The ability to learn and practice new skills applies to the computation of abstract properties such as geometric form and neuroprosthetic control, as well as physical actions. We trained adult mice to successfully use an optical brain-computer interface (BCI) device so that we could assess the extent to which abstract skill learning disrupts previously acquired cortical function. We found that animals used multiple neural strategies to gain volitional control of the BCI device, and continued to do so after additional training. Despite significant engagement of local neural activity during BCI performance, tuning properties and stimulus encoding were not perturbed by new skill acquisition. These results indicate that flexible selection of neural strategies during goal-directed practice may facilitate the integration of new skills with existing function.