ASSUMING that (a) the underlying population is normally distributed, (b) sampling is random with replacement, and (c) the test statistic you are using is the (standardized) sample mean, you use a two-sided test.
The rejection region is - z_sub_alpha/2 < (sample mean - 200) / (sigma / sqrt(sample size)) < z_sub_alpha/2.
You use a z-test because sigma is known, The level of significance alpha is given to be 0.04. For a standard normal distribution you want z_sub_ 0.02. The two critical values are then, plus or minus z_sub_ 0.02.
To find z_sub_ 0.02, go to OpenOffice Calc and use the function NORMSINV(). (If you do not know, how ask me as a comment.) You get z_sub_ 0.02 = -2.0537.
Check Answer (b) above.