What problem does it solve? AI-generated text carries recognizable patterns—staged openers, forced triads, inflated significance, chatbot residue—that make prose sound generic and machine-written. This Skill detects 25 documented AI writing tells and rewrites the text so it reads like a person wrote it, while preserving every fact, name, number, and claim from the source. ## Core Features & Use Cases - Pattern-based rewriting: Detects and fixes 25 AI writing patterns grouped in five sections (staging, rhythm, inflation, formatting, chat leftovers), based on Wikipedia's "Signs of AI writing". - Fact preservation: Never invents names, numbers, dates, quotes, or citations; asks the user when a needed detail is missing instead of fabricating one. - Voice matching: Accepts a writing sample and matches its sentence length, word choice, punctuation, and quirks, including dash usage. - Three output modes: Pasted text returns draft, critique, and final version; file mode edits prose in place while leaving code, frontmatter, and links untouched; embedded mode returns only final text. - Use Case: Paste an AI-drafted blog post or point the Skill at a Markdown file, and receive a rewrite that keeps every supported claim but removes the tells that mark it as machine-generated. ## Quick Start Ask the agent to humanize your pasted text or a named file, optionally including a short sample of your own writing for voice matching.
