{"paper":{"title":"SkillSmith: Compiling Agent Skills into Boundary-Guided Runtime Interfaces","license":"http://creativecommons.org/licenses/by/4.0/","headline":"SkillSmith compiles agent skills offline into minimal boundary-guided interfaces to cut redundant context and reasoning in LLM systems.","cross_cats":["cs.SE"],"primary_cat":"cs.AI","authors_text":"Bangzheng Pu, Dong Dong, Duling Xu, Jialin Li, Jiawei Guan, Zaifeng Pan, Zheng Chen","submitted_at":"2026-05-12T09:25:25Z","abstract_excerpt":"Recently, skills have been widely adopted in large language model (LLM)-based agent systems across various domains. In existing frameworks, skills are typically injected into the agent reasoning loop as contextual guidance once matched to a runtime task, enabling specialized task-solving capabilities. We find that this execution paradigm introduces two major sources of redundancy: irrelevant context injection and repeated skill-specific reasoning and planning. To this end, we propose SkillSmith, a boundary-first compiler-runtime framework that compiles skill packages offline into minimal execu"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"On SkillsBench, SkillSmith reduces solve-stage token usage by 57.44%, thinking iterations by 42.99%, solve time by 50.57% (2.02x faster), and token-proportional monetary cost by 57.44% compared with using raw-skills; compiled artifacts from a stronger model can be reused by a smaller runtime model to improve accuracy where raw skill interpretation fails.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"The extraction of fine-grained operational boundaries from skill descriptions is both feasible and lossless enough that the resulting minimal interfaces preserve all task-relevant behavior without requiring the agent to fall back to full skill text.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"SkillSmith is a boundary-first compiler-runtime system that turns skill packages into minimal executable interfaces, cutting token usage 57%, thinking iterations 43%, and solve time 51% versus raw skill injection on SkillsBench.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"SkillSmith compiles agent skills offline into minimal boundary-guided interfaces to cut redundant context and reasoning in LLM systems.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"575824838ed80ff472abaf7d076922e3a5f7b8b66b6c2cca008ce8ff205b6778"},"source":{"id":"2605.15215","kind":"arxiv","version":1},"verdict":{"id":"2079733b-b3aa-410d-a593-d1065c919116","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-19T17:50:08.857771Z","strongest_claim":"On SkillsBench, SkillSmith reduces solve-stage token usage by 57.44%, thinking iterations by 42.99%, solve time by 50.57% (2.02x faster), and token-proportional monetary cost by 57.44% compared with using raw-skills; compiled artifacts from a stronger model can be reused by a smaller runtime model to improve accuracy where raw skill interpretation fails.","one_line_summary":"SkillSmith is a boundary-first compiler-runtime system that turns skill packages into minimal executable interfaces, cutting token usage 57%, thinking iterations 43%, and solve time 51% versus raw skill injection on SkillsBench.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"The extraction of fine-grained operational boundaries from skill descriptions is both feasible and lossless enough that the resulting minimal interfaces preserve all task-relevant behavior without requiring the agent to fall back to full skill text.","pith_extraction_headline":"SkillSmith compiles agent skills offline into minimal boundary-guided interfaces to cut redundant context and reasoning in LLM systems."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2605.15215/integrity.json","findings":[],"available":true,"detectors_run":[{"name":"doi_compliance","ran_at":"2026-05-19T18:01:24.804510Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"doi_title_agreement","ran_at":"2026-05-19T18:01:18.640644Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"ai_meta_artifact","ran_at":"2026-05-19T13:33:22.839651Z","status":"skipped","version":"1.0.0","findings_count":0}],"snapshot_sha256":"93c4973eeed9fbaba3e014d7667dc3cca130ec10da48c80aff2f93d12b95bdd8"},"references":{"count":30,"sample":[{"doi":"","year":2022,"title":"Do As I Can, Not As I Say: Grounding Language in Robotic Affordances","work_id":"037320f1-b0a9-4cbe-a639-bfb25409ce71","ref_index":1,"cited_arxiv_id":"2204.01691","is_internal_anchor":true},{"doi":"","year":null,"title":"Equipping agents for the real world with agent skills","work_id":"15cf6aa2-3f42-4207-967f-8d42e48599ae","ref_index":2,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2026,"title":"Agent skills","work_id":"22265264-de03-49d5-945e-0a4ae54442d2","ref_index":3,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2026,"title":"Introducing Claude Opus 4.7","work_id":"0e72af39-e73a-4a5b-95dd-11107a91b2db","ref_index":4,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":1946,"title":"Prompting is programming: A query language for large language models.Proceedings of the ACM on Programming Languages, 7(PLDI):1946–1969, 2023","work_id":"2cb96ab5-06a8-4123-a061-292a5607e8cd","ref_index":5,"cited_arxiv_id":"","is_internal_anchor":false}],"resolved_work":30,"snapshot_sha256":"37e78531853d0ccb8f7f3700f81fe4af35a188830765066857aa83d5e18ea29e","internal_anchors":7},"formal_canon":{"evidence_count":2,"snapshot_sha256":"ba5b942f6840c553204c1cee4a17301b03ce675c022b3dfeb4d40d63e2089cbe"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"}