{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:27AKX2E2OHTD3AVNAAZVZW2DE3","short_pith_number":"pith:27AKX2E2","schema_version":"1.0","canonical_sha256":"d7c0abe89a71e63d82ad00335cdb4326fec5ac724461a9ec6c2882233148b40f","source":{"kind":"arxiv","id":"2602.20867","version":1},"attestation_state":"computed","paper":{"title":"SoK: Agentic Skills -- Beyond Tool Use in LLM Agents","license":"http://creativecommons.org/licenses/by/4.0/","headline":"Agentic skills function as reusable procedural modules that let LLM agents handle long-horizon tasks reliably across domains.","cross_cats":["cs.AI","cs.CE","cs.ET"],"primary_cat":"cs.CR","authors_text":"Baihe Ma, Delong Li, Guangsheng Yu, Haiyu Deng, Qin Wang, Xu Wang, Yanna Jiang","submitted_at":"2026-02-24T13:11:38Z","abstract_excerpt":"Agentic systems increasingly rely on reusable procedural capabilities, \\textit{a.k.a., agentic skills}, to execute long-horizon workflows reliably. These capabilities are callable modules that package procedural knowledge with explicit applicability conditions, execution policies, termination criteria, and reusable interfaces. Unlike one-off plans or atomic tool calls, skills operate (and often do well) across tasks.\n  This paper maps the skill layer across the full lifecycle (discovery, practice, distillation, storage, composition, evaluation, and update) and introduces two complementary taxo"},"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":true},"canonical_record":{"source":{"id":"2602.20867","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CR","submitted_at":"2026-02-24T13:11:38Z","cross_cats_sorted":["cs.AI","cs.CE","cs.ET"],"title_canon_sha256":"8586e819e4dabd6f63deafc404829d9d39f16b79aae0d6894ffc8a81e13ce8f1","abstract_canon_sha256":"b2538c823619093b7d52d31cf22ebe645647f5d5a86d6344fe246b28a40d5b7e"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-17T23:39:19.876981Z","signature_b64":"d/pDauVU7DSC/ILbOoS0GqQZl7GLm9DsWlA/Vks+dYYMtE+QzMYbF6uwKTbZKs19uMkIZ2ggXeNrp7fdl4P7Bg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"d7c0abe89a71e63d82ad00335cdb4326fec5ac724461a9ec6c2882233148b40f","last_reissued_at":"2026-05-17T23:39:19.876413Z","signature_status":"signed_v1","first_computed_at":"2026-05-17T23:39:19.876413Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"SoK: Agentic Skills -- Beyond Tool Use in LLM Agents","license":"http://creativecommons.org/licenses/by/4.0/","headline":"Agentic skills function as reusable procedural modules that let LLM agents handle long-horizon tasks reliably across domains.","cross_cats":["cs.AI","cs.CE","cs.ET"],"primary_cat":"cs.CR","authors_text":"Baihe Ma, Delong Li, Guangsheng Yu, Haiyu Deng, Qin Wang, Xu Wang, Yanna Jiang","submitted_at":"2026-02-24T13:11:38Z","abstract_excerpt":"Agentic systems increasingly rely on reusable procedural capabilities, \\textit{a.k.a., agentic skills}, to execute long-horizon workflows reliably. These capabilities are callable modules that package procedural knowledge with explicit applicability conditions, execution policies, termination criteria, and reusable interfaces. Unlike one-off plans or atomic tool calls, skills operate (and often do well) across tasks.\n  This paper maps the skill layer across the full lifecycle (discovery, practice, distillation, storage, composition, evaluation, and update) and introduces two complementary taxo"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"Agentic skills are reusable procedural modules with explicit applicability conditions, execution policies, termination criteria, and interfaces that operate reliably across tasks; their adoption introduces supply-chain and prompt-injection risks, as shown by the ClawHavoc campaign in which nearly 1,200 malicious skills infiltrated a major marketplace.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That the two proposed taxonomies (seven design patterns and representation-by-scope) provide a sufficiently complete and stable organizing framework for the rapidly evolving space of agentic skills, and that the security implications drawn from the single ClawHavoc case study generalize to other agent platforms.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"The paper systematizes agentic skills beyond tool use, providing design pattern and representation-scope taxonomies plus security analysis of malicious skill infiltration in agent marketplaces.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"Agentic skills function as reusable procedural modules that let LLM agents handle long-horizon tasks reliably across domains.