In recent years, the volume of the commercial mobile robot market has been continuously growing due to the maturity of both hardware and software technology. To build commercial robots, skid-steering mechanical design is of increased popularity due to its manufacturing simplicity and unique hardware properties. However, this causes challenges on software and algorithm design, especially for localization (i.e., determining the robot's rotation and position). While the general localization algorithms have been extensively studied in research communities, there are still fundamental problems that need to be resolved for localizing skid-steering robots. To tackle this problem, we propose a probabilistic sliding-window estimator dedicated to skid-steering robots, using measurements from a monocular camera, the wheel encoders, and optionally an inertial measurement unit (IMU). Specifically, we explicitly model the kinematics of skid-steering robots by both track instantaneous centers of rotation (ICRs) and correction factors, which are capable of compensating for the complexity of track-to-terrain interaction, the imperfectness of mechanical design, terrain conditions and smoothness, and so on. These time and location varying kinematic parameters are estimated online along with other localization states in a tightly-coupled manner. More importantly, we conduct in-depth observability analysis for different sensors and design configurations in this paper, which provides us with theoretical tools in making the correct choice when building real commercial robots. In our experiments, we validate the proposed method by both simulation tests and real-world experiments, which demonstrate that our method outperforms competing methods by wide margins.