One of the best-documented transitions in vertebrate evolution is from a short-bodied, robustly-limbed, pentadactyl form to an elongate, limbless, snake-like form. Such transition is gradual and in some cases, extant intermediate species exist for millions of years. How can these intermediate species do well? One of the possible aspects is how these animals coordinate limb and body movements, combining running via limb propulsion and swimming via body undulation. In this paper, we studied the locomotion pattern of three species of Brachymeles skinks (B. kadwa with hind leg lengths (HLL) = 0.17±0.02 snout-vent length (SVL), B. taylori with HLL = 0.15±0.02 SVL, and B. mungtingkamay with the HLL = 0.09±0.01 SVL) and compared it with the stereotypical lizard, Uma scoparia (HLL = 0.25±0.03 SVL), and the legless lizard, Lerista praepedita. We use new theoretical tools including geometric mechanics and neural net trackers to explain and analyze how they can make forward progress on sand surface via appropriate coordination of aforementioned locomotion modes. Our numerical model shows that both leg movements and body undulation contribute to generating self-propulsion in the intermediate species. Moreover, the body-leg coordination observed in these intermediate species quantitatively agrees with theoretical predictions from our modeling to optimize forward speed. Although evolutionary pressures can shrink limbs and elongate bodies, motor control systems can flexibly accommodate these changes to generate appropriate limb-body coordination for high locomotion performance.