different values which give better performance if you know what to look for.
--- end edit
You should learn how to use impedance to analyze Capacitors for ESR and breakpoint for different types ( e-cap, ceramic, low ESR types.) then apply this to a circuit where graphs are given for PSSR, (250uF) Gain ( 50uF) and spurious resonance f loading since speakers are inductive and RC of 10 ohms and 0.05uF provides 10 Ohm load at some breakpoint where Amp tends to resonate with no load. Hence all these types use this "snubber"
See if you can learn how to analyze graphs and evaluate the function of a capacitor in each type. ( eg. 250uF where ESR*C ranges from 1 to 250 us from quality) thus 250us/250uF= 1 Ohm ( poor) & 1us/250uF = 40 milliohms (good) and compare with 4 Om speaker that is 2 Ohms at DC for distortion 1 Ohm (Cap)/2 Ohms (DCR) is 50% then with bass maybe 25% and PSSR reduces 25% by only ~40dB or 1% of 50% is 0.5% distortion from PS noise from heavy bass current on supply ripple. So a mid range cap quality may be adequate for ESR along with 0.1uF.
It may take some concentration at first, then later, quick answers from practice to any question about RLC values using impedance ratios. Z1/(Z1+Z2) attenuation or simply looking up Breakpoint for HPF or LPF or BPF.