SHIELDA: Structured Handling of Exceptions in LLM-Driven Agentic Workflows

2025-08-11

AI Summary

LLM-driven autonomous workflows frequently suffer from execution failures caused by reasoning errors, yet existing approaches lack root-cause diagnosis capabilities and robust recovery mechanisms. To address this, we propose SHIELDA—a novel framework that enables end-to-end anomaly awareness, classification, and structured recovery across both reasoning and execution phases. SHIELDA introduces a fine-grained taxonomy covering 36 distinct anomaly types; a modular, composable library of exception-handling patterns; and a hierarchical recovery executor supporting phase-aware local handling, workflow redirection, and state rollback. Evaluated on the AutoPR agent, SHIELDA accurately identifies root causes at the reasoning layer and achieves multi-level automatic recovery, significantly improving workflow robustness and success rate. This work establishes the first unified framework for cross-phase anomaly management in LLM-based autonomous systems.

Technology Category

• Humans and AI: Human-Aware Planning and Behavior Prediction
• Knowledge Representation and Reasoning: Diagnosis and Abductive Reasoning
• Natural Language Processing: Safety and Robustness

Application Category

• Responsible Web: Machine-in-the-loop, human agency and autonomy
• Search and Retrieval-Augmented AI: Agentic search
• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions

Abstract

Large Language Model (LLM) agentic systems are software systems powered by LLMs that autonomously reason, plan, and execute multi-step workflows to achieve human goals, rather than merely executing predefined steps. During execution, these workflows frequently encounter exceptions. Existing exception handling solutions often treat exceptions superficially, failing to trace execution-phase exceptions to their reasoning-phase root causes. Furthermore, their recovery logic is brittle, lacking structured escalation pathways when initial attempts fail. To tackle these challenges, we first present a comprehensive taxonomy of 36 exception types across 12 agent artifacts. Building on this, we propose SHIELDA (Structured Handling of Exceptions in LLM-Driven Agentic Workflows), a modular runtime exception handling framework for LLM agentic workflows. SHIELDA uses an exception classifier to select a predefined exception handling pattern from a handling pattern registry. These patterns are then executed via a structured handling executor, comprising local handling, flow control, and state recovery, to enable phase-aware recovery by linking exceptions to their root causes and facilitating composable strategies. We validate SHIELDA's effectiveness through a case study on the AutoPR agent, demonstrating effective, cross-phase recovery from a reasoning-induced exception.

Problem

• Traces execution exceptions to reasoning-phase root causes
• Provides structured escalation for failed recovery attempts
• Classifies and handles 36 exception types in LLM workflows

Innovation

• Comprehensive taxonomy of 36 exception types
• Modular runtime exception handling framework
• Structured handling executor for root cause recovery