Vibe Coding

About 391 wordsAbout 1 min

Overview

This project was largely developed using Vibe Coding — generating code through iterative conversations with an AI assistant. The developer provides requirements, screenshots, ROIs, UI analysis, and validation; the AI implements pipeline logic, Python custom nodes, CI/CD configuration, and documentation.

Toolchain

MaaMCP

MaaMCP is an MCP server in the MaaFramework ecosystem that provides AI assistants with direct access to MaaFramework projects. The AI uses MaaMCP to understand project structure, read pipeline files, and analyze logs, greatly improving development efficiency.

create-maa-project

create-maa-project is the project scaffolding tool for MaaFramework. It handles project initialization, runtime synchronization, schema validation, and release workflows for this project.

AGENTS.md

AGENTS.md is the AI behavior guide for this project. It defines behavioral norms and constraints for the AI assistant during development, including project structure, documentation references, build commands, coding conventions, and strategies for common development scenarios. The AI learns from this file to ensure generated code aligns with the project's conventions.

Reference Repositories

The AI referenced the following repositories during development:

MaaFramework Repository

A local clone of MaaFramework was used to help the AI understand the framework's API design, pipeline protocol specifications, and implementation details. When encountering framework-level issues, the AI could directly consult the source code.

M9A Repository

A local clone of M9A was used as a reference. M9A is an excellent community MaaFramework project that has been restructured with create-maa-project. It provides valuable reference for pipeline design patterns, custom node implementations, CI/CD configuration, and release workflows.

Development Workflow

Requirements: Developer provides feature requirements, screenshots, and ROIs. If the AI has multimodal capabilities (such as Fable 5), it can also use MaaMCP to autonomously perform screenshot recognition and UI analysis.
AI Implementation: AI references MaaFramework API and M9A implementations, writes pipeline and Python code
Validation: Run pnpm check and pnpm check:py for code quality
Iteration: Developer reviews, provides feedback, AI adjusts
Release: Package and publish using the create-maa-project toolchain

Acknowledgments

Thanks to the MaaFramework team for the powerful framework, the M9A project for excellent reference implementations, and tools like MaaMCP and create-maa-project that greatly reduced the development difficulty of this project.