Deslop Prose Humanizer

•
30
•

Removes AI-generated patterns and stylistic tells from documentation, pull requests, and communication before they reach human readers.

Deslop is a specialized capability for Claude Code designed to audit and refine AI-generated text by identifying and stripping common machine artifacts across character, word, and document-structure levels. Unlike generic rewriters, it utilizes a rigorous five-pass methodology—mechanical, structural, rhetorical, sentential, and lexical—to ensure prose sounds human and contextually appropriate for various surfaces like READMEs, PR bodies, and blog posts. It prioritizes the removal of verified AI tells over subjective stylistic improvements, maintaining the author's intended signal while eliminating the repetitive 'slop' that often characterizes model outputs.

주요 기능

01Context-aware processing for PRs, READMEs, Slack, and commit messages
0230 GitHub stars
03Automated 'slopscan' script for rapid mechanical and character-level checks
04Role separation between detection and rewriting to prevent self-grading bias
05Structural analysis targeting high-frequency AI tells like nominalizations
06Rigorous five-pass auditing system (Mechanical through Lexical)

사용 사례

01Humanizing AI-generated documentation and README files for public repositories
02Cleaning up pull request bodies and commit messages to match engineering standards
03Removing repetitive AI vocabulary and 'hedging' patterns from technical blog posts