The latest FEVER challenge of Automated Verification of Textual Claims (AVeriTeC) addresses the problem of chaotic contexts in real-world data using a deterministic approach of evidence question generation and using question-answer pairs for veracity prediction and claim stance classification. In this paper, we present a novel contextual engineering approach to effectively address these challenges. In this work, we propose a new prompt engineering method called Chain-Of-Thought Context Attention Promoting(CoT-CAP) and two novel “prompt-centric” approaches that address the challenges of automated fact verification in chaotic contexts. The first method, User Perspective Prompting(UPP), uses a self-attention prompting method to extract the underlying contexts directly from the claim and evidence sentences. The second method, Contextual Stance and Veracity Prediction(CSVP), uses a cross-attention prompting method on contextual knowledge obtained from UPP to evaluate the veracity and evidence-claim stance predictions. Our process achieved around a 13% increase in performance with the prediction of the veracity and the classification of the gold evidence extracted online compared to the baseline model of AVeriTeC.