No, that argument makes no sense. SQL injection doesn't happen because of where the input comes from; it happens because of how the input is handled. We can avoid Bobby Tables scenarios while receiving input that influences SQL queries from humans, never mind neural networks. We do it by controlling the system that transforms the input into a query (e.g. by using properly parameterized queries).
No, you don't. You have a command generated by auditable, conventional code (in the agent wrapper) rather than by a neural network.