[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-81709-en":3,"doc-seo-81709-105":30,"detail-sidebar-cat-0-en-105":91},{"code":4,"msg":5,"data":6},0,"success",{"doc_id":7,"user_id":8,"nickname":9,"user_avatar":10,"doc_module":4,"category_id":11,"category_name":12,"doc_title":13,"doc_description":14,"doc_content":15,"file_id":16,"file_url":17,"file_type":18,"file_size":19,"view_count":20,"is_deleted":4,"is_public":21,"is_downloadable":21,"audit_status":21,"page_count":22,"language":23,"language_code":24,"site_id":25,"html_lang":24,"table_of_contents":26,"faqs":27,"seo_title":13,"seo_description":14,"update_tm":28,"read_time":29},81709,3848291630094,"Emma Wilson","https://eur-avatar.wpscdn.com/davatar_085a072bc5b1113ac321206ff7593b45",8,"Research & Report","SkillSelect-Serve: Budget-Controllable and QoS-Aware Skill Service Recommendation and Composition for Small LLM Agents","Reusable skill libraries are emerging as key infrastructure for Large Language Model (LLM) agents, but existing skill selection often returns fixed Top-k lists and overlooks budget, tool constraints, and risk preferences. SkillSelect-Serve formulates skill selection as Skill Service Recommendation and Composition. Raw skills are modeled as structured Skill Services covering functionality, I/O assumptions, tool dependencies, cost, risk, and QoS attributes, then service bundles are calibrated under budgeted QoS-aware projection for small agents. Experiments on 35,353 skills and 586 queries show improved recall and utility over Top-k retrieval, approaching a Top-20 upper bound, and reveal an execution-aware learning gap.","SkillSelect-Serve: Budget-Controllable and QoS-Aware Skill Service Recommendation and Composition for Small LLM Agents  \nJingyuan Zheng, Dongjing Wang, Xin Zhang, Butian Huang, Haiping Zhang, Dongjin Yu, and Shuguang Deng  \narXiv :2607 .000 1 1v 1 [ cs .IR] 8 May 2026  \nAbstract—Reusable skill libraries are emerging as an important infrastructure for Large Language Model (LLM) agents. However, existing skill selection methods often treat skills as retrievable documents and return fixed Top-k lists, which is inadequate for small LLM agents with limited context budgets, tool constraints, and risk preferences. This paper presents SkillSelectServe, a budget-controllable and QoS-aware framework that formulates agent skill selection as Skill Service Recommendation and Composition. SkillSelect-Serve profiles raw skills as structured Skill Services with functionality, I/O assumptions, tool dependencies, cost, risk, and QoS-related attributes, and uses a local Micro-Agent Requirement Planner to parse natural language tasks into structured service requirements without directly selecting skill IDs. Based on a shared discovery backbone, SkillSelect-Serve performs dual-granularity service utility modeling: skill-level marginal suitability estimation identifies useful candidate services, while bundle-level calibration estimates coverage, redundancy, cost, and risk trade-offs. A Budgeted QoSaware Projection then outputs service bundles under different service budget regimes using only deployment-observable, labelfree features at inference time. We evaluate SkillSelect-Serve on a registry of 35,353 skill items and 586 task queries. Under the same budget of three services, SkillSelect-Serve improves bundle recall from 0.8163 to 0.8700 and mean utility from 0.6333 to 0.6901 over Top-3 retrieval. Under the budget of five services, it improves bundle recall from 0.8492 to 0.8873 and mean utility from 0.6672 to 0.7078 over Top-5 retrieval, matching the Top-20 candidate-space upper bound. Diagnostic execution results further reveal a recommendation–execution gap, suggesting the need for execution-aware service utility learning. These findings show that reusable agent skills should be managed as budgeted, QoSaware Skill Services rather than plain retrievable documents.  \nIndex Terms—LLM Agents, Skill-as-a-Service, Service Computing, Skill Service Recommendation, QoS-aware Service Composition, Budgeted Recommendation, Service Discovery, Small Language Models.  \nI. INTRODUCTION  \nLARGE Language Model (LLM) agents are evolving from  \nsingle-prompt interactions, limited tool calls, and shorthorizon exchanges into complex intelligent systems capable of task decomposition, tool invocation, model orchestration,  \nJingyuan Zheng, Dongjing Wang, Xin Zhang, Haiping Zhang, and Dongjin Yu are with Hangzhou Dianzi University, Hangzhou, China. E-mail: [zhengjoy@hdu.edu.cn](zhengjoy@hdu.edu.cn), [dongjing.wang@hdu.edu.cn](dongjing.wang@hdu.edu.cn), [zhangxin@hdu.edu.cn](zhangxin@hdu.edu.cn),  \n[zhanghp@hdu.edu.cn](zhanghp@hdu.edu.cn), [yudj@hdu.edu.cn](yudj@hdu.edu.cn).  \nButian Huang is with the School of Cyberspace Security, Hangzhou Dianzi University, Hangzhou, China. E-mail: [butian@hdu.edu.cn](butian@hdu.edu.cn).  \nShuguang Deng is with Zhejiang University, Hangzhou, China. E-mail: [dengsg@zju.edu.cn](dengsg@zju.edu.cn).  \nCorresponding author: Dongjing Wang.  \nand interaction with external APIs [1], [2] . In this emerging paradigm, agent skills serve as a critical intermediate layer between user tasks and executable capabilities. A skill typically contains not only natural-language instructions, but also task workflows, tool-use specifications, code templates, inputoutput assumptions, execution constraints, examples, and risk warnings. As such skills are continuously accumulated, reused, and shared, large-scale agent skill libraries are giving rise to a new service-oriented ecosystem. Within this ecosystem, reusable skills function analogously to service uni","cbCaieiUsIemaOij","https://ap.wps.com/l/cbCaieiUsIemaOij","pdf",14416965,2,1,22,"English","en",105,"# Abstract\n# Introduction\n## Problem background and motivation\n## Limitations of existing retrieval-style methods","[{\"question\":\"What problem does SkillSelect-Serve address for small LLM agents?\",\"answer\":\"It addresses agent skill selection as a budgeted decision, where small context budgets and tool/risk constraints make simple Top-k retrieval insufficient for achieving effective task execution.\"},{\"question\":\"How does SkillSelect-Serve represent skills internally?\",\"answer\":\"It profiles raw skills as structured Skill Services, including functionality, I/O assumptions, tool dependencies, cost, risk, and QoS-related attributes.\"},{\"question\":\"What improvement does SkillSelect-Serve achieve compared with Top-k retrieval?\",\"answer\":\"With the same budget, it improves bundle recall and mean utility over Top-3 or Top-5 retrieval, reaching performance close to the Top-20 candidate-space upper bound.\"}]",1784175547,55,{"code":4,"msg":31,"data":32},"ok",{"site_id":25,"language":24,"slug":33,"title":13,"keywords":34,"description":14,"schema_data":35,"social_meta":86,"head_meta":88,"extra_data":90,"updated_unix":28},"skillselect-serve-budget-controllable-and-qos-aware-skill-service-recommendation-and-composition-for-small-llm-agents","",{"@graph":36,"@context":85},[37,53,68],{"@type":38,"itemListElement":39},"BreadcrumbList",[40,44,47,50],{"item":41,"name":42,"@type":43,"position":21},"https://docshare.wps.com","Home","ListItem",{"item":45,"name":46,"@type":43,"position":20},"https://docshare.wps.com/document/","Document",{"item":48,"name":12,"@type":43,"position":49},"https://docshare.wps.com/document/research-report/",3,{"item":51,"name":13,"@type":43,"position":52},"https://docshare.wps.com/document/skillselect-serve-budget-controllable-and-qos-aware-skill-service-recommendation-and-composition-for-small-llm-agents/81709/",4,{"url":51,"name":13,"@type":54,"author":55,"headline":13,"publisher":57,"fileFormat":60,"inLanguage":24,"description":14,"dateModified":61,"datePublished":62,"encodingFormat":60,"isAccessibleForFree":63,"interactionStatistic":64},"DigitalDocument",{"name":9,"@type":56},"Person",{"url":41,"name":58,"@type":59},"DocShare","Organization","application/pdf","2026-07-22","2026-07-16",true,{"@type":65,"interactionType":66,"userInteractionCount":20},"InteractionCounter",{"@type":67},"ViewAction",{"@type":69,"mainEntity":70},"FAQPage",[71,77,81],{"name":72,"@type":73,"acceptedAnswer":74},"What problem does SkillSelect-Serve address for small LLM agents?","Question",{"text":75,"@type":76},"It addresses agent skill selection as a budgeted decision, where small context budgets and tool/risk constraints make simple Top-k retrieval insufficient for achieving effective task execution.","Answer",{"name":78,"@type":73,"acceptedAnswer":79},"How does SkillSelect-Serve represent skills internally?",{"text":80,"@type":76},"It profiles raw skills as structured Skill Services, including functionality, I/O assumptions, tool dependencies, cost, risk, and QoS-related attributes.",{"name":82,"@type":73,"acceptedAnswer":83},"What improvement does SkillSelect-Serve achieve compared with Top-k retrieval?",{"text":84,"@type":76},"With the same budget, it improves bundle recall and mean utility over Top-3 or Top-5 retrieval, reaching performance close to the Top-20 candidate-space upper bound.","https://schema.org",{"og:url":51,"og:type":87,"og:title":13,"og:site_name":58,"og:description":14},"article",{"robots":89,"canonical":51},"index,follow",{"doc_id":7,"site_id":25},{"code":4,"msg":5,"data":92},[93,97,101,105,110,115,120,123,128,131,135],{"id":21,"doc_module":4,"doc_module_name":46,"category_name":94,"show_sort_weight":95,"slug":96},"Story & Novel",90,"story-novel",{"id":20,"doc_module":4,"doc_module_name":46,"category_name":98,"show_sort_weight":99,"slug":100},"Literature",80,"literature",{"id":52,"doc_module":4,"doc_module_name":46,"category_name":102,"show_sort_weight":103,"slug":104},"Exam",70,"exam",{"id":106,"doc_module":4,"doc_module_name":46,"category_name":107,"show_sort_weight":108,"slug":109},5,"Comic",60,"comic",{"id":111,"doc_module":4,"doc_module_name":46,"category_name":112,"show_sort_weight":113,"slug":114},6,"Technology",50,"technology",{"id":116,"doc_module":4,"doc_module_name":46,"category_name":117,"show_sort_weight":118,"slug":119},7,"Healthcare",40,"healthcare",{"id":11,"doc_module":4,"doc_module_name":46,"category_name":12,"show_sort_weight":121,"slug":122},30,"research-report",{"id":124,"doc_module":4,"doc_module_name":46,"category_name":125,"show_sort_weight":126,"slug":127},9,"Religion & Spirituality",20,"religion-spirituality",{"id":126,"doc_module":4,"doc_module_name":46,"category_name":129,"show_sort_weight":126,"slug":130},"World Cup","world-cup",{"id":132,"doc_module":4,"doc_module_name":46,"category_name":133,"show_sort_weight":132,"slug":134},10,"Lifestyle","lifestyle",{"id":136,"doc_module":4,"doc_module_name":46,"category_name":137,"show_sort_weight":106,"slug":138},19,"General","general"]