[G6Tanuki]

That should work OK, but I'm not sure if you need the two 100K resistors and the capacitor to the overwind on the mod transformer; I would simply feed the signal from the mod transformer to one end of a potentiometer [100K or so] with the slider to the diode; then you can use the pot to set the point at which the drive to the diode gets high enough for it to conduct and so start the back-biasing of the EF92 to reduce the gain.


Just for reference, attached is the LDR-and-neon circuit I mentioned [from Pat Hawker's Amateur Radio Techniques book]. I guess you could use a very small incandescent bulb driven from the overwind on the mod transformer in place of the neon; that might work better because of the non-linear voltage/light-output curve of a filament bulb, and its thermal inertia, would stop it having too much effect at low signal levels.