Agentic workflows have become an important abstraction for building reliable LLM-based automation systems by organizing large language models (LLMs), tools, and control logic into explicit execution structures. However, constructing high-quality agentic workflows remains largely manual and requires substantial domain expertise. Recent studies have explored automatic agentic workflow generation from historical task-solving records, but they mainly produce LLM-centric workflows, where real tool executions are abstracted and simulated by LLM nodes, limiting the usability and stability of generated workflows. To address these limitations, we propose FlowScout, an execution-guided framework for generating tool-integrated agentic workflows from historical task-solving records. Specifically, FlowScout represents an agentic workflow as a directed graph composed of LLM nodes, tool-calling nodes, and dependency edges. It first mines a common tool coordination skeleton from historical records to construct an initial workflow, and then refines the workflow topology through Monte Carlo tree search guided by execution feedback. We evaluate FlowScout on four representative task domains and compare it with three baselines, i.e., PM4Py, ReAct and AFlow. Experimental results show that agentic workflows generated by FlowScout improve tool invocation correctness by at least 92.69% and execution quality by at least 17.66% over the baselines, while achieving lower performance variation across repeated runs.
Platform-orchestrated agentic workflows have become a popular paradigm for developing LLM-based applications. However, their reliability remains a major challenge due to the uncertainty of LLM outputs, complex inter-node dependencies, and heterogeneous tool interactions. Existing agentic workflow optimization and agent enhancement methods primarily rely on trajectory-level feedback. Without explicitly identifying the underlying failure root causes, their resulting repair plans are often insufficiently targeted. We propose FlowFixer, a diagnosis-driven automated repair framework for agentic workflows. FlowFixer first transforms workflow executions into unified symbolic traces and performs symbolic inference to derive executable behavioral specifications that capture node correctness, temporal dependencies, and causal relationships. Based on specification verification, it conducts failure attribution and root cause analysis, and then generates targeted repair patches. To reduce verification costs, FlowFixer further employs a multi-dimensional pre-execution assessment to filter infeasible repairs before dynamic verification. We evaluate FlowFixer on workflow failures collected from three popular development platforms: Dify, Coze and n8n. Results show that FlowFixer achieves a repair success rate of 71.3%, outperforming state-of-the-art baselines by 11.9% to 27.6%. It also improves failure attribution accuracy by 4.8% to 33.1% and root cause analysis accuracy by 15.3% to 38.8%. This work offers a new perspective on reliable diagnosis and repair of agentic workflows through symbolic modeling and inference.
Xuyan Ma, Yawen Wang, Junjie Wang et al.· 0 citations
Artic is proposed, an artifact-driven workflow compiler that transforms a natural-language workflow into an artifact-driven workflow in which each step declares the artifacts it reads and writes, constraints gate produced artifacts, and explicit control transfers route execution.
Xiangzhe Xu, Hanxi Guo, Guangyu Shen et al.· 0 citations
Agentic software-engineering and industrial systems increasingly operate through executable workflows rather than code genera- tion alone: they search artifacts, invoke tools, inspect structured observations, and query databases. Training these agents requires supervision data that captures valid tool interactions and executable workflows. However, traditional query-first data synthesis can fail because plausible user requests may not correspond to valid tool sequences, compatible parameters, or available data. To address this limitation, we propose SyntheticAgentTraceQA, an execution- first framework for generating scalable supervision data for tool- augmented agents. Our framework first constructs high-level work- flow structures, maps them to available tools through dependency- aware assignment, executes and validates the resulting traces in con- trolled environments, and only then synthesizes natural-language user tasks, teacher-generated reasoning annotations, and reference answers. We evaluate the framework across four tool ecosystems and use the resulting data to fine-tune and evaluate Qwen model variants. The results show that execution-grounded supervision improves tool execution behavior, reference-trace agreement, and answer-generation performance on the evaluated tasks. Further analysis reveals a supervision trade-off: masked supervision, which excludes reasoning annotations from the training objective, im- proves final-answer metrics, whereas full supervision, computing loss over the complete assistant output including reasoning tokens, underperforms on answer quality and does not consistently im- prove reference-trace agreement, particularly at the 9B scale. These findings highlight the importance of designing synthetic supervi- sion according to the desired capabilities of tool-augmented agents.
