Language‑model agents increasingly rely on skills—reusable procedural knowledge—for reasoning, tool use, and interaction. Prior work studies skill acquisition, retrieval, compression, or composition, yet 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.
To address this, we introduce SkillAlign, a provider‑agnostic framework. It represents candidate skills as multi‑view procedural cards and renders them through alternative exposure interfaces, such as full instructions, hints, compressed summaries, workflows, or no exposure. This enables counterfactual evaluation where the task, agent, and candidate skills remain constant while only the exposure interface varies.
Experiments on ALFWorld and SkillsBench show that the form of exposure substantially impacts task success and rendered context cost; a compact top‑k exposure can even outperform full‑library injection. A replay‑based policy‑learning analysis on ALFWorld reveals that adaptive exposure contains learnable signal but remains far from oracle selection.
Our findings suggest that skill‑augmented agents should optimize not only which skills to use but also how those skills are presented.
Review