SkillAlign improves task success by changing how skills are shown
SkillAlign: Aligning Skill Interfaces for LLM-based Agents
Artificial Intelligence
Summary
Large language model agents use skills to help complete tasks, but how these skills are shown to the agent can make a big difference. The authors created SkillAlign, which tries different ways of presenting the same skills, like full instructions or short hints. They found that changing the way skills are given to the agent affects success rates and the cost of using the agent. This means it’s not just important which skills are used, but also how they are presented.
What this means in practice
- •For ai system developers: Design AI agents that adapt how they present their skills to improve task completion rates and reduce resource use.
- •For software tool integrators: Optimize interfaces for external tools used by language model agents by experimenting with different exposure methods for improved effectiveness.
Authors
Shuo Ren, Xiaomian Kang, Jiajun Zhang
Abstract
Language-model agents increasingly rely on skills: reusable procedural knowledge for reasoning, tool use, and interaction. Existing work studies how skills are acquired, retrieved, compressed, or composed, but often assumes that once a skill is selected, its interface to the agent is fixed. We argue that this overlooks a key source of skill utility: the same skill can help, distract, or mislead depending on how it is exposed. We propose SkillAlign, a provider-agnostic framework that represents candidate skills as multi-view procedural cards and renders them through alternative exposure interfaces, including full instructions, hints, compressed summaries, workflows, or no exposure. This enables counterfactual evaluation where the task, agent, and candidate skills are fixed while only the exposure interface varies. Across ALFWorld and SkillsBench, we show that exposure form substantially affects task success and rendered context cost, and that compact top-k exposure can outperform full-library injection. We further conduct a replay-based policy-learning analysis on ALFWorld, showing that adaptive exposure contains learnable signal but remains far from oracle selection. Our results suggest that skill-augmented agents should optimize not only which skills to use, but also how those skills are presented.