If you are trying to prove a conditional, a good strategy is to assume the antecedent, prove the consequent, then discharge the assumption by the rule of conditional proof. So, you want to assume ¬A → ¬(A → B) then prove A. Since ¬A → ¬(A → B) itself is a conditional, you can then assume ¬A and prove ¬(A → B).
Which rules you are allowed to use depends on the text you are following, but the proof might look something like this:
1. ¬A → ¬(A → B) Assumption
2. ¬A Assumption
3. ¬(A → B) 1,2 MP
4. A ∧ ¬B 3, Impl
5. A 4, ∧-E
6. A ∧ ¬A 2,5 ∧-I
7. ⊥ 6, ⊥-I
8. ¬A → ⊥ 7,2 CP, discharging assumption 2
9. ¬¬A 8, RAC
10. A 9, DNE
11. (¬A → ¬(A → B)) → A 1,10 CP, discharging assumption 1