What problem does it solve? AI-generated text carries recognizable tells: em dashes, rule-of-three lists, vague attributions, promotional language, and filler phrases. This Skill identifies and removes those patterns from drafts so the writing reads as natural human prose without losing any factual content. ## Core Features & Use Cases - Pattern Detection and Removal: Scans for 33 documented AI writing patterns drawn from Wikipedia's "Signs of AI writing" guide, covering content, language, style, and communication tells. - Fact-Preserving Rewrites: Keeps every claim, name, number, and citation from the source while restructuring sentences, and never invents facts to fill gaps. - Voice Calibration: Analyzes a user-provided writing sample and matches its habits, letting the author's style override the Skill's own rules. - Optional Scoring Pass: Scores the source and rewrite on dimensions like hedging, filler, and passive voice via references/prose-score-dimensions.md when the user requests it and jev is available. - Use Case: Paste a ChatGPT-drafted blog post and receive a rewrite with the AI tells removed, an audit of what changed, and a final version that keeps your arguments intact. ## Quick Start Paste your AI-drafted text and ask the assistant to humanize it while preserving all facts and matching your usual tone.
