Stop-Slop Quality Skill

Inspired by: stop-slop — removing AI tells from prose

Purpose: Train AI agents to catch and remove common AI writing patterns (generic phrases, structural clichés, weak prose).

What's Unique About stop-slop

stop-slop identifies AI "tells"—predictable patterns that reveal AI authorship:

• Banned phrases — "harness-native operator", "it's important to note", "ultimately"
• Structural clichés — binary contrasts, dramatic fragmentation, false agency
• Weak prose — passive voice, vague declaratives, Wh-sentence starters

AI Tells Detected

Throat-Clearing Phrases

• "It's important to note that..."
• "As you may know..."
• "It's worth mentioning..."

Emphasis Crutches (Banned Adverbs)

• truly, really, very, literally, certainly, definitely

Business Jargon

• leverage, synergy, paradigm shift, ecosystem

Meta-Commentary

• "The following code demonstrates"
• "As mentioned above"

Implementation

Create agents/quality_checker.py to check/clean text:

class StopSlopChecker:
    def check_text(self, text: str) -> list[dict]:
        """Find AI tells in text"""
    
    def clean_text(self, text: str) -> str:
        """Remove AI tells"""

Integration Points

• Pre-commit hook to check commit messages
• Pre-PR hook to check descriptions
• Real-time filtering during agent output

References

• stop-slop: https://github.com/hardikpandya/stop-slop
• Quick-Note Issue: #229