← ClaudeAtlas

prose-linterlisted

Check prose for the mechanical tells of AI writing, and fix them. Use this whenever generating or editing prose meant for people (documents, reports, markdown, README/docs, emails, proposals, deliverables) and whenever asked to "de-AI" text, make writing sound human or natural, remove AI tells, or review a draft for AI-generated patterns. Covers the patterns catalogued in Wikipedia's "Signs of AI writing": inflated-importance claims, name-dropping, vague sources, formulaic sayings, fake alternatives, passive voice, nominalisation, symmetry, bullet overuse, missing opinions, filler openers, semicolon and em-dash overuse, and more. Includes voice matching, false-positive guidance, a mechanical CLI scanner (scripts/scan.py) that can gate a build, and a-priori guidance for writing human from scratch.
tarvanitis/ai-prose-tools · ★ 0 · Code & Development · score 67
Install: claude install-skill tarvanitis/ai-prose-tools
# Humanize Prose Goal: text a reader cannot flag as AI-written. --- ## Step 0 — Voice matching (do this first) If you have a writing sample from the author, read it before touching anything else. Note: - Sentence length range (short punchy? long periodic?) - Whether they use contractions - How they start paragraphs (anecdote? claim? question?) - Punctuation habits (serial comma? em-dash? parentheticals?) - Register (formal? casual? blunt?) Apply those habits throughout the rewrite. A text can be free of every pattern below and still feel wrong if it doesn't sound like the author. --- ## The 36 patterns ### Category A — Content #### A1. Inflated-importance claims Asserting that something is revolutionary, unprecedented, or game-changing without evidence. **Before:** This framework represents a paradigm shift that will revolutionize how organizations think about data. **After:** This framework changes how organisations structure their data pipelines. Watch for: revolutionary, unprecedented, paradigm shift, game-changing, transformative, groundbreaking, landmark. --- #### A2. Name-dropping without substance Dropping famous names, companies, or concepts that add prestige but no insight. **Before:** As Aristotle understood, and as Google has demonstrated, the key to success is iteration. **After:** Rapid iteration shortens feedback loops and reduces the cost of mistakes. Watch for: name + "understood", "recognized", "has shown", "has demonstrated" when the name a