| name | ai-prose-cleaner |
| description | Detect and clean AI-typical writing patterns in documents — em dashes, filler phrases, over-formality |
| tags | ["writing","cleanup","documents"] |
When to Use
- Documents sound "too AI" — em dashes, "delve", "it's worth noting", hedging stacks
- Before sending proposals, READMEs, or reports to clients/reviewers
- After AI-assisted writing that needs a human voice pass
- Batch scan agent/skill definitions for AI fingerprints
- User says "deslop prose", "clean AI writing", "humanize"
Workflow
Delegates to @omg:ai-prose-cleaner agent, which operates in two modes:
- Detect (default): Scan files, report AI patterns with line numbers, severity rating (LOW → HEAVY)
- Clean: Rewrite flagged patterns in place, preserving meaning. Tier 1 patterns auto-fixed, Tier 3 structural changes suggested for manual review.
Supported Languages
English and German AI patterns detected (Halbgeviert-Striche, "Es ist erwähnenswert", etc.)
Trigger Keywords
clean AI writing, humanize, deslop prose, AI speech, prose cleaner
Examples
@omg:ai-prose-cleaner scan docs/proposal.md
@omg:ai-prose-cleaner clean README.md
@omg:ai-prose-cleaner scan plugin/agents/*.md
Quality Contract
- Every rewrite preserves factual content — only voice changes
- Report includes line numbers, pattern tier, and suggestions
- Rating scale: patterns per 1,000 words (LOW < 1, MILD 1-3, MODERATE 3-6, HEAVY > 6)
