{"paper":{"title":"Demystifying Agent Skills: Why They Work-Until They Don't","arxiv_id":"2608.14036","primary_cat":"cs.AI","submitted_at":"2026-08-14T07:26:38+00:00","version_compiled":"v1"},"notice":"Every quoted sentence is the authors' text from the arXiv listing abstract. Labels and probabilities are the outputs of classifier-1.13.0 on the question bank abstract-p1-20260923, exactly as returned. No sentence in this record was written by a language model.","schema":"pith.compiled-record/v0","status":"compiled","abstract":{"source":"arXiv listing abstract v1","sentences":[{"i":0,"p":1.0,"id":"s0","own":false,"role":"background","text":"Skills have emerged as a practical and effective approach for enhancing LLM agents at inference time through structured packages of knowledge.","hedge":{"p":1.0,"label":"asserted"},"own_p":0.08,"main_claim_p":0.04,"quantitative":false,"quantitative_p":0.02,"role_probabilities":{"gap":0.0,"method":0.0,"result":0.0,"background":1.0,"limitation":0.0,"future_work":0.0,"contribution":0.0,"interpretation":0.0}},{"i":1,"p":0.94,"id":"s1","own":false,"role":"gap","text":"However, existing evaluations largely measure whether skills improve aggregated task success, leaving a more fundamental question underexplored: \\emph{\\textbf{When do skills help, why do they work, and where do they fail?}} Through controlled experiments across various benchmarks, agent harnesses and LLMs, we isolate the effects of representation, outcome annotation, retrieval difficulty, and cross-framework robustness of skills.","hedge":{"p":0.99,"label":"asserted"},"own_p":0.32,"main_claim_p":0.07,"quantitative":false,"quantitative_p":0.08,"role_probabilities":{"gap":0.94,"method":0.0,"result":0.0,"background":0.0,"limitation":0.0,"future_work":0.0,"contribution":0.06,"interpretation":0.0}},{"i":2,"p":0.88,"id":"s2","own":true,"role":"method","text":"To further answer this question, we design a contrastive study that combines controlled quantitative experiments with paired trajectory analysis.","hedge":{"p":0.87,"label":"asserted"},"own_p":0.93,"main_claim_p":0.03,"quantitative":false,"quantitative_p":0.06,"role_probabilities":{"gap":0.0,"method":0.88,"result":0.0,"background":0.0,"limitation":0.0,"future_work":0.0,"contribution":0.12,"interpretation":0.0}},{"i":3,"p":0.99,"id":"s3","own":true,"role":"method","text":"We normalize 8,135 trial records from controlled experiments and retain 238 valid unique labels from 240 open-coded records.","hedge":{"p":1.0,"label":"asserted"},"own_p":0.95,"main_claim_p":0.03,"quantitative":true,"quantitative_p":0.98,"role_probabilities":{"gap":0.0,"method":0.99,"result":0.0,"background":0.0,"limitation":0.0,"future_work":0.0,"contribution":0.01,"interpretation":0.0}},{"i":4,"p":0.41,"id":"s4","own":true,"role":"interpretation","text":"We consolidate these observations into a taxonomy of three high-level categories and twelve skill-use modes: skills work when noisy trajectories become procedural anchors that stabilize execution.","hedge":{"p":1.0,"label":"asserted"},"own_p":0.93,"main_claim_p":0.42,"quantitative":true,"quantitative_p":0.93,"role_probabilities":{"gap":0.0,"method":0.01,"result":0.29,"background":0.0,"limitation":0.0,"future_work":0.0,"contribution":0.29,"interpretation":0.41}},{"i":5,"p":1.0,"id":"s5","own":true,"role":"result","text":"Skills improve over Workflow Memory by 6.06 points in matched comparisons.","hedge":{"p":1.0,"label":"asserted"},"own_p":0.92,"main_claim_p":0.08,"quantitative":true,"quantitative_p":0.99,"role_probabilities":{"gap":0.0,"method":0.0,"result":1.0,"background":0.0,"limitation":0.0,"future_work":0.0,"contribution":0.0,"interpretation":0.0}},{"i":6,"p":0.64,"id":"s6","own":true,"role":"result","text":"Procedural anchoring accounts for 65.7\\% of skill cases, versus 4.5\\% for explicit knowledge injection, showing that skills stabilize action rather than inject missing facts.","hedge":{"p":1.0,"label":"asserted"},"own_p":0.94,"main_claim_p":0.31,"quantitative":true,"quantitative_p":0.99,"role_probabilities":{"gap":0.0,"method":0.0,"result":0.64,"background":0.0,"limitation":0.0,"future_work":0.0,"contribution":0.0,"interpretation":0.36}},{"i":7,"p":1.0,"id":"s7","own":true,"role":"result","text":"Retrieval is a separate bottleneck: as pools grow from 5 to 100, actual-use precision falls from 29.6\\% to 