The rule applies to the normal distribution.
For any normal distribution, 68% of the observations lie within 1 standard deviation (SD) either side of the mean that is, between (mean - SD) and (mean + SD); 95% of the observations lie between (mean - 2*SD) and (mean + 2*SD); 99.7% between (mean - 3*SD) and (mean + 3*SD).
To find the probability that a normally distributed random variable, with mean = m and SD = s, you calculate its z-score = (X - m)/s and look up the relevant probability value in tables.
In doing this, you may have to use the symmetry of the normal distribution or the probability of complementary events (or both). This will depend on what exactly is tabulated.
Mean 70, SD = 15
Pr[X < 55] = Pr[Z < (55 - 70)/15] = Pr[Z < -1] = 0.5*(1-0.68) = 0.5*0.32 = 0.16
Pr[X < 40] = Pr[Z < (40 - 70)/15] = Pr[Z < -2] = 0.5*(1-0.95) = 0.5*0.05 = 0.025
Pr[X < 85] = Pr[Z < (85 - 70)/15] = Pr[Z < 1] = 1 - 0.16 = 0.84
Pr[X < 100] = Pr[Z < (100 - 70)/15] = Pr[Z < 2] = 1 - 0.025 = 0.975
Pr[X > 85] = 1 - Pr[X < 85] = 1 - 0.84 = 0.16