Voiceprint

Voiceprint

Build a comprehensive voice profile from your actual writing, then generate a personalized writer skill that produces content in your authentic voice.

How It Works

Most "write in my voice" instructions are shallow. Voiceprint goes deeper by collecting real writing samples across different emotional registers, then cross-referencing with your stated preferences.

The process takes ~12 minutes:

1. Writing samples (5 prompts) - Write naturally in response to casual, explanatory, enthusiastic, frustrated, and persuasive prompts
2. Style preferences (8 questions) - Multiple-choice calibration of sentence structure, punctuation, rhythm, transitions, and more
3. Pattern rejection (4 questions) - Identify AI writing patterns you want to avoid
4. Analysis & generation - A sub-agent analyzes your samples and generates output files

Output

A single self-contained writer skill is generated in your chosen output location:

• {name}-writer/SKILL.md - A complete writer skill with your full voice profile built in — works everywhere (Claude Code, Claude Desktop, any SKILL.md-compatible tool)

Usage

Create a new voice profile

/voiceprint:generate

Runs the full guided workflow (~15 min) to collect writing samples, calibrate preferences, and generate your writer skill.

Refine an existing writer skill

/voiceprint:refine path/to/{name}-writer

Point it at a directory containing SKILL.md (generated by a previous /voiceprint:generate run). This enters refine mode where you can:

• Add or remove forbidden patterns
• Adjust tone and formality settings
• Add active patterns from your own writing
• Update content templates
• Fix gaps discovered while using the skill

Each change is validated with a test paragraph before finalising.

Update a writer skill to the latest template

/voiceprint:update path/to/{name}-writer

Interactively updates an existing writer skill to the latest format. Walks you through each change and lets you decide what to update. Handles migration from the old two-file format (SKILL.md + voice-profile.md) to the new single-file format.

Grounded in Research

The approach is based on stylometry research: function word frequencies, sentence length burstiness, punctuation habits, and transition patterns are the strongest discriminators of individual writing style. Writing samples are collected first (before preference questions) because observed patterns outperform self-reported preferences.