Extract personal facts from your ChatGPT conversation history and generate a comprehensive profile. All processing happens locally using your own LLM server - your conversations never leave your machine.

1. Extracts personal facts from your ChatGPT conversations (career, values, interests, relationships, etc.)
2. Merges duplicate and overlapping facts into a consolidated set
3. Synthesizes a readable personal profile document

Your ChatGPT conversations contain years of context about who you are. This skill distills that into a structured profile that can help AI assistants understand you better.

Install the facts-from-openai skill from https://github.com/getaifs/facts-from-openai

# Install this skill
git clone https://github.com/getaifs/facts-from-openai.git ~/.claude/skills/facts-from-openai

# Install dependencies (other skills this relies on)
git clone https://github.com/getaifs/chatgpt-export.git ~/.claude/skills/chatgpt-export
git clone https://github.com/getaifs/qwen3-skill.git ~/.claude/skills/qwen3

git clone https://github.com/getaifs/facts-from-openai.git ~/.claude/skills/facts-from-openai
git clone https://github.com/getaifs/chatgpt-export.git ~/.claude/skills/chatgpt-export
git clone https://github.com/getaifs/qwen3-skill.git ~/.claude/skills/qwen3

1. Get your ChatGPT export: ChatGPT > Settings > Data Controls > Export data

   • You'll receive an email with a download link
   • The downloaded file is typically named something like 2024-01-15-conversations.zip

2. Set up local LLM (if you don't have one running):

   python3 ~/.claude/skills/qwen3/setup.py install

3. Run the pipeline (use the actual path to your export ZIP):

   python3 ~/.claude/skills/facts-from-openai/facts.py run \
     --export /path/to/your/chatgpt-export.zip \
     --output ~/my-profile

4. View your profile: ~/my-profile/profile.md

Or just ask Claude Code:

Build a personal profile from my ChatGPT export at /path/to/export.zip

The script accepts any valid path - absolute, relative, or with ~ for home directory.

• All processing happens locally - conversations are processed by your local LLM
• Claude never sees your conversations - only orchestrates the pipeline
• Intermediate files are protected - extraction files are marked so Claude won't read them
• Only the final profile is shared - the synthesized output contains no raw conversation data

• Python 3.8+
• Local LLM server (ollama or llama.cpp)
• ChatGPT data export (ZIP file)

MIT License - See LICENSE for details.