titanwings/ex-skill

What it solves

It allows users to create a digital simulation of a former partner (or any specific person) to maintain a sense of connection or find closure. It solves the problem of losing the specific communication style, tone, and emotional patterns of a person by synthesizing their digital footprint into a conversational AI persona.

How it works

Users provide chat histories (from WeChat or iMessage), screenshots, or text descriptions along with personality traits (MBTI, astrology, attachment styles). The system analyzes this data to build a two-part digital persona:

1. Relationship Memory: A database of shared memories, inside jokes, and conflict patterns.
2. Persona Structure: A six-layer personality model covering core rules, identity, expression, emotional behavior, conflict boundaries, and triggers.

This persona is then deployed as a "Skill" compatible with agent frameworks like OpenClaw, Claude Code, or DeepSeek Harness, allowing the user to chat with the simulation in real-time.

Who it’s for

People who want to simulate conversations with an ex-partner or a specific individual, and users of AI agent frameworks who want to create highly personalized, emotionally accurate digital personas.

Highlights

• Automatic Data Extraction: Supports auto-decryption and extraction from WeChat (Windows/macOS) and iMessage (macOS).
• Multi-Layered Personality: Uses a sophisticated 6-layer structure to mimic not just what a person says, but how they react to conflict and when they go silent.
• Iterative Refinement: Users can add more messages to update the persona or provide conversational corrections (e.g., "they wouldn't say that") to refine behavior.
• Version Control: Includes the ability to archive and rollback to previous versions of the generated persona.