Remove Ai Style

This does exactly what the name says: scans your text for patterns that scream "AI wrote this" and rewrites accordingly. It comes with three intensity levels (moderate, heavy, full), defaulting to heavy, which means it will actively restructure formulaic paragraphs and swap out robotic transitions. The Chinese version even includes quick-scan scripts to flag quote misuse before you do a full read-through. Works for both English and Chinese. Honestly, the "full" mode is the interesting one because it treats your draft as raw material and rebuilds it from scratch while keeping the meaning intact. If you're shipping AI-assisted content and need it to pass the smell test, this handles the tedious rewrite work.

skills.sh installs: 4.2K
last updated: Aug 21, 2026

View on GitHub

Install

$ npx -y skills add zc277584121/marketing-skills --skill remove-ai-style --agent claude-code

Installs into .claude/skills of the current project.

Files

Remove AI Style

Use a two-layer workflow:

Run the deterministic analyzer to locate repeatable lexical, punctuation, structural, and assistant-residue signals.
Read the full article and make contextual judgments that rules cannot cover.

Apply one strong default. Fix confirmed findings at every severity, rewrite awkward passages, and remove unflagged formulaic rhythm while preserving facts, meaning, required structure, and the author's intended voice. Do not ask the user to choose an intensity and do not offer light, moderate, heavy, or full modes.

Most analyzer hits still require contextual judgment. Three rule families are hard publication constraints in editable prose:

Remove all exclamation marks.
Remove all em dashes, en dashes, and double hyphens used as dashes.
Rewrite all formulaic binary contrasts, including “不是……而是……”, “并非……更是……”, “看似……其实……”, “not just X but Y”, and close variants, as direct claims.

SKILL.md on GitHub