The AntiSlop

A Claude Code skill that detects and fixes AI-generated writing patterns. More comprehensive than detection alone — it actually rewrites the problems.

Why This Exists

AI-generated text has tells. Wikipedia's Signs of AI Writing documents 24 patterns. But detection alone isn't enough. You need a tool that:

1. Detects comprehensively — 35+ patterns across content, language, style, and structure
2. Scores severity — Not all patterns are equally bad (tiered system)
3. Actually fixes problems — Editor mode rewrites, not just flags
4. Catches structural tells — Staccato fragments, manufactured personality, uniformity

Installation

Option 1: Skills CLI

npx skills add aplaceforallmystuff/the-antislop

Option 2: Clone to skills directory

mkdir -p ~/.claude/skills
git clone https://github.com/aplaceforallmystuff/the-antislop.git ~/.claude/skills/the-antislop

Usage

In Claude Code, Claude Desktop, or any Claude interface with skill support:

/antislop

[paste your text here]

Or ask Claude directly:

Run antislop on this draft: [your text]

The Horoscope Test

"Could anyone have written this, for anyone?"

If yes, it's slop. Like a horoscope — technically applicable to everyone, resonant with no one.

Detection Categories

Tier 1: Remove Immediately

• delve, game-changer, revolutionary, leverage, unlock potential
• "it's worth noting," "in today's digital landscape"
• moreover, furthermore, cutting-edge

Tier 2: Suspicious When Repeated

• here's the thing, at the end of the day, the bottom line
• paired adjectives ("comprehensive and thorough")
• template openings ("In this post, we'll cover...")

Tier 3: Watch for Clusters

• transition words (however, firstly, moving forward)
• corporate buzzwords (robust, seamless, scalable)

Structural Patterns (Often Missed)

• Staccato fragments: "Short. Punchy. Exhausting."
• Sentence uniformity: Every sentence 10-15 words
• Comparator sentences: "This isn't X. It's Y."
• Manufactured personality: Fake developer snark
• Self-promotional framing: Author's achievements as headline

Scoring System

• 0-5: Low risk
• 6-12: Medium risk (significant editing needed)
• 13+: High risk (likely unedited AI)

Editor Mode

Unlike detection-only tools, The AntiSlop fixes problems by default:

1. Removes Tier 1 phrases
2. Deduplicates Tier 2 phrases
3. Combines staccato fragments into flowing prose
4. Replaces comparator sentences with direct statements
5. Varies sentence lengths

Request "audit only" if you just want detection without edits.

Example

Before:

Let's delve into how AI is revolutionizing the landscape. It's worth noting that these game-changing tools are unlocking potential at scale. The speed is impressive. The quality is enhanced. The adoption is growing.

After:

Here's how teams are using AI coding tools. In a 2024 Google study, developers completed simple functions 55% faster, but showed no improvement on debugging. The tools handle boilerplate well — config files, test scaffolding, repetitive refactors — but can't tell when they're wrong.

Further Reading

For more on AI writing patterns and maintaining your authentic voice, see the AI Writing Field Guide.

Related Skills

Part of the aplaceforallmystuff skills collection:

• voice-editor — 6-pass editing to match your voice profile
• voice-analyzer — Create the VOICE.md that personalizes detection
• slop-detector — Lightweight detection-only variant

Credits

• Pattern research from Wikipedia: Signs of AI Writing
• WikiProject AI Cleanup
• Finnish study on "delve" usage (56,878 essays)
• Georgia Tech analysis (168.3M articles)

License

MIT

Author

Jim Christian (@jimchristian)