{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:AM5XLQQYV7ROFBPT4PRZSUIGDT","short_pith_number":"pith:AM5XLQQY","schema_version":"1.0","canonical_sha256":"033b75c218afe2e285f3e3e39951061cd8d853943ae3f86d6256db09c9d0af65","source":{"kind":"arxiv","id":"2605.13716","version":1},"attestation_state":"computed","paper":{"title":"SkillOps: Managing LLM Agent Skill Libraries as Self-Maintaining Software Ecosystems","license":"http://creativecommons.org/publicdomain/zero/1.0/","headline":"SkillOps maintains LLM skill libraries via Skill Contracts and ecosystem graphs, raising ALFWorld task success to 79.5% as a standalone agent and improving retrieval baselines by up to 2.9 points with near-zero library-time LLM cost.","cross_cats":["cs.MA"],"primary_cat":"cs.SE","authors_text":"Hongji Pu, Liang Zhao, Xinyuan Song","submitted_at":"2026-05-13T16:02:25Z","abstract_excerpt":"Large language model agents increasingly rely on skill libraries for multi-step tasks, yet these libraries can accumulate persistent defects as skills are added, reused, patched, and linked to changing dependencies. We call this failure mode skill technical debt: library-level defects that may not break a single skill locally but can harm future retrieval, composition, and execution. Existing skill-based agents mainly focus on task-time retrieval, planning, and repair, while library-time maintenance remains underexplored. We propose SkillOps, a method-agnostic plug-in framework for maintaining"},"verification_status":{"content_addressed":true,"pith_receipt":true,"author_attested":false,"weak_author_claims":0,"strong_author_claims":0,"externally_anchored":false,"storage_verified":false,"citation_signatures":0,"replication_records":0,"graph_snapshot":true,"references_resolved":true,"formal_links_present":false},"canonical_record":{"source":{"id":"2605.13716","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/publicdomain/zero/1.0/","primary_cat":"cs.SE","submitted_at":"2026-05-13T16:02:25Z","cross_cats_sorted":["cs.MA"],"title_canon_sha256":"67ae17563200187eebb2a3646aea99b7bda6549df51e411205482f583c928f22","abstract_canon_sha256":"dee088cbdfcb1c39f37c813f25f4828eecc3179bf53ade38ee5656e9bc1b8ad6"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-18T02:44:16.705104Z","signature_b64":"RWWvANxh4FZ1GniWv4tTyds1tKO5M1gDk0ZjtrVCGS5FeD9+p/ynJUK87QQt5/H0sHLYtHMdxej9av8+1YaDCg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"033b75c218afe2e285f3e3e39951061cd8d853943ae3f86d6256db09c9d0af65","last_reissued_at":"2026-05-18T02:44:16.704665Z","signature_status":"signed_v1","first_computed_at":"2026-05-18T02:44:16.704665Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"SkillOps: Managing LLM Agent Skill Libraries as Self-Maintaining Software Ecosystems","license":"http://creativecommons.org/publicdomain/zero/1.0/","headline":"SkillOps maintains LLM skill libraries via Skill Contracts and ecosystem graphs, raising ALFWorld task success to 79.5% as a standalone agent and improving retrieval baselines by up to 2.9 points with near-zero library-time LLM cost.","cross_cats":["cs.MA"],"primary_cat":"cs.SE","authors_text":"Hongji Pu, Liang Zhao, Xinyuan Song","submitted_at":"2026-05-13T16:02:25Z","abstract_excerpt":"Large language model agents increasingly rely on skill libraries for multi-step tasks, yet these libraries can accumulate persistent defects as skills are added, reused, patched, and linked to changing dependencies. We call this failure mode skill technical debt: library-level defects that may not break a single skill locally but can harm future retrieval, composition, and execution. Existing skill-based agents mainly focus on task-time retrieval, planning, and repair, while library-time maintenance remains underexplored. We propose SkillOps, a method-agnostic plug-in framework for maintaining"},"claims":{"count":3,"items":[{"kind":"strongest_claim","text":"On ALFWorld, SkillOps achieves 79.5 percent task success as a standalone agent, outperforming the strongest baseline by 8.8 percentage points with no additional task-time large language model calls.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That rule-based diagnosis across the four health dimensions (utility, compatibility, risk, validation) can reliably detect and repair library-level defects without task-specific LLM calls or human oversight.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"SkillOps maintains LLM skill libraries via Skill Contracts and ecosystem graphs, raising ALFWorld task success to 79.5% as a standalone agent and improving retrieval baselines by up to 2.9 points with near-zero library-time LLM cost.