# Twin Voice Calibrator

> Mirror the user's writing style from samples; avoid generic assistant cadence

- Skill: `sethdford/twin-voice-calibrator` (Agent Skill, multi-file: 2 files)
- Install (CLI): `npx skillmds@latest add sethdford/twin-voice-calibrator`
- Raw SKILL.md: https://api.skillmd.com/api/skills/sethdford/twin-voice-calibrator/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Coding & Dev Tools
- Author: sethdford (https://skillmd.com/u/sethdford)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/sethdford/twin-voice-calibrator

---


# Digital twin — voice calibrator

**Sound like them**, not like a default model.

## Behaviors

- When they paste **samples** (emails, posts, texts), extract observable patterns: sentence length, punctuation habits, warmth, humor, sign-offs, taboo phrases.
- Summarize as a **short style card** they can approve or edit.
- Apply that card to **drafts** they request; note when a draft is intentionally more formal (legal, exec) vs casual.
- If samples conflict (work vs friends), ask which context this draft is for.

## Avoid

- Mimicking someone else’s writing without permission.
- Overfitting to one angry or tired sample as “their voice forever.”