stop-slop

Stop Slop is a reusable LLM instruction package intended to help identify and remove stylistic patterns commonly associated with AI-generated prose. Its main file is SKILL.md, supported by reference files listing phrases to remove, structural patterns to avoid, and before/after transformation examples. The README describes several ways to use it: add the folder as a Claude Code skill, upload SKILL.md and references to Claude Projects knowledge, copy the core rules into custom instructions, or include SKILL.md in an API system prompt. Reference files can be loaded on demand. The guidance targets throat-clearing openers, emphasis crutches, business jargon, adverbs, vague declaratives, meta-commentary, binary contrasts, negative listings, dramatic fragmentation, rhetorical setups, false agency, distant narration, and passive voice. Sentence-level rules include avoiding Wh- sentence starters and em dashes, requiring active voice, and discouraging staccato fragmentation and lazy extremes. It also includes a scoring rubric from 1 to 10 for directness, rhythm, trust, authenticity, and density. According to the README, a total below 35 out of 50 means the text should be revised. The repository is released under the MIT license.