humanizer and anti slop

Standard LLM drafts often suffer from predictable rhythms, bloated corporate jargon, and formulaic structures that scream "AI-generated." This skill identifies these synthetic patterns and provides concrete rewrites to restore a punchy, authentic human voice.

What it does

• Pattern identification scans for eight specific "slop" archetypes including throat-clearing openers and binary contrasts.
• Rhythmic analysis calculates sentence burstiness to break up the monotonous 14-18 word drone typical of AI.
• Vocabulary mapping replaces high-risk tells like "delve," "tapestry," and "leverage" with direct technical terms.
• Contraction auditing identifies unnatural formality and suggests conversational adjustments for better flow.
• Structural refactoring strips away formulaic persona bookends and mechanical conclusion paragraphs.

How it works

1. Analyze the target text by providing a draft or piping a file through the audit_slop.py utility.
2. Review the burstiness index and risk score to identify specific sentences requiring rhythmic variation.
3. Execute the humanize commands or MCP tools to swap prohibited vocabulary for precise, grounded alternatives.

Frameworks & tools

The skill includes a standalone Python 3.8+ audit script for CLI or CI/CD pipelines. It also supports Model Context Protocol (MCP) integrations via the humanizer and de_slop server tools.

Why this beats prompting it yourself

Generic "write like a human" prompts often result in overly casual or quirky text that loses professional authority. This tool uses a structural approach based on sentence variance and specific banned patterns to ensure technical precision remains intact while the "slop" is removed.

Use cases

• Docs review to strip robotic preamble from technical guides and API documentation.
• Content publishing to ensure blog posts pass style consistency benchmarks and AI detection.
• Product announcements to replace artificial grandiosity with concrete performance metrics.
• CI/CD gating to automatically flag PRs that introduce low-burstiness or high-jargon text.

Known limitations

Does not support raw code, API specs, or machine-readable configs. Explicitly avoids legal contracts where boilerplate is mandatory and math proofs where formulaic transitions are standard.