Vai al contenuto
TakoAPI
Torna alle skill

twinify

di neobotjan2026

Web & Frontend Development
955Download riportati da clawskills.sh
0Visualizzazioni
0 likeAccedi per mettere like a questa skill
—Non ancora valutato
La tua valutazione

Sintesi

Create AI digital twins of real people from WhatsApp chat history exports.

Descrizione

Cosa fa questa skill

Creates AI digital twins of real people by analyzing WhatsApp chat export files. The skill parses messages from a .txt export, builds a structured personality profile, and registers a new OpenClaw agent that mimics the target person's communication style. Explicit consent from the person being cloned is required before proceeding.

Derives personality profiles directly from real message history rather than requiring manual persona writing or structured interviews.

Quando usarla

  • Clone a friend's texting style into a chatbot you can message anytime
  • Check how a colleague might phrase a response before sending a draft
  • Build a personal AI modeled on your own saved chat history
  • Reconstruct communication patterns of someone no longer reachable
  • Test persona fidelity by comparing agent replies to real archived messages

Esempio di workflow

INPUT

User asks: Create a twin of my friend Alex from this WhatsApp export

AGENT
  1. 1

    Asks the user to confirm they have Alex's explicit, informed consent before proceeding

  2. 2

    Runs parse_chat.py on the exported .txt file to extract Alex's messages into parsed_messages.json

  3. 3

    Analyzes parsed messages to identify recurring phrases, tone patterns, emotional shifts, and shared history

  4. 4

    Generates SOUL.md, EXAMPLES.md, ANTI-EXAMPLES.md, and MEMORY.md profile files in the twin's workspace

  5. 5

    Creates and registers a new OpenClaw agent in ~/.openclaw/ configured with the generated profile

OUTPUT

A new OpenClaw agent that responds as Alex, ready for test messages and iterative refinement

Installazione

clawhub install neobotjan2026/twinify

Esegui il comando CLI oppure incolla il prompt in Claude Code, Cursor, Codex o nell'agente di coding che usi.

Link