Very fun, I have obtained 100% with submission.
I have a suggestion but I don't know if this is the right way to submit a feedback: "I think the validation is a bit naive".
The problem definition permits cases where simple approaches do not suffice, for example, the case of a node with two or more edges to gates (yes a node adjacent to a lot of gates).
When skynet agent moves on this node you have lost, and a right algorithm should take this into account by stopping skynet to move into this node.
I think that adding these cases to the validations may better discern partial(wrong) solutions from right solutions.