Proposing a zero-shot approach for linguistic steganography based on in-context learning to achieve better imperceptibility.