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"643a05f8cb838d312fb08ffeca6faade81891ef6cc85615ffa1197fd805db0b9"},"source":{"id":"2602.20867","kind":"arxiv","version":1},"verdict":{"id":"0fdfa029-e6cf-41cf-a5cf-9662e7d0c728","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-14T23:14:01.776364Z","strongest_claim":"Agentic skills are reusable procedural modules with explicit applicability conditions, execution policies, termination criteria, and interfaces that operate reliably across tasks; their adoption introduces supply-chain and prompt-injection risks, as shown by the ClawHavoc campaign in which nearly 1,200 malicious skills infiltrated a major marketplace.","one_line_summary":"The paper systematizes agentic skills beyond tool use, providing design pattern and representation-scope taxonomies plus security analysis of malicious skill infiltration in agent marketplaces.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That the two proposed taxonomies (seven design patterns and representation-by-scope) provide a sufficiently complete and stable organizing framework for the rapidly evolving space of agentic skills, and that the security implications drawn from the single ClawHavoc case study generalize to other agent platforms.","pith_extraction_headline":"Agentic skills function as reusable procedural modules that let LLM agents handle long-horizon tasks reliably across domains."},"references":{"count":76,"sample":[{"doi":"","year":2024,"title":"WebArena: A Realistic Web Environment for Building Autonomous Agents","work_id":"7058ffd2-a339-4102-89eb-248eeb074652","ref_index":1,"cited_arxiv_id":"2307.13854","is_internal_anchor":true},{"doi":"","year":2024,"title":"SWE-agent: Agent-Computer Interfaces Enable Automated Software Engineering","work_id":"01826cd9-a652-403c-a2ec-531da9fe2b6a","ref_index":2,"cited_arxiv_id":"2405.15793","is_internal_anchor":true},{"doi":"","year":2025,"title":"Measuring and augmenting large language models for solving capture-the-flag challenges,","work_id":"6582892a-62dd-4abd-9f10-d220464d0559","ref_index":3,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2023,"title":"HuggingGPT: Solving AI Tasks with ChatGPT and its Friends in Hugging Face","work_id":"f20ed1da-2676-4598-a11b-54549718735b","ref_index":4,"cited_arxiv_id":"2303.17580","is_internal_anchor":true},{"doi":"","year":2024,"title":"Can large language model agents simulate human trust behavior?","work_id":"7da91674-be8c-4f1b-a952-856157a43024","ref_index":5,"cited_arxiv_id":"","is_internal_anchor":false}],"resolved_work":76,"snapshot_sha256":"3e102cbc69b46ca7d2e547eee6fa622ec283903d617fcad4a60bd14312c7eab7","internal_anchors":34},"formal_canon":{"evidence_count":3,"snapshot_sha256":"5a05e6977da5bd9fdd24a577bdcb125b1087bfb4043638e7bc290f978a9f2245"},"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":"2602.20867","created_at":"2026-05-17T23:39:19.876496+00:00"},{"alias_kind":"arxiv_version","alias_value":"2602.20867v1","created_at":"2026-05-17T23:39:19.876496+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2602.20867","created_at":"2026-05-17T23:39:19.876496+00:00"},{"alias_kind":"pith_short_12","alias_value":"27AKX2E2OHTD","created_at":"2026-05-18T12:33:37.589309+00:00"},{"alias_kind":"pith_short_16","alias_value":"27AKX2E2OHTD3AVN","created_at":"2026-05-18T12:33:37.589309+00:00"},{"alias_kind":"pith_short_8","alias_value":"27AKX2E2","created_at":"2026-05-18T12:33:37.589309+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":31,"internal_anchor_count":31,"sample":[{"citing_arxiv_id":"2605.23904","citing_title":"SkillOpt: Executive Strategy for Self-Evolving Agent Skills","ref_index":8,"is_internal_anchor":true},{"citing_arxiv_id":"2605.22321","citing_title":"Benchmarking Autonomous Agents against Temporal, Spatial, and Semantic Evasions","ref_index":7,"is_internal_anchor":true},{"citing_arxiv_id":"2605.07358","citing_title":"A Comprehensive Survey on Agent Skills: Taxonomy, Techniques, and Applications","ref_index":70,"is_internal_anchor":true},{"citing_arxiv_id":"2605.14102","citing_title":"ChromaFlow: A Negative Ablation Study of Orchestration Overhead in Tool-Augmented Agent Evaluation","ref_index":10,"is_internal_anchor":true},{"citing_arxiv_id":"2605.16508","citing_title":"The Scaling Laws of Skills in LLM Agent Systems","ref_index":18,"is_internal_anchor":true},{"citing_arxiv_id":"2605.17169","citing_title":"Responsible Agentic AI Requires Explicit Provenance","ref_index":31,"is_internal_anchor":true},{"citing_arxiv_id":"2605.18401","citing_title":"SkillsVote: Lifecycle Governance of Agent Skills from