1
微软优化 Edge 浏览器：网页提示不乱跑、扩展 AI 智能体技能
IT之家 9 月 29 日消息，微软于 9 月 21 日发布更新日志，面向 Microsoft Edge 浏览器开发者，引入一系列新功能 / 新工具。相关改进落地后， 可以让网页应用更流畅，AI 智能体可以驾驭更多浏览器任务。 这些新功能 / 新工具虽然是“开发者向”，不过开发者通过调用这些新特性，…
2
SkillMOO: Multi-Objective Optimization of Agent Skills for Software Engineering
面向软件工程智能体的技能多目标优化新方法，值得一读
arXiv:2604.09297v3 Announce Type: replace-cross Abstract: Agent skills are increasingly used to configure coding agents for software engineering (SE) …
3
Detecting Malicious Agent Skills in the Wild using Attention
用注意力机制检测野外恶意Agent技能，为AI安全提供新思路。
arXiv:2606.23416v1 Announce Type: cross Abstract: LLM agents increasingly load skills, file-based packages of natural-language instructions written by…
4
Trace2Skill: Distill Trajectory-Local Lessons into Transferable Agent Skills
从智能体轨迹中提炼可迁移技能，提出一种基于局部经验的知识蒸馏新方法，助力通用智能体学习。
arXiv:2603.25158v5 Announce Type: replace Abstract: Large Language Model (LLM) agents increasingly rely on domain-specific skills, yet manually author…
5
Skill-RM: Unifying Heterogeneous Evaluation Criteria via Agent Skill
这篇论文提出Skill-RM框架，通过智能体技能统一异构评估标准，推动AI评估体系革新。
arXiv:2606.03980v1 Announce Type: cross Abstract: Reward models (RMs) provide critical feedback signals for LLM post-training, notably in reinforced f…
6
MOCHA: Multi-Objective Chebyshev Annealing for Agent Skill Optimization
多目标优化新方案：Chebyshev退火算法让智能体技能训练更高效，兼顾精度与泛化能力。
arXiv:2605.19330v1 Announce Type: cross Abstract: LLM agents organize behavior through skills - structured natural-language specifications governing h…