I've thought it this way. ##E^*## is the vector space of all linear functions ##φ : ℝ^2[t] → ℝ##. The canonical basis of ##E## is ##\{1,t,t^2\}##. So the goal is to write any ##φ∈E^*## as a linear combination of ##F_0,F_1,F_2##, i.e. solve the linear equation ##φ=x_0F_0+x_1F_1+x_2F_2##, i.e. ##φ = (F_i(V_j))_{j,i} \vec{x}##.
In coordinates it is ##φ(1) = φ(1,0,0), φ(t) = φ(0,1,0), φ(t^2) = φ(0,0,1),## and it's a clearer (to me) to write ##F_i(p)=F_i(a_0,a_1,a_2)## instead.
To decide whether the ##F_i## form a basis you will probably need another property given at the beginning. You can either solve the linear equation(s) or find an argument why the matrix is regular.
(For further talk remind me please what is meant by anti-dual basis of ##B^*##. Is it a basis of ##(E^*)^*##?)