No AI Slop
Peter Yang's writing Skill identifies and edits more than 20 named AI-slop patterns while requiring minimal changes and preservation of the writer's vocabulary, cadence, humor, and imperfections.
Project overview
No AI Slop is a writing editor Skill published by Peter Yang. Instead of claiming to identify AI authorship, it breaks visible problems into named patterns such as binary contrasts, throat-clearing openers, faux-insight setups, colon reveals, importance puffery, synonym cycling, dramatic fragments, and fake-profound endings. Edit mode makes the minimum effective changes while preserving voice. Detect mode quotes and names findings without rewriting, scoring, or guessing authorship.
Core capabilities
Minimum effective edit
The Skill identifies the draft's point and voice signals, then changes only formulaic, incorrect, repetitive, or unclear material while preserving distinctive wording, cadence, humor, uncertainty, and structure.
Named-pattern audit
Detect mode names each rule, quotes the affected line, and suggests a short fix without rewriting, scoring, or claiming AI authorship detection.
Writing fundamentals
Beyond the named patterns, the rules check active voice, concrete details, direct verbs, sentence portability, and empty abstraction.
Built-in evaluation
After an edit, eval.md checks that no facts were added, the writer's voice remains, the edit is proportionate, and the listed patterns are resolved before output.
Installation and usage
Use the README's npx command to install only the `no-ai-slop` Skill globally in a compatible host, or send the repository's natural-language install prompt to ChatGPT, Claude Code, Codex, or another coding agent.
Let an AI Agent install it
Send this prompt to Codex, Claude Code, or another AI agent that can work with your local environment.
Help me install the No AI Slop writing editor Skill from https://github.com/petergyang/no-ai-slop. Read README.md, skills/no-ai-slop/SKILL.md, skills/no-ai-slop/eval.md, PRIVACY.md, and LICENSE first, confirm the current version and host compatibility, then run `npx skills add petergyang/no-ai-slop --skill no-ai-slop --global --yes`. Ask before replacing an existing Skill or plugin. After installation, confirm that the host discovers `/no-ai-slop`, then use a nonsensitive sample to test edit and detect modes. Verify that edit mode returns the full draft plus What changed and detect mode only names and quotes patterns. Ask before credentials, additional permissions, overwriting files, or any risky action.- ChatGPT, Claude Code, Codex, or another coding agent that supports Agent Skills
- An environment that can run the documented npx command
- A nonsensitive sample draft for verification
- A backup and confirmation before replacing an existing Skill or plugin
npx skills add petergyang/no-ai-slop --skill no-ai-slop --global --yes- 1Install the single Skill
Run `npx skills add petergyang/no-ai-slop --skill no-ai-slop --global --yes` in a compatible host environment. Ask before replacing an existing directory with the same name.
- 2Verify edit mode
Run `/no-ai-slop (your writing)`. Check that the output preserves meaning and recognizable voice, makes proportionate changes, and includes the full draft plus a short What changed section.
- 3Verify detect mode
Run `/no-ai-slop is this slop? (your writing)`. The response should name patterns, quote affected lines, and suggest short fixes without rewriting, scoring, or guessing authorship.
- 4Review before publishing
Check every fact, quote, voice choice, and intentional rhetorical device. If a rule harms the writer's style, give the agent sentence-specific feedback instead of accepting the whole rewrite.
Confirm that the host discovers `/no-ai-slop`. Run edit mode on a short draft containing `Here's the thing`, `It's not X. It's Y.`, and a fake-profound ending; the result should include the full edited draft and `What changed`. Then run detect mode and confirm it names and quotes patterns without rewriting or guessing AI authorship.
- The project does not require an account or operate its own server, but text is still processed by the ChatGPT, Codex, or other host where the Skill runs
- Do not use unpublished, confidential, or personally sensitive writing as a casual test sample
- Detect mode reports reviewable writing patterns and does not determine AI authorship
- The repository's only static image is a brand logo, so this page intentionally has no body media
- The current Windows build check has a path-separator issue while the latest Ubuntu CI validation succeeds
Use cases
Remove formulaic phrasing from posts and email
Cut throat-clearing, faux insight, binary contrast, and fake-profound endings while preserving spoken cadence and opinion.
Audit without rewriting
Ask for named patterns and quoted evidence, then let the author decide which lines to change.
Edit product updates and long-form writing
Protect numbers, mechanisms, and specific facts while removing importance puffery, synonym cycling, and abstract filler.
Review team writing
Use named patterns as discussable editorial evidence instead of an opaque score claiming that AI wrote the text.
Assessment
Based on README.md, SKILL.md, eval.md, the privacy policy, and the terms, No AI Slop's useful distinction is that it turns editorial judgment into named, quotable patterns instead of claiming to detect AI. Voice preservation and minimum effective change are part of both the workflow and the self-check. It remains an opinionated style guide: banned words, fragment rules, and rhythm choices can reject deliberate authorial decisions, so the writer must review the final draft. This assessment uses public materials and a build-script check; the Skill was not installed or tested on real writing.
Why it may be useful
- Clear separation between editing and detection without opaque AI-authorship scoring
- Explicit protection for vocabulary, cadence, humor, uncertainty, digressions, and concrete facts
- eval.md checks voice preservation, proportional editing, and factual integrity
- No project account or external server, with source available under the MIT license
What to know first
- Named patterns reflect an editorial viewpoint rather than an objective rule across every language and genre
- Banned words and formatting rules can conflict with brand voice, advertising, poetic writing, or intentional fragments
- Text privacy remains governed by the ChatGPT, Codex, or other host processing the draft
- The Windows package check has a path-separator issue, and actual installation or editing quality was not tested here
README
No AI Slop
Overview
Peter Yang's writing Skill identifies and edits more than 20 named AI-slop patterns while requiring minimal changes and preservation of the writer's vocabulary, cadence, humor, and imperfections. Removes 20+ patterns of AI slop from any piece of writing.
Getting started
- Use the README's npx command to install only the `no-ai-slop` Skill globally in a compatible host, or send the repository's natural-language install prompt to ChatGPT, Claude Code, Codex, or another coding agent.
- The README positions the project as removing more than 20 AI-slop patterns without erasing the writer's vocabulary, cadence, humor, or imperfections.
- Installation options include a natural-language prompt for ChatGPT, Claude Code, Codex, or another coding agent and an exact global npx command.
- Confirm that the host discovers `/no-ai-slop`. Run edit mode on a short draft containing `Here's the thing`, `It's not X. It's Y.`, and a fake-profound ending; the result should include the full edited draft and `What changed`. Then run detect mode and confirm it names and quotes patterns without rewriting or guessing AI authorship.
Configuration
npx skills add petergyang/no-ai-slop --skill no-ai-slop --global --yes