Due to its low hardware cost and ease of deployment, WiFi fingerprinting has become a prominent research direction in indoor positioning. However, traditional methods based on Received Signal Strength (RSS) still face three critical challenges: susceptibility to noise interference, low retrieval efficiency as fingerprint databases scale up, and trajectory instability in dynamic environments. These challenges are inherently rooted in the stochastic fluctuation of RSS signals, the high-dimensional and non-Euclidean nature of fingerprint space, and the unpredictability of user movement patterns. To address these limitations, an adaptive trajectory-assisted dynamic indoor positioning algorithm based on RSS fingerprinting, termed AT-WKNN (Adaptive Trajectory-assisted Weighted K-Nearest Neighbor), is proposed. Specifically, a hybrid distance metric incorporating adaptive distance constraints is first designed to identify high-quality neighboring fingerprints while filtering out noisy samples, thereby improving positioning accuracy. Subsequently, a Hierarchical Navigable Small-World (HNSW) structure is employed to enable efficient fingerprint retrieval. In addition, a Kalman filter is utilized to smooth trajectory estimation and suppress dynamic noise. Experimental results demonstrate that the proposed AT-WKNN algorithm achieves a 43.0% improvement in positioning accuracy and a 6.90 $\times$ increase in retrieval efficiency compared with the baseline WKNN method. Furthermore, validation on the large-scale UJIIndoorLoc benchmark dataset confirms the scalability and generalization capability of the proposed method.