From Skill Text to Skill Structure: The Scheduling-Structural-Logical Representation for Agent Skills
Qiliang Liang, Hansi Wang, Zhong Liang, Yang Liu
2026-05-04
Summary
This paper focuses on how to better organize and represent the 'skills' that AI agents use to complete tasks, moving beyond simply describing them in text.
What's the problem?
Currently, AI agent skills are often described using lengthy text documents, making it difficult for the AI itself to understand *how* to actually use those skills. Important details about what a skill does, how it works, and what it affects are all mixed together in natural language, making it hard for the AI to find the right skill or predict what will happen when it's used. Essentially, the AI struggles to 'reason' about skills because the information isn't clearly structured.
What's the solution?
The researchers developed a new way to represent skills called SSL, which stands for Scheduling-Structural-Logical. This method breaks down each skill into three separate parts: how to initiate it (scheduling signals), the steps involved in carrying it out (structural execution), and the specific actions and resources it uses (logical action/resource use). They used a large language model to automatically organize existing skill descriptions into this SSL format and then tested it on tasks like finding relevant skills and assessing potential risks associated with using them.
Why it matters?
This work is important because it makes AI skills more understandable and usable for AI agents. By providing a clear, structured representation, the AI can more easily search for, evaluate, and apply skills, leading to more reliable and effective performance. It’s a step towards building AI systems where skills are reusable, inspectable, and predictable, rather than being hidden within blocks of text.
Abstract
LLM agents increasingly rely on reusable skills, capability packages that combine instructions, control flow, constraints, and tool calls. In most current agent systems, however, skills are still represented by text-heavy artifacts, including SKILL.md-style documents and structured records whose machine-usable evidence remains embedded largely in natural-language descriptions. This poses a challenge for skill-centered agent systems: managing skill collections and using skills to support agent both require reasoning over invocation interfaces, execution structure, and concrete side effects that are often entangled in a single textual surface. An explicit representation of skill knowledge may therefore help make these artifacts easier for machines to acquire and leverage. Drawing on Memory Organization Packets, Script Theory, and Conceptual Dependency from Schank and Abelson's classical work on linguistic knowledge representation, we introduce what is, to our knowledge, the first structured representation for agent skill artifacts that disentangles skill-level scheduling signals, scene-level execution structure, and logic-level action and resource-use evidence: the Scheduling-Structural-Logical (SSL) representation. We instantiate SSL with an LLM-based normalizer and evaluate it on a corpus of skills in two tasks, Skill Discovery and Risk Assessment, and superiorly outperform the text-only baselines: in Skill Discovery, SSL improves MRR from 0.573 to 0.707; in Risk Assessment, it improves macro F1 from 0.744 to 0.787. These findings reveal that explicit, source-grounded structure makes agent skills easier to search and review. They also suggest that SSL is best understood as a practical step toward more inspectable, reusable, and operationally actionable skill representations for agent systems, rather than as a finished standard or an end-to-end mechanism for managing and using skills.