prose-plugin-development

cc-prose

A Claude Code plugin for proactive content generation with human voice and AI slop prevention.

Overview

This plugin helps you write content that sounds authentically human while following all style rules. It complements the copyedit plugin, sharing its configuration but with a clear separation of concerns:

• copyedit = Review and fix existing content (editing)
• prose = Generate new content with human voice (creation)

Key features:

• Automatic activation: Activates when you request content creation
• Voice profiles: 8 built-in templates plus custom profile creation
• Voice extraction: Derive voice profiles from existing writing samples
• Humanizer depth: Wikipedia's 24-pattern methodology for authentic writing
• Proactive guidance: Pattern prevention during writing (not just post-editing)
• Easy humanizer updates: Sync from upstream source with make sync-humanizer

Note: If you previously used copyedit's content-writer skill, this plugin's content-generator supersedes it. The copyedit plugin should focus on editing, not content creation.

Installation

Via Marketplace (recommended):

# Add the marketplace (once)
/plugin marketplace add rhuss/cc-rhuss-marketplace

# Install the plugin
/plugin install prose@cc-rhuss-marketplace

From source:

git clone https://github.com/rhuss/cc-prose.git
cd cc-prose
make install

# Verify installation
claude plugin list

Commands

Command	Description
/prose:write <topic>	Generate content on a topic
/prose:rewrite <file>	Humanize existing text
/prose:voice [create|apply|list]	Manage voice profiles
/prose:check <file>	Pre-submission validation
/prose:init [--global]	Initialize style configuration

Quick Start

Generate Content

The content-generator skill activates automatically when you request content creation:

# Just ask naturally - the skill activates automatically
Write a section about Kubernetes Services with human voice

# Or use the explicit command
/prose:write a section explaining how Kubernetes Services work

# Write using a specific voice
/prose:write an intro to container networking using the tutorial voice

The content-generator will:

1. Load your copyedit configuration
2. Apply your active voice profile
3. Generate content following all style rules
4. Validate against humanizer patterns
5. Present publication-ready text

Humanize Existing Text

The humanizer detects and removes AI writing patterns from existing text. It works across 24 categories of AI-specific patterns, from vocabulary choices ("delve", "leverage") to structural habits (significance inflation, promotional language, excessive hedging).

# Rewrite a file to remove AI patterns
/prose:rewrite docs/chapter03.md

# Rewrite with a specific voice applied
/prose:rewrite docs/intro.md using the conversational voice

# Or just describe what you need
Tighten up the writing in docs/overview.md, it sounds too robotic

The humanizer operates in two modes:

• Reactive: Cleans up existing AI-sounding text on demand
• Proactive: The content-generator invokes the humanizer's rules during generation, so AI patterns never appear in the first place

Create a Voice Profile

# Create a new voice profile interactively
/prose:voice create technical-friendly

# List available voice profiles
/prose:voice list

# Apply a voice to the current project
/prose:voice apply narrative

You'll be prompted for:

• Formality level (casual to formal)
• Personality level (neutral to opinionated)
• Pronoun preferences (you vs we)
• Contractions usage

Extract a Voice from Existing Writing

The voice-extractor analyzes writing samples and derives a reusable voice profile from them:

# Extract voice from a single document
Extract a voice profile from docs/best-practices.md

# Extract from a directory of samples
Derive a voice from all the posts in content/blog/

# Create a named profile from an author's writing
Learn my voice from ~/writing-samples/ and save it as "my-style"

The extractor reads Markdown, AsciiDoc, and PDF files, analyzing sentence structure, vocabulary choices, formality level, and personality markers. For multi-file corpora, it performs incremental analysis and presents aggregate confidence scores before saving.

Validate Before Submission

# Run comprehensive validation
/prose:check docs/chapter03.md

The pre-validator checks:

• AI pattern scan (24 categories)
• Stoplist word check
• Style compliance
• Flow quality
• Consistency
• Voice consistency
• Readability metrics

Initialize Configuration

# Set up project-level style configuration
/prose:init

# Set up global configuration (applies to all projects)
/prose:init --global