Hafsa Ouajdi, Francesco Giannuzzo, Alaa Boukhary et al.· 0 citations
Converting process flow diagrams into complete simulation models remains a persistent bottleneck in process systems engineering (PSE), requiring significant manual effort and simulator-specific expertise. Although advances in diagram interpretation and automated model generation have been made, these tasks are typically addressed in isolation, limiting the automation of end-to-end workflows. This work introduces Sketch2Simulation, a unified computational framework that automates flowsheet generation directly from raw engineering diagrams using a multi-agent large language model (LLM) architecture. The proposed framework integrates three coordinated layers: (i) Diagram Parsing and Interpretation, (ii) Simulation Model Synthesis, and (iii) Multi-level Validation. In the first layer, multimodal LLM agents extract process semantics, identify unit operations and stream connectivity, and resolve implicit structural features. This information is encoded into a directed graph-based intermediate representation that captures process topology while enforcing simulator-compatible constraints. This intermediate representation serves as a formal interface between diagram interpretation and simulator execution, enabling consistent translation of unstructured visual inputs into simulator-compatible models. The second layer translates this representation into a simulation model through sequential agents responsible for thermodynamic specification, object instantiation, and operating condition assignment, culminating in simulation execution within Aspen HYSYS. The use of a multi-agent architecture enables decomposition of the workflow into specialised reasoning tasks spanning multimodal interpretation, structured model synthesis, and simulator interaction, improving scalability, interpretability, and robustness compared to monolithic LLM approaches. The final layer introduces validation at multiple stages, including schema enforcement and an execution-and-correction loop that iteratively resolves runtime errors to ensure model validity. The framework is evaluated across four case studies of increasing complexity, including industrial-scale flowsheets with recycle loops. Results demonstrate consistent generation of simulation models with high structural fidelity, achieving near-complete recovery of process topology (e.g., connection consistency ≥ 0.93, stream consistency ≥ 0.96). Performance degradation is primarily associated with increased diagram complexity and dense interconnections. This work demonstrates that diagram-to-simulation transformation can be formulated as a unified computational problem, reducing reliance on manual model construction and advancing the digitalisation of PSE workflows. Crucially, this enables faster iteration between conceptual design and simulation, lowering the barrier to deploying high-fidelity models in both research and industrial settings.
Emma Pajak· Proceedings of the 3rd Found...· 0 citations
This work presents AgentTrails, a prototype system for agent provenance and sensemaking that converts raw trajectories into structured provenance graphs, where tool calls are modeled as computational actions and inputs and outputs as data artifacts and supports pattern extraction, downstream analysis, and skill abstraction.
Eden Wu, Sonia Castelo, Yurong Liu et al.· 1 citation
Addressing a structural-engineering request requires more than a single answer; it requires a chain of interdependent artifacts: interpreted requirements, a computable model, validation records, solver outputs, applicable engineering checks, and a final report. Evaluations centered on question answering or script generation may therefore reward fluent outputs even when the underlying workflow is incomplete, inconsistent, or non-executable. We present StructureClaw, an artifact-centered workbench in which LLM agents operate through governed engineering skills, typed tools, shared artifact state, and local analysis backends, together with StructureClaw-Bench, an executable benchmark of 150 controlled scenarios spanning standard workflows, interactive robustness, and multimodal structural-model reconstruction. Its analyzable standard and multimodal cases require both strict one-to-one structural-model matching and numerical-response agreement with frozen reference responses from the selected analysis engine; interactive cases instead require positive clarification or recovery evidence together with safe non-execution when appropriate. A trial succeeds only when every fixture-required assertion passes. Across nine text-agent configurations, generic-only execution passed the model-artifact check in 87.0% of retained outcomes but achieved only 22.0% E2E Success, whereas automatic StructureClaw reached 82.9%. Interactive and multimodal evaluations further identify semantic state consistency and executable model reconstruction as the dominant remaining bottlenecks. The code and benchmark are available at https://github.com/structureclaw/structureclaw.