{"about":{"site":"https://codewithpapers.app","non_affiliation":"Code with Papers and Syntology are not affiliated with, endorsed by, or sponsored by Papers with Code, Meta, or the pwc-archive mirror.","licence":"CC BY-SA 4.0","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","attribution":"https://codewithpapers.app/attribution","modified":"archive material modified by Syntology; see the attribution page"},"url":"/paper/arxiv-2604-04804","title":"SkillX: Automatically Constructing Skill Knowledge Bases for Agents","arxiv_id":"2604.04804","date":"2026-04-06","proceeding":null,"authors":["Chenxi Wang","Zhuoyun Yu","Xin Xie","Wuguannan Yao","Runnan Fang","Shuofei Qiao","Kexin Cao","Guozhou Zheng","Xiang Qi","Peng Zhang","Shumin Deng"],"abstract":"Learning from experience is critical for building capable large language model (LLM) agents, yet prevailing self-evolving paradigms remain inefficient: agents learn in isolation, repeatedly rediscover similar behaviors from limited experience, resulting in redundant exploration and poor generalization. To address this problem, we propose SkillX, a fully automated framework for constructing a \\textbf{plug-and-play skill knowledge base} that can be reused across agents and environments. SkillX operates through a fully automated pipeline built on three synergistic innovations: \\textit{(i) Multi-Level Skills Design}, which distills raw trajectories into three-tiered hierarchy of strategic plans, functional skills, and atomic skills; \\textit{(ii) Iterative Skills Refinement}, which automatically revises skills based on execution feedback to continuously improve library quality; and \\textit{(iii) Exploratory Skills Expansion}, which proactively generates and validates novel skills to expand coverage beyond seed training data. Using a strong backbone agent (GLM-4.6), we automatically build a reusable skill library and evaluate its transferability on challenging long-horizon, user-interactive benchmarks, including AppWorld, BFCL-v3, and $τ^2$-Bench. Experiments show that SkillKB consistently improves task success and execution efficiency when plugged into weaker base agents, highlighting the importance of structured, hierarchical experience representations for generalizable agent learning. Our code will be publicly available soon at https://github.com/zjunlp/SkillX.","url_abs":"https://arxiv.org/abs/2604.04804","url_pdf":"https://arxiv.org/pdf/2604.04804","source":{"archive":null,"snapshot":"2025-07-28","note":"not in the Papers with Code archive (frozen at the snapshot)","row_kind":"graph","title_abstract_authors_date":"arXiv metadata, CC0 1.0 (https://info.arxiv.org/help/license)"},"code_links":[],"tasks":[],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=2604.04804","mcp":{"get_harvested_code_for_paper":{"arxiv_id":"2604.04804"}},"developers":"https://syntology.ai/developers","read_at":"2026-09-24T18:15:14+00:00","read_at_is":"when the build read Syntology's graph, not when any sample ran","claim":"Per-sample execution status on synthesized fixtures; not a correctness claim about the paper. Samples come from repositories linked to the paper, official or community; repo_kind says which.","repos":[{"mentioned_in_github":null,"is_official":null,"provenance":"deterministic:regex_extraction","mentioned_in_paper":null,"url":"https://github.com/zjunlp/SkillX","reach":null}],"summary":{"unverified":4},"by_repo_kind":{"found_in_text":{"samples":4,"ran":0,"repositories":1}},"repo_kind_vocabulary":{"official":"The archive marks this repository official for the paper","named_in_paper":"The archive records that the paper mentions this repository; it is not marked official","listed":"In the archive's code links for this paper, not marked official and not recorded as mentioned in the paper","found_in_text":"Syntology found this repository in the paper's own text; whether it is the authors' implementation is not asserted","community":"Not in the archive's code links for this paper; a community repository Syntology harvested"},"n_pointer_only_for_licence":0,"samples":[{"code_sha256_prefix":"03907d8433ebb2b8","entry":"collect_skills_from_plan_metadata","repo":"zjunlp/SkillX","repo_kind":"found_in_text","path":"core/skill_schema.py","file_url":"https://github.com/zjunlp/SkillX/blob/HEAD/core/skill_schema.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"03907d8433ebb2b8"}},{"code_sha256_prefix":"7b7c3951bfd53ca7","entry":"group_trajectories_by_task","repo":"zjunlp/SkillX","repo_kind":"found_in_text","path":"core/trajectory.py","file_url":"https://github.com/zjunlp/SkillX/blob/HEAD/core/trajectory.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"7b7c3951bfd53ca7"}},{"code_sha256_prefix":"3a8456e390d0bbe6","entry":"normalize_skill_output","repo":"zjunlp/SkillX","repo_kind":"found_in_text","path":"core/skill_schema.py","file_url":"https://github.com/zjunlp/SkillX/blob/HEAD/core/skill_schema.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"3a8456e390d0bbe6"}},{"code_sha256_prefix":"f07c8d39f7fde6b0","entry":"skills_to_json","repo":"zjunlp/SkillX","repo_kind":"found_in_text","path":"core/skill_schema.py","file_url":"https://github.com/zjunlp/SkillX/blob/HEAD/core/skill_schema.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"f07c8d39f7fde6b0"}}]},"arxiv_metadata":{"licence":"arXiv metadata, CC0 1.0 (https://info.arxiv.org/help/license)","fields":["title","abstract","authors","date"],"primary_category":"cs.LG","source":"arxiv_2026.jsonl"},"syntology_extracted_results":null}