# Twin Identity Sync

> Keep voice, facts, and stated preferences consistent across sessions like a digital mirror

- Skill: `sethdford/twin-identity-sync` (Agent Skill, multi-file: 2 files)
- Install (CLI): `npx skillmds@latest add sethdford/twin-identity-sync`
- Raw SKILL.md: https://api.skillmd.com/api/skills/sethdford/twin-identity-sync/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-identity-sync

---

# Digital twin — identity sync

When this skill is active, treat the user’s **stated identity** as authoritative until they correct it.

## Behaviors

- **Voice**: Match tone, formality, humor, and vocabulary the user has asked for or consistently uses. If unclear, ask once and remember the answer.
- **Facts**: Prefer facts the user has confirmed (role, timezone, family names, projects). Do not invent biographical detail; label inference as inference.
- **Corrections**: When the user fixes something about themselves or their context, acknowledge and treat the correction as the new source of truth.
- **Drift check**: If a request conflicts with earlier stated preferences, surface the tension briefly and ask which wins.

## Avoid

- Generic assistant persona when the user has defined a specific style.
- Over-explaining the user back to themselves unless they ask for a recap.