GoalForge AI: A Seven-Agent Orchestration Framework for Autonomous and Adaptive Study Planning

This paper presented GoalForge AI, a seven-agent orchestration framework that reframes study planning as a coordinated pipeline of specialized, individually auditable agents rather than a single opaque model call. By combining LLM-driven agents for goal decomposition, research, and reporting with a deterministic, rule-based Decision agent, the system achieves a balance between adaptability and explainability that is difficult to obtain from a single end-toend model. The working implementation, spanning a Node.js and Express backend, an SQLite persistence layer, and a React frontend, demonstrates the technical feasibility of this architecture, and the persistent audit log provides a concrete mechanism for inspecting an agentic system's reasoning — a property of growing importance as such systems are adopted in education and other human-centered domains. More broadly, the design pattern demonstrated here — pairing LLM-driven agents for open-ended reasoning tasks with deterministic, rule-based agents for safety- or trustcritical decisions — is not specific to study planning and may generalize to other agentic applications where explainability and predictable behavior are as important as adaptability. Future work will focus on four directions: empirical evaluation with real learners over a full academic term to measure plan adherence against a static-timetable baseline; an automatic replanning trigger so the Validation agent can invoke the Planner and Decision agents again when a learner falls significantly behind, rather than requiring manual goal recreation; richer memory modeling capable of capturing individual learning styles and historically effective task densities rather than a single last-used strategy; and migration of the persistence layer to a networked database engine to support concurrent, multi-institution deployment at scale.

[1]S. Russell and P. Norvig, "Artificial Intelligence: A Modern Approach," 4th ed., Pearson, 2021. [2]T. Brown et al., "Language Models are Few-Shot Learners," in Proc. NeurIPS, 2020. [3]Google AI, "Gemini API Documentation," Google DeepMind, 2025. [4]OpenJS Foundation, "Node.js SQLite Built-in Module Documentation," Node.js Docs, 2025. [5]   React Core Team, "React Documentation," Meta Open Source, 2025. [6]M. Wooldridge, "An Introduction to MultiAgent Systems," 2nd ed., Wiley, 2009. [7]   IEEE Computer Society, "IEEE Standard for Software and System Test Documentation," IEEE Std 829-2008, 2008. [8]   OWASP Foundation, "OWASP Authentication Cheat Sheet," OWASP Foundation, 2024.