De-AI-ify Text Refinement

What it fixes

AI-generated technical writing is often structurally clean but too smooth, overly certain, and over-packaged. This skill polishes that “AI flavor” without rewriting the argument or adding new claims. It focuses on wording, syntax, tone, and rhetoric so the text reads more like a real engineer or author.

How it works

It runs detection, rewriting, and verification. Detection targets patterns such as absolute statements, unsupported numbers, flowery metaphors, formulaic summaries, repeated three-item parallelism, excessive emoji, dash-plus-snark, and “not X, but Y” framing. Rewriting stays at the surface: headings, lists, bold text, tables, and information density are preserved. For example, “300% efficiency gains” may become a more cautious phrasing, and abstract “cognitive upgrade logic” becomes a concrete explanation. Verification checks meaning preservation, length inflation, structural flattening, and whether the output becomes a new template.

Boundaries

It is useful for cleaning machine-like prose in existing articles, blog posts, docs, or release notes. It is not for drafting from scratch, adding facts, inventing case studies, or expanding short notes into long-form content. It preserves register, keeps technical writing technical, avoids converting fullwidth and halfwidth punctuation, and does not introduce unsupported conclusions. If the source relies on fabricated data or anonymous anecdotes, it tends to remove them rather than keep them.