If you've ever read something that technically says all the right things but still screams AI, this is the fix. Built from Wikipedia's AI cleanup guide, it catches the telltale patterns like "nestled in," em dash abuse, vague "experts say" attributions, and those -ing phrase tack-ons that add fake profundity. But the real gem here is the personality section, which doesn't just strip out slop but pushes you to add actual voice back in. Varies sentence rhythm, acknowledges uncertainty, lets you have opinions. Originally by @blader, now at v2.1.1. Works across Read, Write, Edit, so you can either fix existing text or have Claude draft with this awareness baked in.
npx skills add https://github.com/davila7/claude-code-templates --skill humanizer