Voice Extraction Skill

Extract a user's writing voice through SICO comparative analysis.

Method: Comparative Feature Extraction

Rather than measuring surface metrics, this skill uses SICO Phase 1: feed the model user writing samples alongside its own default output on the same topics. The model describes what the user does differently. This produces voice descriptions that encode implicit structural patterns no metrics can capture.

Key Principles (from research)

1. Anonymize samples: Label as "Sample 1", "Sample 2", etc. Context labels cause the extractor to anchor on content rather than reading a unified voice.