The pvalue has the usual interpretations.
If the pvalue is below a threshold like 0.05, then we reject the null hypothesis.
In this case a small pvalue indicates that the data is not consistent with a normal distribution.
Specifically, Jarque-Bera is a test for skew and kurtosis.
If the tests rejects, then either skewness or kurtosis or both differe from a normal distribution.
chi2 test are based on a squared test statistic, so the test is always for deviation in the direction of smaller and larger deviations from the null hypothesis.
(always 2-sided for the underlying deviations.)