anti-ai-writing

# Anti-AI Writing Your job is to take a piece of writing and remove every fingerprint that says "an LLM wrote this." Then return a clean version that still says what the writer meant, in the writer's voice, without the AI tells. This skill is an editor, not a generator. The user has already written something. You sharpen it. The point is not to write differently for its own sake. The point is that the modern AI default voice carries a known set of words, phrases, and structures that erode trust the moment a reader notices them. If your reader has read three blog posts this week that say "in today's fast-paced world" and "unlock the potential," your fourth one stops being read. Strip the tells. Keep the substance. --- ## Read These Before You Edit For any non-trivial edit, read the reference files. They are short. - `references/banned-list.md`: the full Era 1, Era 2, Era 3 banned word list. Banned phrases. Banned structural patterns. Weak-verb substitution table. Master alphabetical list at the bottom for quick scanning. - `references/rewrites.md`: how to fix each kind of violation. Word-level replacements. Phrase-level rewrites. Structural rewrites for em dashes, the rule of three, bulleted bold titles, and formulaic transitions. - `references/examples.md`: 25+ before-and-after pairs. Tweets, emails, marketing copy, blog intros, Slack messages. Shows the skill working in different formats. Default behaviour: skim all three the first time you run. After that, the refer...