writing-humanize_skill

Overview

This skill removes signs of AI-generated writing and rewrites text to sound natural and human. It detects 24 documented AI writing patterns and applies targeted fixes so the result reads as specific, varied, and human-authored. Use it to clean docs, blog posts, READMEs, or commits without changing meaning.

How this skill works

The skill scans input for a catalog of 24 AI tells (inflated language, fake depth, chatbot artifacts, unnatural grammar, formatting tells, weak endings, and developer-specific patterns). It classifies content type and applies content-type rules, then performs a rewrite at the selected intensity (light, standard, heavy) while preserving facts. Finally, it returns the rewritten text and a concise summary of detected patterns and applied fixes.

When to use it

• Preparing a blog post or article for publication to ensure a natural voice
• Polishing README or project docs to remove promotional or formulaic language
• Cleaning technical docs or API reference to remove filler while keeping precision
• Shortening PR descriptions or commit messages to remove padding
• Batch-scanning project docs to find and fix mechanical formatting tells

Best practices

• Choose content type accurately so voice rules are applied correctly
• Start with a light pass for minor cleanup; use heavy only when structure needs rebuilding
• Preserve facts: do not add new technical details during rewriting
• Review the change summary to confirm no meaning drift occurred
• For files, run a scan first and pick specific files to avoid unwanted edits

Example use cases

• Paste a draft blog post to remove hedging, vary sentence rhythm, and inject personality
• Scan a repository's docs to remove emoji headings, curly quotes, and inflated README copy
• Rewrite API reference snippets to remove filler while keeping precise behavior notes
• Compress a verbose PR description into a clear, action-focused commit message
• Run a heavy rewrite on marketing copy to replace buzzword stacks with concrete claims