AgentSkiller: Scaling Generalist Agent Intelligence through Semantically Integrated Cross-Domain Data Synthesis

2026-02-10

AI Summary

The development of general-purpose agents is hindered by the scarcity of high-quality, long-horizon, cross-domain interactive data. To address this challenge, this work proposes AgentSkiller, a novel framework that introduces a directed acyclic graph (DAG)-based state transition architecture and a cross-domain service integration mechanism. By leveraging domain ontologies, persona-centered entity graphs, service blueprints to define MCP tool interfaces, and persona-driven simulators, AgentSkiller automatically generates multi-turn, verifiable, and state-explicit high-fidelity interaction data. The approach ensures data recoverability and diversity, producing approximately 11K high-quality synthetic samples. Empirical evaluation demonstrates significant performance gains on function-calling tasks, with particularly pronounced improvements observed in large-scale language models.

Technology Category

• Cognitive Modeling & Cognitive Systems: Agent Architectures
• Multiagent Systems: Modeling other Agents
• Humans and AI: Human-Aware Planning and Behavior Prediction

Application Category

• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• Search and Retrieval-Augmented AI: Agentic search
• Economics, Online Markets and Human Computation: Trust and reliance of crowd workers and data experts on GenAI

Abstract

Large Language Model agents demonstrate potential in solving real-world problems via tools, yet generalist intelligence is bottlenecked by scarce high-quality, long-horizon data. Existing methods collect privacy-constrained API logs or generate scripted interactions lacking diversity, which struggle to produce data requisite for scaling capabilities. We propose AgentSkiller, a fully automated framework synthesizing multi-turn interaction data across realistic, semantically linked domains. It employs a DAG-based architecture with explicit state transitions to ensure determinism and recoverability. The pipeline builds a domain ontology and Person-Centric Entity Graph, defines tool interfaces via Service Blueprints for Model Context Protocol servers, and populates environments with consistent databases and strict Domain Policies. A cross-domain fusion mechanism links services to simulate complex tasks. Finally, the pipeline creates user tasks by verifying solution paths, filtering via execution-based validation, and generating queries using a Persona-based Simulator for automated rollout. This produces reliable environments with clear state changes. To demonstrate effectiveness, we synthesized $\approx$ 11K interaction samples; experimental results indicate that models trained on this dataset achieve significant improvements on function calling over baselines, particularly in larger parameter regimes.

Problem

• generalist agent
• data scarcity
• long-horizon interaction
• cross-domain synthesis
• high-quality training data

Innovation

• AgentSkiller
• cross-domain data synthesis
• semantic integration
• DAG-based architecture
• Persona-based Simulator