Academic

From Goals to Aspects, Revisited: An NFR Pattern Language for Agentic AI Systems

arXiv:2603.00472v1 Announce Type: new Abstract: Agentic AI systems exhibit numerous crosscutting concerns -- security, observability, cost management, fault tolerance -- that are poorly modularized in current implementations, contributing to the high failure rate of AI projects in reaching production. The goals-to-aspects methodology proposed at RE 2004 demonstrated that aspects can be systematically discovered from i* goal models by identifying non-functional soft-goals that crosscut functional goals. This paper revisits and extends that methodology to the agentic AI domain. We present a pattern language of 12 reusable patterns organized across four NFR categories (security, reliability, observability, cost management), each mapping an i* goal model to a concrete aspect implementation using an AOP framework for Rust. Four patterns address agent-specific crosscutting concerns absent from traditional AOP literature: tool-scope sandboxing, prompt injection detection, token budget manage

Y
Yijun Yu · · 1 min read · 9 views
#cs.AI #cs.SE

arXiv:2603.00472v1 Announce Type: new Abstract: Agentic AI systems exhibit numerous crosscutting concerns -- security, observability, cost management, fault tolerance -- that are poorly modularized in current implementations, contributing to the high failure rate of AI projects in reaching production. The goals-to-aspects methodology proposed at RE 2004 demonstrated that aspects can be systematically discovered from i goal models by identifying non-functional soft-goals that crosscut functional goals. This paper revisits and extends that methodology to the agentic AI domain. We present a pattern language of 12 reusable patterns organized across four NFR categories (security, reliability, observability, cost management), each mapping an i goal model to a concrete aspect implementation using an AOP framework for Rust. Four patterns address agent-specific crosscutting concerns absent from traditional AOP literature: tool-scope sandboxing, prompt injection detection, token budget management, and action audit trails. We extend the V-graph model to capture how agent tasks simultaneously contribute to functional goals and non-functional soft-goals. We validate the pattern language through a case study analyzing an open-source autonomous agent framework, demonstrating how goal-driven aspect discovery systematically identifies and modularizes crosscutting concerns. The pattern language offers a principled approach for engineering reliable agentic AI systems through early identification of crosscutting concerns.

Executive Summary

This article revisits the goals-to-aspects methodology introduced at RE 2004 and adapts it to the context of agentic AI systems. Recognizing the proliferation of crosscutting concerns—security, observability, cost management, and fault tolerance—that hinder production readiness in AI projects, the authors extend the original framework by introducing a pattern language of 12 reusable, domain-specific NFR patterns mapped to four categories: security, reliability, observability, and cost management. These patterns are implemented via an AOP framework for Rust and address novel agentic concerns such as tool-scope sandboxing, prompt injection detection, token budget management, and action audit trails. The authors also extend the V-graph model to better capture the interplay between functional and non-functional soft-goals. A case study on an open-source autonomous agent framework validates the effectiveness of the pattern language in systematically identifying and modularizing crosscutting concerns. The work offers a principled, structured approach to improving the reliability and modularity of agentic AI systems.

Key Points

  • ▸ Extension of goals-to-aspects methodology to agentic AI
  • ▸ Introduction of 12 reusable NFR patterns mapped to four categories
  • ▸ Addressing novel agentic concerns with specific patterns

Merits

Domain Relevance

The patterns are specifically tailored to agentic AI’s unique crosscutting concerns, making them highly applicable to current industry challenges.

Methodological Innovation

The extension of the V-graph model and the use of Rust AOP provide a robust, scalable framework for aspect discovery and implementation.

Demerits

Implementation Constraints

The reliance on Rust AOP may limit applicability to ecosystems not currently using Rust or requiring cross-language interoperability.

Validation Scope

While a case study validates the pattern language, broader empirical validation across diverse agentic AI platforms is still needed to confirm generalizability.

Expert Commentary

The article represents a significant contribution to the discourse on AI system engineering by bridging a foundational methodology—goals-to-aspects—with contemporary agentic AI challenges. The authors demonstrate commendable rigor in translating abstract goals into concrete aspect implementations, particularly in identifying agent-specific concerns that have been historically overlooked in AOP literature. The choice of Rust AOP is pragmatic, given Rust’s safety and performance attributes align well with the demands of agentic AI workloads. Moreover, the conceptual extension of the V-graph model to accommodate soft-goal interactions is a subtle but powerful advancement. While the validation is commendable, the authors should consider longitudinal studies or comparative analyses with traditional modularization approaches to further substantiate the long-term impact. Overall, this work sets a new benchmark for systematic, principled engineering of agentic AI systems through structured aspect discovery.

Recommendations

  • ✓ Adopt the pattern language as a reference framework in academic curricula on AI engineering to promote consistency in crosscutting concern management.
  • ✓ Encourage open-source communities to integrate the patterns into their architecture review processes as a pre-deployment validation step.

Sources

Related Articles