3.3\\%.","hedge":{"p":1.0,"label":"asserted"},"own_p":0.91,"main_claim_p":0.07,"quantitative":true,"quantitative_p":0.99,"role_probabilities":{"gap":0.0,"method":0.0,"result":1.0,"background":0.0,"limitation":0.0,"future_work":0.0,"contribution":0.0,"interpretation":0.0}},{"i":8,"p":0.79,"id":"s8","own":true,"role":"result","text":"Confusable distractors impair offline identification, yet downstream success remains stable; exact ground-truth invocation is neither sufficient nor necessary.","hedge":{"p":1.0,"label":"asserted"},"own_p":0.9,"main_claim_p":0.09,"quantitative":false,"quantitative_p":0.12,"role_probabilities":{"gap":0.0,"method":0.0,"result":0.79,"background":0.0,"limitation":0.0,"future_work":0.0,"contribution":0.0,"interpretation":0.21}},{"i":9,"p":0.88,"id":"s9","own":true,"role":"result","text":"Skills fail under brittle assumptions, incompatible contexts, or insufficient adaptation.","hedge":{"p":1.0,"label":"asserted"},"own_p":0.81,"main_claim_p":0.17,"quantitative":false,"quantitative_p":0.03,"role_probabilities":{"gap":0.0,"method":0.0,"result":0.88,"background":0.0,"limitation":0.03,"future_work":0.0,"contribution":0.0,"interpretation":0.09}},{"i":10,"p":0.91,"id":"s10","own":true,"role":"interpretation","text":"These findings move evaluation beyond aggregate success rates and guide reliable self-evolving agents.","hedge":{"p":0.98,"label":"asserted"},"own_p":0.56,"main_claim_p":0.41,"quantitative":false,"quantitative_p":0.03,"role_probabilities":{"gap":0.0,"method":0.0,"result":0.0,"background":0.0,"limitation":0.0,"future_work":0.07,"contribution":0.02,"interpretation":0.91}}]},"compiled":{"main_claim":{"p":0.76,"text":"We consolidate these observations into a taxonomy of three high-level categories and twelve skill-use modes: skills work when noisy trajectories become procedural anchors that stabilize execution.","location":"abstract s4","hedge_strength":{"p":1.0,"label":"asserted"}},"provenance":{"route":{"via":"TypeSafe API","model":"classifier-1.13.0","endpoint":"https://api.typesafe.ai/v1/systemone","provider":"TypeSafe"},"packing":"one request per paper","requests":1,"input_tokens":8240,"cost_usd_reported":0.0},"compiled_at":"2026-09-23","bank_version":"abstract-p1-20260923","model_version":"classifier-1.13.0","abstract_states":{"claims_first":0.07,"states_limitation":0.46,"human_participants":0.21,"controlled_comparison":0.91,"randomized_assignment":0.04,"states_quantitative_result":0.99,"compares_with_prior_results":0.93,"mentions_code_or_data_availability":0.04},"evidential_basis":{"p":1.0,"value":"observation or experiment","probabilities":{"argument":0.0,"literature":0.0,"proof_or_derivation":0.0,"analysis_of_existing_data":0.0,"observation_or_experiment":1.0,"simulation_or_computation":0.0}},"contribution_type":{"p":1.0,"value":"empirical study","probabilities":{"review":0.0,"method_or_tool":0.0,"empirical_study":1.0,"simulation_study":0.0,"theoretical_result":0.0,"dataset_or_benchmark":0.0,"position_or_commentary":0.0}},"contribution_statement":{"p":0.77,"text":"To further answer this question, we design a contrastive study that combines controlled quantitative experiments with paired trajectory analysis.","location":"abstract s2"}},"coverage":{"tier":"abstract","levels_run":["P abstract"],"not_examined":["full text","figures and tables","references and citing sentences","relations to other papers"],"sources_read":["arXiv listing abstract v1"],"levels_not_run":["S sections","T body sentences","C citations","R cross-paper relations"],"sentences_examined":11,"sentences_in_abstract":11},"measured_accuracy":{"url":"https://pith.science/accuracy","measured_on":"2026-10-02","method":"two independent model families labelled a fresh random sample blind; the figure is how often the published label matches their consensus","sentence_role_agreement":0.92,"main_claim_agreement":0.908,"typed_citation_agreement":0.933,"typed_citation_agreement_95_interval":[0.917,0.945],"review_comments_accurate":0.758,"review_comments_accurate_95_interval":[0.684,0.82],"review_caught_admitted_error_found_or_related":0.448,"review_comments_judged":{"judged_by":["gpt-5.6-luna","grok-4.7"],"judged_against":"Pith's own text copy of each paper: text extracted from the PDF, the first 90,000 characters","measured_on":"2026-10-04","accurate":113,"decided":149,"sampled_reviews":200,"split_by_lane":false,"by_lane":"https://pith.science/accuracy#by-lane"}}}