SkillProx: Self-Evolving Agent Skills via Proximal Textual Gradient Descent
2026-08-07 • Artificial Intelligence
Artificial IntelligenceComputation and Language
AI summaryⓘ
The authors introduce SkillProx, a method to help AI language models improve their repeated task-solving skills more effectively. Their approach uses a two-step process: first, it identifies and fixes mistakes while keeping track of feedback, and second, it reviews the skill components to keep only the most useful parts. This framework balances improving task performance and keeping skills simple. Experiments show SkillProx leads to better accuracy than previous methods, and tests confirm that both its diagnosis and refinement steps contribute to its success.
LLM agentsprocedural knowledgeskill refinementproximal gradientclosed-loop diagnosistask lossknowledge consolidationleave-one-out utilityin-distributionout-of-distribution
Authors
Mingxuan Zheng, Yujin Zhou, Chuxue Cao, Boqin Yin, Yuyao Zhang, Jiapeng Sun, Shuaishuai Gong, Sirui Han, Yike Guo
Abstract
LLM agents increasingly adapt to recurring tasks by accumulating procedural knowledge in skills. These skills are lightweight, reusable textual artifacts that are loaded into the agent's context without weight updates. Recent methods refine skills through iterative task execution, failure diagnosis, and trajectory-guided text-space updates. However, existing frameworks lack explicit diagnosis--outcome feedback and treat deletion as a generic edit operation rather than a dedicated mechanism for consolidating accumulated knowledge. We introduce SkillProx, a proximal-gradient-inspired forward--backward framework that couples closed-loop diagnostic evolution with utility-aware proximal refinement. Motivated by a composite objective balancing task loss and skill complexity, the forward stage re-executes diagnosis-driven edits on the same task batch, rolls back regressions, and feeds measured outcomes into subsequent diagnoses. The backward stage decomposes the resulting skill into auditable knowledge units, estimates their contributions using a frozen leave-one-out utility audit, and applies validation-gated consolidation, demotion, or removal. Experiments on in-distribution and out-of-distribution benchmarks across multiple backbone LLMs show that SkillProx improves average accuracy by 3.0 percentage points over the strongest gradient-based baseline. Component ablations demonstrate the complementary effects of closed-loop diagnosis and proximal refinement.