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"}],"snapshot_sha256":"da5db9419fd00949ff6d02fd2b3d1c5d1ae8f1ec6d2f533e762322941adcb9fe"},"source":{"id":"2605.13716","kind":"arxiv","version":1},"verdict":{"id":"e2e6d8fd-d41e-4068-884b-085a8b06e6ca","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-14T17:48:34.580765Z","strongest_claim":"On ALFWorld, SkillOps achieves 79.5 percent task success as a standalone agent, outperforming the strongest baseline by 8.8 percentage points with no additional task-time large language model calls.","one_line_summary":"SkillOps maintains LLM skill libraries via Skill Contracts and ecosystem graphs, raising ALFWorld task success to 79.5% as a standalone agent and improving retrieval baselines by up to 2.9 points with near-zero library-time LLM cost.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That rule-based diagnosis across the four health dimensions (utility, compatibility, risk, validation) can reliably detect and repair library-level defects without task-specific LLM calls or human oversight.","pith_extraction_headline":""},"references":{"count":52,"sample":[{"doi":"","year":null,"title":"Fundamenta Mathematicae , volume =","work_id":"b9f9b3d0-5ed0-40c6-90c3-7a7178f5ff66","ref_index":1,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":null,"title":"SkillsBench: Benchmarking How Well Agent Skills Work Across Diverse Tasks","work_id":"b477d12a-2ca6-4894-9f90-3fb479635e98","ref_index":2,"cited_arxiv_id":"2602.12670","is_internal_anchor":true},{"doi":"","year":2019,"title":"Alessandro Berti and Sebastiaan van Zelst and Wil M. P. van der Aalst , title =. 2019 , eprint =","work_id":"3558efaf-50ca-452a-9797-013780bd02be","ref_index":3,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2026,"title":"2026 , eprint =","work_id":"c48a44e6-e564-4a40-8c55-300d7ebb64ff","ref_index":4,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2025,"title":"2025 , eprint =","work_id":"4b2a1659-5333-4b9f-9d27-b923200d6690","ref_index":5,"cited_arxiv_id":"","is_internal_anchor":false}],"resolved_work":52,"snapshot_sha256":"faf21b8fdaaf407beebd18a373608d0a858f2362771af9da79a92ee3e67fa219","internal_anchors":16},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"},"aliases":[{"alias_kind":"arxiv","alias_value":"2605.13716","created_at":"2026-05-18T02:44:16.704735+00:00"},{"alias_kind":"arxiv_version","alias_value":"2605.13716v1","created_at":"2026-05-18T02:44:16.704735+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.13716","created_at":"2026-05-18T02:44:16.704735+00:00"},{"alias_kind":"pith_short_12","alias_value":"AM5XLQQYV7RO","created_at":"2026-05-18T12:33:37.589309+00:00"},{"alias_kind":"pith_short_16","alias_value":"AM5XLQQYV7ROFBPT","created_at":"2026-05-18T12:33:37.589309+00:00"},{"alias_kind":"pith_short_8","alias_value":"AM5XLQQY","created_at":"2026-05-18T12:33:37.589309+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":0,"internal_anchor_count":0,"sample":[]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/AM5XLQQYV7ROFBPT4PRZSUIGDT","json":"https://pith.science/pith/AM5XLQQYV7ROFBPT4PRZSUIGDT.json","graph_json":"https://pith.science/api/pith-number/AM5XLQQYV7ROFBPT4PRZSUIGDT/graph.json","events_json":"https://pith.science/api/pith-number/AM5XLQQYV7ROFBPT4PRZSUIGDT/events.json","paper":"https://pith.science/paper/AM5XLQQY"},"agent_actions":{"view_html":"https://pith.science/pith/AM5XLQQYV7ROFBPT4PRZSUIGDT","download_json":"https://pith.science/pith/AM5XLQQYV7ROFBPT4PRZSUIGDT.json","view_paper":"https://pith.science/paper/AM5XLQQY","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2605.13716&json=true","fetch_graph":"https://pith.science/api/pith-number/AM5XLQQYV7ROFBPT4PRZSUIGDT/graph.json","fetch_events":"https://pith.science/api/pith-number/AM5XLQQYV7ROFBPT4PRZSUIGDT/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/AM5XLQQYV7ROFBPT4PRZSUIGDT/action/timestamp_anchor","attest_storage":"https://pith.science/pith/AM5XLQQYV7ROFBPT4PRZSUIGDT/action/storage_attestation","attest_author":"https://pith.science/pith/AM5XLQQYV7ROFBPT4PRZSUIGDT/action/author_attestation","sign_citation":"https://pith.science/pith/AM5XLQQYV7ROFBPT4PRZSUIGDT/action/citation_signature","submit_replication":"https://pith.science/pith/AM5XLQQYV7ROFBPT4PRZSUIGDT/action/replication_record"}},"created_at":"2026-05-18T02:44:16.704735+00:00","updated_at":"2026-05-18T02:44:16.704735+00:00"}