Independently assess whether a reproduction evidence package supports, contradicts or cannot resolve a specific paper claim. Use after execution and mechanical verification, preferably in a fresh context; do not use it to repair the experiment while judging it.
