If you want the amplifier to be small-signal stable with any passive load:
1. Determine the amplifier's output impedance without the L and R
2. Look for frequency regions where the real part of the output impedance goes negative. I'll assume there is a negative peak at frequency f_neg.
3. Choose an R that is greater than two times the opposite of the most negative real part of the output impedance, for example four times the opposite of the most negative real part of the output impedance
4. Connect an L in parallel with a reactance equal to R at f_neg
5. Determine the impedance of the amplifier with the L and R included and fine tune their values as needed to make the real part of the impedance positive at all frequencies
I took a more pragmatic approach when I designed my amplifier: I came up with a target for the magnitude of the output impedance at 20 kHz, divided that by 2 pi * 20 kHz to find the inductance, put a resistor in parallel with a value close to the nominal load impedance (8.2 ohm) and shunted the output with a capacitor with value C ~= L/R^2 to turn the whole thing into a first-order series filter. (Making it a first-order series filter was recommended in an article of A. N. Thiele, as it helps to keep RF signals out of the amplifier.) I have no idea if it is stable with any passive load, but it certainly is with all loudspeakers and cables I ever connected to it.