
Stop Slop
github.com/hardikpandya/stop-slop- Category
- AI Agents
- Rank
- No. 420Tools index
- Listed in
- #3 Make agent writing sound human
- Pricing
- Open Source
- Platform
- cli
- Type
- TOOL
- Builder
- hardikpandya
- GitHub
- 16.9k stars
- Date
About
A skill file that teaches AI models like Claude to identify and remove predictable AI writing patterns, banned phrases, and structural clichés from prose. Helps create more natural, human-sounding text by eliminating common AI tells.
What it does
Stop Slop supplies an editorial rulebook for revising prose. It flags filler, vague claims, canned contrasts, dramatic fragments, passive constructions, false agency, and repetitive rhythm. Writers can apply its checklist during drafting or editing, then assess directness, rhythm, trust, authenticity, and density.
Why it's ranked here
Its value comes from unusually specific guidance rather than a broad request to sound human. The phrase tables, structural diagnoses, replacement advice, and transformations make the rules usable. Its rigid prohibitions also limit the range of prose it can handle without human judgment.
What's good
The guidance explains why patterns fail and often supplies a better approach. It asks writers to name actors, replace vague importance claims with concrete consequences, vary sentence lengths, and remove rhetorical scaffolding. Before-and-after samples make several edits easy to understand and repeat.
Tradeoffs
Several rules overreach. Banning every adverb, every passive construction, every sentence beginning with a question word, and every three-item list can remove useful language. One revised example even contains an em dash despite the explicit prohibition. The scoring rubric also depends on subjective judgments without calibration guidance.
How to use it well
Use it as a second-pass checklist for essays, product writing, and other prose that has grown formulaic. Treat its bans as prompts for inspection, then keep exceptions that serve meaning or voice. It does not provide grammar checking, factual verification, citation support, or automated editing software.
Technical notes+
The repository is a Markdown instruction package rather than executable software. SKILL.md contains the core rules, quick checks, and evaluation rubric. references/phrases.md catalogs wording to remove, references/structures.md describes recurring constructions and alternatives, and references/examples.md supplies transformations. README.md documents use through Claude Code skills, Claude Projects knowledge, custom instructions, and API system prompts. LICENSE grants MIT terms. The provided repository tree shows no tests, package manifest, CLI implementation, library code, or service endpoint.
Observed
- License
- MIT
- Primary format
- Markdown instruction and reference documents
- Install surface
- Folder-based skill, project knowledge upload, copied custom instructions, or system-prompt content
- Interfaces
- Prompt and skill content; no CLI, library, MCP server, or standalone API implementation shown
- Model support
- Described for Claude or any LLM that can follow supplied instructions
- Repository structure
- Core skill document plus phrase, structure, and example references
- Testing
- No test directory or test files appear in the provided repository tree
Read from README.md, LICENSE, SKILL.md, CHANGELOG.md, references/phrases.md, references/examples.md, references/structures.md.
What it can do
Identify predictable AI writing patterns in text
AI-generated prose text → List of detected AI writing patterns and issues
Remove banned phrases from AI writing
Text containing common AI phrases → Text with banned phrases eliminated
Eliminate structural clichés from prose
Text with predictable AI structural patterns → Text with varied, natural structure
Score text for AI detectability
Written text sample → Scoring rubric results indicating AI tells
Transform AI-generated text to sound more human
Robotic or formulaic AI writing → Natural, human-sounding prose
Train AI models to avoid common writing tells
AI model and skill file → AI model with improved natural writing capabilities
Tags
Comments (0)
No comments yet
Editorially curated, with community endorsements as a secondary signal. Corrections welcome.