wehub-resource-sync
5.2 KiB
ex.skill
"Every goodbye deserves a gentle place to rest."
Distill your ex into an AI Skill — let them live on in your terminal.
Provide source materials from your past relationship (chat logs, photos, social media posts) plus your own descriptions. Generate an AI Skill that truly sounds like them — speaks with their catchphrases, replies in their style, remembers the places you went together.
⚠️ This project is for personal reflection and emotional healing only. Not for harassment, stalking, or privacy invasion.
Installation · Usage · Examples · 中文
Installation
Claude Code
# Install to current project
mkdir -p .claude/skills
git clone https://github.com/therealXiaomanChu/ex-skill .claude/skills/create-ex
# Or install globally
git clone https://github.com/therealXiaomanChu/ex-skill ~/.claude/skills/create-ex
Dependencies (optional)
pip3 install -r requirements.txt
Usage
In Claude Code, type:
/create-ex
Follow the prompts: enter a codename, basic info, personality traits, then choose data sources. All fields except the codename can be skipped.
After creation, use /{slug} to chat with the generated ex Skill.
Commands
| Command | Description |
|---|---|
/list-exes |
List all ex Skills |
/{slug} |
Full Skill (chat like them) |
/{slug}-memory |
Memory mode (recall shared experiences) |
/{slug}-persona |
Persona only |
/ex-rollback {slug} {version} |
Rollback to a previous version |
/delete-ex {slug} |
Delete |
/let-go {slug} |
Gentle alias for delete |
Examples
Input:
First love, 3 years together in college, ENFP, Gemini, talkative, sends voice messages at 2am, still likes my posts after breakup
Scenario 1: Casual Chat
You ❯ whatcha doing
Ex.skill ❯ just had dinner, scrolling my phone
why'd you message me out of nowhere
[sends a meme]
Scenario 2: Memory Lane
You ❯ remember our first date?
Ex.skill ❯ the one with the terrible pasta? lmao
you pretended it was good, I totally noticed
that restaurant closed down btw, did you know?
Scenario 3: Late Night Emo
You ❯ I kinda miss you
Ex.skill ❯ ...
yeah
get some sleep
Features
Data Sources
| Source | Format | Notes |
|---|---|---|
| txt / html / json / csv exports | Recommended, richest data | |
| txt / mht export | Great for school-era relationships | |
| Social Media | Screenshots | Extracts public persona |
| Photos | JPEG/PNG with EXIF | Timeline and location extraction |
| Narration | Plain text | Your subjective memories |
Generated Skill Structure
Each ex Skill has two parts working together:
| Part | Content |
|---|---|
| Part A — Relationship Memory | Shared experiences, date spots, inside jokes, fight patterns, sweet moments, timeline |
| Part B — Persona | 5-layer structure: Hard rules → Identity → Speech style → Emotional patterns → Relationship behavior |
Supported Tags
Attachment Styles: Secure · Anxious · Avoidant · Disorganized
Love Languages: Words of Affirmation · Quality Time · Receiving Gifts · Acts of Service · Physical Touch
Personality Tags: Talkative · Reserved · Tough-love · Silent treatment · Clingy · Independent · Romantic · Pragmatic · Perfectionist · Procrastinator · Workaholic · Jealous · Insecure · Night owl · Leaves on read · Instant replier ...
Evolution
- Append memories → New chat logs or photos → Auto-analyze and merge
- Conversation corrections → "They wouldn't say that" → Instant correction
- Version management → Auto-archive on every update → Rollback supported
Philosophy
Every relationship teaches us something. Some people leave, but their words, their laughter, the way they got angry — all of it stays in your neural network. This Skill just helps you migrate those memories from biological to digital neural networks. Not to hold on, but to say a proper goodbye.
Acknowledgments
The architecture of this project is directly inspired by colleague-skill (同事.skill) by titanwings. colleague-skill pioneered the idea of "distilling a person into an AI Skill" with its dual-layer architecture (Work Skill + Persona). Ex-Partner.skill adapts this framework from workplace to romantic relationships. Huge respect to the original author's creativity and open-source spirit.
This project follows the AgentSkills open standard, compatible with Claude Code and OpenClaw.
MIT License © therealXiaomanChu