Skip to content

StayInno/authentic-style-writer

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

11 Commits
 
 
 
 

Repository files navigation

authentic-style-writer

A Claude Code skill that analyzes text for AI-like writing patterns, scores them, and rewrites the text in an authentic human voice.

What it does

  • Detects 14 types of AI-typical signals (hedge clusters, assistant openers, abstract filler, forced symmetry, redundancy, pronoun-deficit, syntactic-repetition, etc.)
  • Scores AI-likeness on a 0–10 scale
  • Rewrites toward authentic, specific, human-sounding prose
  • Optionally learns your personal style from writing samples
  • Optionally flags possible mixed authorship within a single draft (Stage 1.7 Boundary Detection, for drafts ≥ 400 words)
  • Evaluates rewrite quality across four dimensions (genericity reduction, meaning preservation, style match, stylometry match) with a detection-resistance reference anchored to the StyloAI / AuTexTification baseline

Scientific grounding

The skill is built on published stylometric research:

  • Sadasivan et al. 2023 — paraphrasing attacks drop detector accuracy from 97% to 57–80%, validating the entire humanization approach.
  • Mindner & Schaaff 2024 — detector F1 falls from > 96% on raw ChatGPT to 78% on paraphrased text.
  • Alshammari 2025 — humanization brings GPTZero, Copyleaks, QuillBot down to 52–71% on DeepSeek-generated text.
  • Opara 2024 (StyloAI) — F1 = 0.81 benchmark on AuTexTification, used as detection-resistance anchor in Stage 4.
  • Petryshak & Rybchak 2025 — Simpson's D and TTR are top-2 features for Human/ChatGPT/DeepSeek separation, used as Stage 3 rewrite targets.
  • Mitrović, Andreoletti & Ayoub 2023 — SHAP-identified signals (redundancy, pronoun-deficit, syntactic-repetition) added to the taxonomy.
  • Kumarage et al. 2023 — change-point detection inspired the Stage 1.7 Boundary Detection.

For an external validator with ~120 stylometric vectors and Ukrainian-language support, see StyloMetrix (Okulska, Stetsenko et al. 2023).

Install

Option 1 — clone the repo into your project:

git clone https://github.com/StayInno/authentic-style-writer.git
cp -r authentic-style-writer/.claude/skills/authentic-style-writer .claude/skills/

Option 2 — copy the skill folder directly:

Copy .claude/skills/authentic-style-writer/ into your project's .claude/skills/ directory. Claude Code picks it up on the next session start.

Usage

/authentic-style-writer [your draft text]

With your own writing samples (for style-matched rewriting):

/authentic-style-writer
===MY WRITING===
[3–10 samples of your own writing, 600+ words total]
===DRAFT===
[text to rewrite]

See .claude/skills/authentic-style-writer/README.md for the full usage guide.

Files

.claude/skills/authentic-style-writer/
├── SKILL.md          # Pipeline instructions for Claude
├── ai-signals.md     # Signal taxonomy with examples
└── README.md         # Usage guide

About

AI skill - writing system that learns the user's real style from prior texts, diagnoses generic assistant-like signals, rewrites toward the user's authentic style, and evaluates style match separately from meaning preservation.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors