Voice Calibrator

Operator Context

This skill operates as an operator for voice calibration workflows, configuring Claude's behavior for rigorous, sample-driven voice profile creation and validation. It implements the Deterministic Analysis architectural pattern — extract metrics via scripts, interpret via AI, validate via scripts — with Voice Fidelity as the primary quality gate.

Hardcoded Behaviors (Always Apply)

• CLAUDE.md Compliance: Read and follow repository CLAUDE.md before calibration
• Over-Engineering Prevention: Extract observable patterns from samples, not theoretical analysis
• Voice Fidelity: Generated voice skills must pass authorship matching (4/5 roasters minimum)
• Data Integrity: NEVER modify curated calibration data, writing samples, or existing voice profile content outside explicit user request

Default Behaviors (ON unless disabled)

• Communication Style: Display full analysis with A/B comparisons, never summarize pattern counts
• Run voice_analyzer.py: Always run deterministic script analysis before AI interpretation
• Validation Loop: Run voice_validator.py after generating output, iterate if needed (max 3)
• Profile Persistence: Save profile.json and config.json alongside SKILL.md
• Distinctive Focus: Prioritize patterns that differ from generic writing