update.

rl-introduction.bbl

-\begin{thebibliography}{142}
+\begin{thebibliography}{143}
 \providecommand{\natexlab}[1]{#1}
 \providecommand{\url}[1]{\texttt{#1}}
 \expandafter\ifx\csname urlstyle\endcsname\relax
@@ -697,6 +697,11 @@ Da~Yin, Faeze Brahman, Abhilasha Ravichander, Khyathi Chandu, Kai-Wei Chang, Yej
 \newblock Agent lumos: Unified and modular training for open-source language agents.
 \newblock In \emph{Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)}, pp.\  12380--12403, 2024.
 
+\bibitem[Yu et~al.(2026)Yu, Zhu, Lin, Cui, Ding, and Li]{yu-etal:skill}
+Jianxiang Yu, Jiapeng Zhu, Bochen Lin, Qier Cui, Zichen Ding, and Xiang Li.
+\newblock Skill is not one-size-fits-all: Model-aware skill alignment for llm agents.
+\newblock \emph{arXiv preprint arXiv:2605.30723}, 2026.
+
 \bibitem[Yu et~al.(2024{\natexlab{a}})Yu, Yao, Zhang, He, Han, Cui, Hu, Liu, Zheng, Sun, et~al.]{yu-etal:2024rlhf}
 Tianyu Yu, Yuan Yao, Haoye Zhang, Taiwen He, Yifeng Han, Ganqu Cui, Jinyi Hu, Zhiyuan Liu, Hai-Tao Zheng, Maosong Sun, et~al.
 \newblock Rlhf-v: Towards trustworthy mllms via behavior alignment from fine-grained correctional human feedback.

rl-introduction.bib

@@ -6,6 +6,13 @@
 
 
 
+@article{yu-etal:skill,
+  title={Skill is Not One-Size-Fits-All: Model-Aware Skill Alignment for LLM Agents},
+  author={Yu, Jianxiang and Zhu, Jiapeng and Lin, Bochen and Cui, Qier and Ding, Zichen and Li, Xiang},
+  journal={arXiv preprint arXiv:2605.30723},
+  year={2026}
+}
+
 @article{wang-etal:mem,
   title={Mem-$\{$$\backslash$alpha$\}$: Learning memory construction via reinforcement learning},
   author={Wang, Yu and Takanobu, Ryuichi and Liang, Zhiqi and Mao, Yuzhen and Hu, Yuanzhe and McAuley, Julian and Wu, Xiaojian},

section6/section6.tex

@@ -429,6 +429,13 @@ Solve complex mathematical, logical, tabular, and constraint reasoning tasks. Us
 \end{tcolorbox}
 
 
+A straightforward approach to constructing skills for an agent is to manually write them according to specific task requirements. However, there can be many different ways to design a skill for the same task, similar to the process of writing prompts for LLMs. For example, one may define a skill with detailed instructions, including explicit workflows and execution procedures, while another may provide only high-level guidance and allow the agent to determine the detailed execution strategy. Such variations in skill design can significantly affect agent performance \citep{yu-etal:skill}.
+
+
+One limitation of manual skill construction is that the quality and diversity largely depend on human experience. Therefore, if we want LLMs to handle a broad range of tasks, relying on human-annotated data for LLM fine-tuning is often inefficient. To address this limitation, an alternative approach is to automatically generate skills from agentic experiences. For example, we can prompt an LLM to summarize successful trajectories and extract reusable skills. These generated skills can then be added to the agent's skill bank and reused in future tasks.
+
+The above way of generating skills often suffer from quality issues. First, a general LLM does not inherently know how to design effective skills without additional optimization. Second, the skill generator may optimize for the wrong objective because skill generation and skill utilization involve different contexts. Specifically, an LLM may consider a generated skill effective because it satisfies the instructions provided in the prompt, while another agent may fail to use this skill effectively in real-world tasks. As a result, skill generation should not only focus on producing well-structured skills, but also consider whether these skills can actually improve agent performance during task execution. Recent work resorts to RL to achieve this goal, where the skill generation is optimized based on the actual performance of generated skills.
+