What Is the Purpose of the no-ai-slop Repository?

The no-ai-slop repository implements a Codex skill and ChatGPT plugin that detects and removes "AI-slop"—over-used, generic, or artificially polished patterns common in AI-generated text—while preserving the author's personal voice.

The petergyang/no-ai-slop repository addresses the homogenization of digital writing by providing automated tools to strip clichéd AI phrasing from drafts. This open-source project serves as both a technical implementation and a stylistic guardrail, helping writers maintain authenticity in an era of generative text. Understanding the purpose of the no-ai-slop repository reveals how specialized editing rules can enforce voice preservation while eliminating robotic patterns.

Core Functionality: Edit and Detect Modes

The repository provides two primary operational modes designed to handle AI-slop remediation at different stages of the editing workflow.

Edit Mode

In Edit mode, the skill accepts user drafts and applies minimal surgical edits to remove more than 20 predefined AI-slop patterns. These patterns include binary contrasts ("It's not X. It's Y."), throat-clearing openers ("Here's the thing,"), faux-insight setups, and colon reveals. The system returns both the cleaned draft and a "What changed" summary documenting every modification.

Detect Mode

Detect mode operates as a diagnostic scanner that identifies slop patterns without rewriting the text. The skill lists each detected pattern with a quoted line reference and provides targeted fix suggestions, allowing authors to manually approve changes while learning to recognize their own AI-adjacent habits.

Repository Architecture and Key Files

The implementation relies on four critical components that define rules, validate output, and package the distribution for the Codex ecosystem.

SKILL.md: The Rule Engine

Located at skills/no-ai-slop/SKILL.md, this file contains the human-readable specification that governs the editing workflow. It enumerates the specific linguistic patterns to excise and establishes the editorial philosophy of minimal intervention. According to the source code, this document serves as the authoritative instruction set that the Codex agent follows when processing text.

eval.md: Quality Assurance

The skills/no-ai-slop/eval.md file implements post-edit validation checks that verify the edited text preserves the original point, voice, and structural integrity. These safeguards ensure that slop removal never compromises the author's intended meaning or unique stylistic markers.

plugin.json: Integration Metadata

Found at .codex-plugin/plugin.json, this manifest provides the plugin metadata required by Codex and ChatGPT integration points. It specifies the skill name, version, description, capabilities, and default prompts necessary for the agent to invoke the functionality correctly within conversational interfaces.

build_plugin.py: Distribution Builder

The scripts/build_plugin.py utility validates the manifest integrity, copies skill files to the appropriate structure, and generates a ZIP archive ready for global installation. This automation ensures consistent packaging across releases.

Installation and Usage Examples

Users can install the skill globally in a coding agent using the command:

npx skills add petergyang/no-ai-slop --skill no-ai-slop --global --yes

Example: Editing a Draft

Input command:

/no-ai-slop
It's not X. It's Y. Here's the thing, I think we should move forward.

Result:

[Edited draft]
It's Y. I think we should move forward.

What changed
- Removed binary contrast "It's not X. It's Y." → kept only the meaningful clause.
- Cut throat-clearing opener "Here's the thing,".

Example: Detecting Slop Patterns

Input:

/no-ai-slop is this slop? 
The future isn't coming. It's already here.

Result:

Found patterns:
- Binary contrast: "The future isn't coming. It's already here." → keep only "The future is already here."

Summary

• The no-ai-slop repository provides a Codex skill and ChatGPT plugin specifically engineered to identify and eliminate AI-generated linguistic patterns while maintaining authorial voice.
• It operates through two distinct modes: Edit (automated cleaning with change summaries) and Detect (diagnostic scanning with manual fix suggestions).
• The architecture relies on skills/no-ai-slop/SKILL.md for rule definitions, skills/no-ai-slop/eval.md for post-edit validation, .codex-plugin/plugin.json for platform integration, and scripts/build_plugin.py for automated packaging.
• The tool targets over 20 specific slop patterns including binary contrasts, throat-clearing phrases, and faux-insight constructions.
• Installation supports global agent integration via npx skills add, making the utility accessible across development environments.

Frequently Asked Questions

What exactly qualifies as "AI-slop"?

AI-slop refers to over-used, generic, or artificially polished linguistic patterns that frequently appear in AI-generated text. According to the petergyang/no-ai-slop source code, these include binary contrasts (e.g., "It's not X. It's Y."), throat-clearing openers (e.g., "Here's the thing,"), colon reveals, and faux-insight setups. These patterns create a homogenized writing style that lacks personal voice.

How does no-ai-slop ensure it doesn't change the author's intended meaning?

The repository implements a validation layer defined in skills/no-ai-slop/eval.md that performs post-edit checks. These checks verify that the edit respects the original point, voice, and structure of the draft. Additionally, the Edit mode follows a philosophy of minimal intervention, only removing specific predefined patterns rather than performing broad rewrites.

Can I use no-ai-slop without ChatGPT or Codex?

While the repository is specifically architected as a Codex skill and ChatGPT plugin through the .codex-plugin/plugin.json manifest, the underlying rules in skills/no-ai-slop/SKILL.md and validation logic in skills/no-ai-slop/eval.md are human-readable specifications. Developers could theoretically implement these rules in other editing pipelines, though the primary distribution mechanism requires either the Codex agent environment or ChatGPT plugin interface.

What specific patterns does the tool detect and remove?

The skill targets more than 20 predefined patterns, including binary rhetorical contrasts, unnecessary throat-clearing openers, faux-insight setups, and colon reveals. The complete enumeration of these patterns resides in skills/no-ai-slop/SKILL.md, which serves as the authoritative rule set for the editing engine.

{
  "mcpServers": {
    "instagit": {
      "command": "npx",
      "args": ["-y", "instagit@latest"]
    }
  }
}