Collection, Recommendation to Evolution","ref_index":19,"is_internal_anchor":true},{"citing_arxiv_id":"2605.20023","citing_title":"When Skills Don't Help: A Negative Result on Procedural Knowledge for Tool-Grounded Agents in Offensive Cybersecurity","ref_index":7,"is_internal_anchor":true},{"citing_arxiv_id":"2605.14102","citing_title":"ChromaFlow: A Negative Ablation Study of Orchestration Overhead in Tool-Augmented Agent Evaluation","ref_index":10,"is_internal_anchor":true},{"citing_arxiv_id":"2605.02900","citing_title":"Safety in Embodied AI: A Survey of Risks, Attacks, and Defenses","ref_index":162,"is_internal_anchor":true},{"citing_arxiv_id":"2604.03733","citing_title":"SoK: Blockchain Agent-to-Agent Payments","ref_index":6,"is_internal_anchor":true},{"citing_arxiv_id":"2605.10990","citing_title":"Skill Drift Is Contract Violation: Proactive Maintenance for LLM Agent Skill Libraries","ref_index":13,"is_internal_anchor":true},{"citing_arxiv_id":"2605.06130","citing_title":"Skill1: Unified Evolution of Skill-Augmented Agents via Reinforcement Learning","ref_index":32,"is_internal_anchor":true},{"citing_arxiv_id":"2605.11781","citing_title":"Five Attacks on x402 Agentic Payment Protocol","ref_index":35,"is_internal_anchor":true},{"citing_arxiv_id":"2605.12015","citing_title":"SkillSafetyBench: Evaluating Agent Safety under Skill-Facing Attack Surfaces","ref_index":61,"is_internal_anchor":true},{"citing_arxiv_id":"2605.11169","citing_title":"OLIVIA: Online Learning via Inference-time Action Adaptation for Decision Making in LLM ReAct Agents","ref_index":34,"is_internal_anchor":true},{"citing_arxiv_id":"2605.08526","citing_title":"Skill-CMIB: Multimodal Agent Skill for Consistent Action via Conditional Multimodal Information Bottleneck","ref_index":11,"is_internal_anchor":true},{"citing_arxiv_id":"2605.08887","citing_title":"Ace-Skill: Bootstrapping Multimodal Agents with Prioritized and Clustered Evolution","ref_index":24,"is_internal_anchor":true},{"citing_arxiv_id":"2604.23505","citing_title":"Uncertainty Propagation in LLM-Based Systems","ref_index":8,"is_internal_anchor":true},{"citing_arxiv_id":"2605.06130","citing_title":"Skill1: Unified Evolution of Skill-Augmented Agents via Reinforcement Learning","ref_index":32,"is_internal_anchor":true},{"citing_arxiv_id":"2605.05726","citing_title":"SkillRet: A Large-Scale Benchmark for Skill Retrieval in LLM Agents","ref_index":12,"is_internal_anchor":true},{"citing_arxiv_id":"2605.05274","citing_title":"Sealing the Audit-Runtime Gap for LLM Skills","ref_index":19,"is_internal_anchor":true},{"citing_arxiv_id":"2604.13180","citing_title":"SciFi: A Safe, Lightweight, User-Friendly, and Fully Autonomous Agentic AI Workflow for Scientific Applications","ref_index":21,"is_internal_anchor":true},{"citing_arxiv_id":"2604.08224","citing_title":"Externalization in LLM Agents: A Unified Review of Memory, Skills, Protocols and Harness Engineering","ref_index":63,"is_internal_anchor":true},{"citing_arxiv_id":"2605.06978","citing_title":"Group of Skills: Group-Structured Skill Retrieval for Agent Skill Libraries","ref_index":24,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":3,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/27AKX2E2OHTD3AVNAAZVZW2DE3","json":"https://pith.science/pith/27AKX2E2OHTD3AVNAAZVZW2DE3.json","graph_json":"https://pith.science/api/pith-number/27AKX2E2OHTD3AVNAAZVZW2DE3/graph.json","events_json":"https://pith.science/api/pith-number/27AKX2E2OHTD3AVNAAZVZW2DE3/events.json","paper":"https://pith.science/paper/27AKX2E2"},"agent_actions":{"view_html":"https://pith.science/pith/27AKX2E2OHTD3AVNAAZVZW2DE3","download_json":"https://pith.science/pith/27AKX2E2OHTD3AVNAAZVZW2DE3.json","view_paper":"https://pith.science/paper/27AKX2E2","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2602.20867&json=true","fetch_graph":"https://pith.science/api/pith-number/27AKX2E2OHTD3AVNAAZVZW2DE3/graph.json","fetch_events":"https://pith.science/api/pith-number/27AKX2E2OHTD3AVNAAZVZW2DE3/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/27AKX2E2OHTD3AVNAAZVZW2DE3/action/timestamp_anchor","attest_storage":"https://pith.science/pith/27AKX2E2OHTD3AVNAAZVZW2DE3/action/storage_attestation","attest_author":"https://pith.science/pith/27AKX2E2OHTD3AVNAAZVZW2DE3/action/author_attestation","sign_citation":"https://pith.science/pith/27AKX2E2OHTD3AVNAAZVZW2DE3/action/citation_signature","submit_replication":"https://pith.science/pith/27AKX2E2OHTD3AVNAAZVZW2DE3/action/replication_record"}},"created_at":"2026-05-17T23:39:19.876496+00:00","updated_at":"2026-05-17T23:39:19.876496+00:00"}