Reviewing generated code means carefully examining the code produced by a model before using it: understanding it, checking that it meets the request, handling edge cases, looking for errors and flaws, and running it with tests. Generated code is never reliable by default.
Example
Check that the suggested function handles an empty list and only calls libraries that exist.