Perancangan Framework Vibe Coding yang Bertanggung Jawab dalam Pengembangan Perangkat Lunak

Abstract

The advancement of large language models (LLM) has given rise to vibe coding, a prompt-based programming paradigm that enhances productivity but introduces serious risks including reduced code comprehension, low maintainability, accumulation of technical debt, and application security vulnerabilities. This research aims to design a Responsible Vibe Coding Framework that integrates software engineering principles (SOLID), computational thinking, human-AI collaboration, and secure, maintainable software development practices. Employing a qualitative-conceptual approach through systematic literature review of publications from 2015-2025, the study identifies the phenomenon, analyzes supporting theories, and synthesizes a framework comprising five core components: Input Layer (intent specification), AI Assistance Layer (LLM code generation), Human Oversight Layer (human verification), Engineering Guardrails Layer (automated testing and security scanning), and Output Layer (production-ready code). The framework ensures human-centered decision-making, integrated software engineering discipline, and transparency with accountability, comprehensively addressing vibe coding challenges while maintaining code quality, security, and comprehension in educational and industrial contexts.