de-slop_skill

Overview

This skill removes LLM-isms and AI writing patterns to restore natural, human-sounding prose. It locates immediate ‘tells’, clustered overused phrasing, and deeper structural issues, then applies prioritized edits to preserve meaning and voice. Use it when text reads like a chatbot or needs a human rewrite.

How this skill works

I scan the text to diagnose three levels of issues: red (clear AI tells), yellow (statistical clusters of overused words or constructions), and green (structural patterns that need rewriting). I remove or replace red items first, then break up yellow clusters, and finally rewrite or reweight green patterns to restore natural rhythm and specificity. The final pass checks for overcorrection, meaning preservation, and voice consistency.

When to use it

• Editing prose that reads like chatbot output
• Reviewing or polishing AI-assisted drafts before publishing
• Fixing copy that feels generic, repetitive, or overly hedged
• Self-check when writing feels ‘off’ but the cause is unclear
• Preparing content for human audiences or editorial review

Best practices

• Run a full read before editing to preserve intent and context
• Fix red flags (chatbot phrases, grandiose filler) first — they’re unambiguous
• Break up word clusters rather than ban single words; prefer plain verbs and nouns
• For structural issues, rewrite sections instead of doing blind word swaps
• Do a final read to avoid overcorrection and to keep a consistent voice

Example use cases

• Humanize blog posts produced with an AI draft
• Clean up marketing copy that uses cliché AI phrases
• Polish documentation that overuses formal, balanced structure
• Edit academic or technical prose to remove hedging and restore direct claims
• Prepare social posts or newsletters to sound like a specific author