What problem does it solve? AI-generated text carries recognizable tells — decorative vocabulary, template openings, uniform sentence rhythm, em-dash overuse — that make prose sound machine-written. This Skill detects those patterns, presents the findings for approval, then rewrites the text so it reads as if a human wrote it. ## Core Features & Use Cases - Pattern Analysis: Scans input for vocabulary spikes, opening/closing templates, structural tells, punctuation habits, and sentence-rhythm uniformity, then classifies severity as Light, Medium, or Heavy AI. - Approval-Gated Rewriting: Presents the full analysis for user approval before dispatching the expert-naturalizer agent to rewrite, preserving code blocks, tables, proper names, and domain-specific terms. - Optional Web Research: A research:yes flag pulls current AI-pattern data from sources like Wikipedia and Pangram Labs before analysis. - Use Case: You have a blog draft generated by an LLM that sounds robotic. Run the Skill on the file, review the detected patterns, approve the strategy, and optionally overwrite the original file with the rewritten version. ## Quick Start Humanize the text in @file:draft.md and show me the detected AI patterns before rewriting.
