Hardik Pandya · @hardikpandya ↗ Stop Slop
Strip the phrases and rhythms that mark writing as AI-generated so drafts read like a person wrote them. It rewrites against concrete banned-pattern rules and scores the result for human voice.
The roster
What it does
Stop Slop teaches your agent to catch and remove the patterns that mark writing as AI-generated. It rewrites prose against concrete rules: banned phrases and throat-clearing openers, structural cliches like binary contrasts and dramatic fragmentation, and sentence-level bans on em dashes and staccato fragments.
Then it scores the result from 1 to 10 on five dimensions of human voice: directness, rhythm, trust, authenticity, and density. Anything below the threshold gets another pass. Add it as a skill in Claude Code or any LLM that reads the SKILL.md format.
When to use it
- Your writing has that AI smell
It rewrites against concrete banned-pattern rules so drafts read like a person wrote them.
- You want a human-voice score before publishing
It grades the result so you can tell whether the slop is actually gone.
Common questions
How is Stop Slop different from just asking the model to sound human?
Stop Slop rewrites against concrete banned-pattern rules covering throat-clearing openers, em dashes, and business jargon, then scores the result out of 10 for human voice rather than relying on a vague instruction.
How do I use Stop Slop?
Stop Slop is a single Markdown skill. You add the folder as a skill in Claude Code, upload SKILL.md to a Claude Project, or include it in your API system prompt so the reference files load on demand.
Is Stop Slop free?
Yes. Stop Slop is open-source under the MIT license with no paid tier, and it runs inside whatever model or agent you already use.