Where the boilerplate came from
Early diffusion models were trained on captions where words like "masterpiece" and "trending on artstation" correlated with higher-quality images. Including them genuinely helped.
Current models are trained differently and largely ignore them. The terms survive because prompts get copied, and they still occupy the early positions where weighting is strongest.
What to do with the space
Removing five noise terms from the front of a prompt moves your subject and your medium into the positions that matter. That is usually a bigger improvement than anything you could add.
The suggested prompt in the output is simply your terms with the noise removed, in the same order.