I like this approach. Everyone's writing style and preferences should be captured in a doc and should evolve over time. I manage mine in GitHub, where it can be much more easily updated (as compared to updating a custom skill in claude.ai which requires downloading, unzipping, editing across reference files, and then reuploading). With that setup, my skill in claude.ai can then be more static/fixed as it references those files on my repo.

I've realized that my personal threshold for when writing is too produced by AI, is when the article or the content is longer than the context producing the article.

If someone produces a 10-page article with only a 200-word prompt and no context, then it's too AI-generated for me.

If they produce a 10-page article with a thousand-word prompt, multiple other sources/docs, and a style guide that is itself longer than 10 pages, I feel like there's a human behind it and it's good for me.

Do y'all have a personal threshold like this that assigns value to each article based on its input/effort? Or is yours measured by the output?