lingyichen-ai/next-chat-skills

Next-Chat-Skills is an AI chat application that can actually do things for you. Instead of just answering questions in plain text, it can autonomously decide to run scripts, install software dependencies, and generate files based on your conversation. Think of it as a conversational AI that has hands, it can write and execute code on your behalf, rather than just suggesting what you should do. The way it works is through a concept called "Skills." You define these skills as files in a local folder on your computer. Each skill is described in a markdown file with some metadata at the top that tells the AI what the skill does and how to trigger it. When you chat with the assistant, it reads through your available skills and decides on its own when one is relevant. If it decides to use a skill, it runs the associated script in a sandboxed environment, safely executing whatever needs to happen and handing the result back to you in the chat. This would be useful for anyone who wants a personal assistant that goes beyond text responses. For example, if you're a founder or PM prototyping a product, you could describe what you want, and the AI could generate the actual project files and install the libraries you need. If you have recurring tasks that involve running scripts, processing data, formatting files, setting up environments, you could register those as skills and just ask the AI to handle it conversationally. The project is built with Next.js and supports two storage backends: SQLite for quick local setup with no configuration, and PostgreSQL if you need to run it in production across multiple instances. Tables are created automatically on startup, so there's no manual database setup required. The Docker deployment image comes bundled with both Node.js and Python 3, reflecting that skills can execute scripts in either language. The README doesn't go into detail on how skills are structured beyond the markdown-with-YAML-frontmatter format, so you'd need to look at examples in the skills directory to understand the authoring process.