An Agent-Driven Framework for Task-Level SDLC Automation

Abstract

Recent advances in agentic artificial intelligence have enabled partial automation of software engineering workflows; however, most existing systems focus on isolated SDLC phases or high-level project planning. This paper proposes a task-centric, agent-driven framework that automates the Software Development Life Cycle (SDLC) end-to-end by decomposing requirements into systems, tasks, and dependencies while dynamically assigning priorities, deadlines, and execution order. Unlike prior approaches that operate at phase or project granularity, the proposed framework enables fine-grained task-level orchestration with parallel execution and dependency-aware scheduling. Implemented using a multi-agent architecture built on large language model (LLM) orchestration, the system demonstrates how autonomous agents can collaboratively manage SDLC activities while maintaining human-aligned control and cost awareness. The framework is evaluated conceptually through realistic software project scenarios and positioned as a scalable alternative to traditional project management tools such as Jira.