I'm struggling with the proof of the following, well-known result for the Levi-Civita connection of a Lie Group with biinvariant metric, i.e. satisfying \begin{equation} g(D_bL_a X_b, D_b L_a Y_b)= g(X_a, Y_a) \end{equation} with $L_a$ the left (or right) translation by $a$.
The claim is \begin{equation} \nabla_X Y= \frac{1}{2}[X,Y] \end{equation} which one should (supposedly) be able to prove through Koszul's formula. My first question, though, is more basic: why is this a connection at all? I should have $\nabla_{\phi X}Y=\phi\nabla_XY$ and $\nabla_X\phi Y = \phi\nabla_X Y + X(\phi)Y$: but the commutator is antisymmetric in $X$, $Y$: how can they follow different rules? Having $X(\phi)=0$ for every function and left-invariant vector field would seem to be the only way out, but I can't find any reason for this to hold. As for Koszul, I just keep getting something different - but I have the feeling there's something I really failed to understand. What are the terms one can simplify in the formula? One has $g(X,Y)=const.$ from the biinvariance - is there more?