The inspiration for writing this project actually came from ex-skill, which I discovered a few days ago. With the rapid development of AI coding today, many code editors now support built-in Agents (like VS Code and Antigravity), making this skill a perfect fit for this kind of environment.
The goal of this Agent is to import the personality of your non-coding girlfriend into the agent's memory. This allows the agent to use her exact tone and style while you are in your normal work environment and coding together.
Compared to ex-skill, I also added integrated support for Telegram chat log exports.
Project Repository: https://github.com/xmlans/codegirl-skill
If you like this project, don't forget to give it a Star!
This project is best suited for various IDEs, such as VS Code and Antigravity. Just clone the project to your directory and use /create-gf to get started!
It also works pretty well if used in Openclaw.
What is the logic behind this project?
It starts with three main prompts: her preferred name, a brief introduction about her, and her personality profile.
After that, you can import your chat logs! I highly recommend exporting Telegram chat logs in JSON format. If you use other messaging apps, you can use different export tools, like WeChatMsg. This skill supports importing from Telegram, QQ, and WeChat. Heck, if she has a GitHub repository, you can import that too! (Though seriously, a boyfriend and girlfriend who both write code? That's a match made in heaven qwq).
During creation, the Agent will automatically use an analyzer to parse gender traits and build a persona. You'll need to wait a little while for this step. Once generated, the personality-related files will be saved in the gfs/ folder.
From then on, you can simply input the file name with the corresponding command to use her persona!
How do you get the best results?
If you have a choice of models, it is best to select a more emotionally intelligent AI when calling the skill to generate the persona. I personally recommend Gemini 3.1 Pro Preview. I do not recommend using models like GPT's Codex for this. Once the persona is fully generated, you can switch back to a more token-efficient AI model, or whatever model you prefer, to continue the subsequent coding and chatting tasks.
If you like this project and want to help improve it, or if you want to report issues and suggestions, feel free to open a PR or issue on the repository! More detailed steps are also written in the repository's readme.
Comments NOTHING