What problem does it solve?
AI-assisted drafts often sound generic or unlike the researcher's own writing, and tools that only remove AI tells leave prose merely less bad. This Skill builds a written profile of how you actually write from your own published papers, so new drafts can be measured against a concrete target rather than a list of taboos.
Core Features & Use Cases
- Corpus-based profiling: Processes three to twelve of your own papers one document at a time via subagents, producing per-document notes on lexicon, rhythm, openings, transitions, hedging, and quirks.
- Reusable voice profile: Synthesizes the notes into a
voice-profile.md file capturing signature vocabulary, avoid lists, sentence rhythm, structural habits, hedging register, and deliberate quirks.
- Draft auditing: Run with
--audit draft.tex to compare an existing draft against the profile, reporting per-section distance with evidence tied to specific profile lines.
- Use Case: Before drafting a new paper's introduction, build a profile from your last six published papers so the draft matches your actual register, and let
/humanize respect your documented deliberate quirks instead of flagging them as AI tells.
Quick Start
Ask the AI to build a voice profile from the folder containing your published papers, then audit your current draft against it.
