SkillHone：基于持久决策历史的持续智能体技能演进工具

View PDF HTML (experimental)

Abstract:Agent skills extend language-model agents with task-specific procedures, scripts, and references, but the tasks and environments they target continually change. Existing methods improve skills in bounded runs and retain only the final artifact, discarding the decision history that later agents need to interpret prior revisions, evaluations, and rejected alternatives. We introduce SkillHone, a harness for continual agent skill evolution grounded in persistent decision history. SkillHone pairs skill revisions with evaluation-side evidence that supplies practice feedback, recording structured histories of diagnoses, revisions, evidence, and outcomes. Role-separated subagents run candidate skills on practice probes with redacted reporting and propose revisions informed by prior decisions, enabling cross-session refinement without rediscovering past rationale. On deep-research benchmarks, SkillHone runs without a pre-integrated search stack and outperforms the commercially backed deep-research agent by 15.8 points on GAIA and 3.2 points on WebWalkerQA-EN, while also exceeding prior skill-evolution methods. We further deploy SkillHone on internal tool-mediated analysis scenarios, where it improves accuracy by an average of 18.8 points across seven settings.

Subjects:	Machine Learning (cs.LG)
Cite as:	arXiv:2606.08671 [cs.LG]
	(or arXiv:2606.08671v3 [cs.LG] for this version)
	https://doi.org/10.48550/arXiv.2606.08671 arXiv-issued DOI via DataCite

Submission history

From: Zhiwei Li [view email]
[v1] Sun, 7 Jun 2026 15:21:08 UTC (935 KB)
[v2] Tue, 23 Jun 2026 08:10:11 UTC (472 KB)
[v3] Mon, 6 Jul 2026 18:50:16 UTC (472 KB)