Humanizer

Based on Wikipedia's AI writing patterns guide, this skill scans your text for telltale signs of LLM output and rewrites them to sound human. It catches the usual suspects like "pivotal moment," "vibrant community," and those fake-depth participial phrases ("reflecting the community's connection to..."). But the interesting part is the second half: it doesn't just remove bad patterns, it actively adds voice. The examples show how to inject opinion, vary sentence rhythm, and acknowledge uncertainty instead of neutral reporting. Worth running on anything that came out of ChatGPT before you ship it, especially if you're editing Wikipedia articles or trying to avoid that telltale algorithmic smoothness.

Install to Claude Code

npx -y skills add shipshitdev/library --skill humanizer --agent claude-code