Can LLMs Write Faithfully? An Agent-Based Evaluation of LLM-generated Islamic Content