GitHub – hardikpandya/stop-slop: A skill file for removing AI tells from prose

If you’ve spent any time reading AI generated content, you start to notice it. The rhythm. The tidy structure. The slightly over-polished tone that feels… just a bit off.

I’ve caught myself doing it too. Writing something, rereading it, and thinking, “This sounds like a robot trying very hard to be helpful.”

That’s exactly the problem Stop Slop by Hardik Pandya is trying to solve.

**Stop Slop** is a skill file designed to remove what it calls *AI tells* from prose. In simple terms, it teaches language models to recognize their own bad habits and cut them out. Think throat-clearing openers. Overused emphasis. Business jargon. Vague declarations that say a lot without saying much. Even structural clichés like dramatic fragmentation, fake contrasts, and that distant narrator voice we’ve all seen.

It goes deeper than just banning a few phrases. The project outlines:

• **Banned phrases and adverbs**
• **Structural clichés to avoid**
• **Sentence-level rules** like no passive voice, no lazy extremes, no Wh- sentence starters
• A scoring system, rate your writing across dimensions and revise if you score below 35 out of 50

What I appreciate is how practical it is. You can plug it into Claude Code, upload the SKILL.md into projects, or even drop the core rules into custom instructions. It’s modular. Flexible. Not locked to one tool.

There’s also something bigger happening here. As AI writing becomes more common, readers are getting sharper. We can feel when something lacks texture. Real human writing has rough edges. Small contradictions. Slight detours. Personality.

Tools like Stop Slop aren’t about hiding AI. They’re about raising the bar.

And honestly, that’s a good direction. Because whether we’re using AI or not, the goal stays the same. Write clearly. Sound human. Mean what you say.

If you’re experimenting with AI assisted writing, this repository is worth exploring. Not as a shortcut, but as a mirror.