How to Fix Rhetorical Setups and Socratic Posturing in Text Using the Stop‑Slop Skill

The stop‑slop repository is a lightweight, Markdown‑based skill that packages editorial rules for eliminating AI‑generated writing clichés—specifically rhetorical setups and Socratic‑style prompts—through a simple file structure designed for integration with Claude and other LLMs.

Fixing rhetorical setups and Socratic posturing in text requires more than instinct; it requires systematic pattern recognition. The hardikpandya/stop-slop repository provides exactly that: a self‑contained, extension‑based rule set that treats editorial standards as structured data rather than vague guidance. By separating core principles from specific pattern libraries, it enables large language models to filter out formulaic structures in real time.

Understanding the Stop‑Slop Architecture

The repository contains no executable code. Instead, it functions as a self‑documenting data bundle where rules are stored as Markdown files and consumed by LLMs as behavioral guidelines.

Core Editorial Principles in SKILL.md

The file SKILL.md serves as the central manifest. It enumerates eight high‑level principles—such as cutting filler, breaking formulaic structures, and enforcing active voice—and points to granular reference files for implementation details. This design ensures that the core philosophy remains stable while the specific patterns it targets can evolve.

Pattern Libraries in the References Directory

Concrete enforcement lives in the references/ folder:

• references/structures.md – Enumerates structural clichés to avoid, including the classic Socratic setup ("What if …?") and rhetorical questioning patterns. Each entry includes a short "instead" rewrite demonstrating the transformation.
• references/phrases.md – Maintains the list of banned filler phrases and adverbs (e.g., "Here’s what I mean") that signal AI‑generated prose.
• references/examples.md – Provides concrete before‑and‑after edits showing the skill applied to real paragraphs.

This separation of concerns allows maintainers to add new patterns to structures.md or phrases.md without modifying the main skill definition in SKILL.md.

How to Fix Rhetorical Setups and Socratic Posturing

The skill targets specific mechanical patterns that trigger reader fatigue. When fixing rhetorical setups and Socratic posturing in text, the system references two specific prohibition categories.

Identifying Structural Violations

According to references/structures.md, the primary targets include formulaic openers like "What if we could…?" and other Socratic questioning techniques that create artificial distance between the writer and the reader. The file maps each pattern to a direct rewrite, converting hypothetical questions into declarative statements.

Eliminating Filler Phrases

The references/phrases.md file maintains a running list of signifier phrases that pad AI output. By cross‑referencing generated text against this list, the skill ensures removals like "Here’s what I mean" or "It’s important to note that" are cut before delivery.

Integration Methods for Claude and LLM Workflows

The skill supports three consumption models, all leveraging the raw GitHub URLs for immediate access.

Method 1: Upload as a Claude Skill

Package the entire repository folder and upload it as a skill in Claude Projects. The model will ingest SKILL.md and automatically reference the linked pattern files.

# system prompt snippet

name: stop-slop
description: |
  Remove AI writing patterns from prose. See the detailed rules at:
  https://github.com/hardikpandya/stop-slop/blob/main/SKILL.md

Method 2: Embed Directly in System Prompts

For one‑off implementations, paste the contents of SKILL.md directly into the system prompt, or reference the specific pattern files via URL:

You are an editor. Apply the Stop‑Slop guidelines (see https://github.com/hardikpandya/stop-slop/blob/main/references/structures.md) to the following paragraph and rewrite it without rhetorical setups or Socratic questions:

> What if we could make the process faster? Here's what I mean: …

Method 3: Inline Rule Lookup

For quick spot checks during a conversation, reference the principle number and file path:

# Quick reference (from SKILL.md)

2️⃣ Break formulaic structures – see https://github.com/hardikpandya/stop-slop/blob/main/references/structures.md

Summary

• The stop‑slop skill fixes rhetorical setups and Socratic posturing by treating editorial rules as structured Markdown data rather than soft guidelines.
• SKILL.md contains the eight core principles, while references/structures.md and references/phrases.md hold the specific patterns to eliminate.
• The repository supports three integration paths: Claude Skill upload, direct system prompt embedding, or URL‑based referencing.
• Because the architecture separates core rules from detailed patterns, extending the ban list requires only editing the relevant file in references/ without touching the main skill definition.

Frequently Asked Questions

What specific rhetorical setups does stop-slop target?

The skill primarily targets Socratic‑style openers like "What if…?" and structural patterns that rely on hypothetical questioning to introduce concepts. According to references/structures.md, these are flagged for conversion into direct statements to eliminate the artificial "classroom" tone common in AI-generated text.

How do I add stop-slop to an existing Claude project?

Download or clone the repository, then upload the folder as a custom skill in your Claude Project settings. Alternatively, copy the contents of SKILL.md into your system prompt and ensure the model can access the linked reference files via their GitHub URLs.

Can I customize the rules to ban specific phrases from my organization?

Yes. The modular structure allows you to fork the repository and edit references/phrases.md or references/structures.md to include industry‑specific jargon or internal clichés. Because SKILL.md points to these files generically, your custom patterns will be enforced without changing the core skill logic.

Does stop-slop work with LLMs other than Claude?

Absolutely. Any LLM that can process Markdown and follow URL references can utilize the skill. Simply provide the file contents or links to SKILL.md and the references/ directory in your system prompt, and the model will apply the editorial rules during generation.

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