# SkillProx: Self-Evolving Agent Skills via Proximal Textual Gradient Descent

arXiv: https://arxiv.org/abs/2608.07449 · HTML: https://mlsanity.com/paper/2608.07449/

## Key facts

- Cited by 6 papers in the map (data through 14 Oct 2026).
- Placed in Agent skill learning and evolution, one of 16 directions in LLM agents and tool use.
- Builds on 13 of its 21 references that are in the map.
- First posted on arXiv 7 Aug 2026.

## Placed in

- LLM agents and tool use / [Agent skill learning and evolution](https://mlsanity.com/agents/agent-skill-learning-and-evolution/)
- Synthetic data and self-improvement for LLMs / [Self-evolving LLM agents](https://mlsanity.com/synth/self-evolving-llm-agents/)

## Builds on (most cited first, 13 of 13)

- [ReAct: Synergizing Reasoning and Acting in Language Models](https://mlsanity.com/paper/2210.03629/) (2022)
- [Compositional Semantic Parsing on Semi-Structured Tables](https://mlsanity.com/paper/1508.00305/) (2015)
- [ReTool: Reinforcement Learning for Strategic Tool Use in LLMs](https://mlsanity.com/paper/2504.11536/) (2025)
- [Trace2Skill: Distill Trajectory-Local Lessons into Transferable Agent Skills](https://mlsanity.com/paper/2603.25158/) (2026)
- [EASYTOOL: Enhancing LLM-based Agents with Concise Tool Instruction](https://mlsanity.com/paper/2401.06201/) (2024)
- [SkillOpt: Executive Strategy for Self-Evolving Agent Skills](https://mlsanity.com/paper/2605.23904/) (2026)
- [SoK: Agentic Skills -- Beyond Tool Use in LLM Agents](https://mlsanity.com/paper/2602.20867/) (2026)
- [CoEvoSkills: Self-Evolving Agent Skills via Co-Evolutionary Verification](https://mlsanity.com/paper/2604.01687/) (2026)
- [Skill-it! A Data-Driven Skills Framework for Understanding and Training Language Models](https://mlsanity.com/paper/2307.14430/) (2023)
- [Externalization in LLM Agents: A Unified Review of Memory, Skills, Protocols and Harness Engineering](https://mlsanity.com/paper/2604.08224/) (2026)
- [SkillGrad: Optimizing Agent Skills Like Gradient Descent](https://mlsanity.com/paper/2605.27760/) (2026)
- [LRAS: Advanced Legal Reasoning with Agentic Search](https://mlsanity.com/paper/2601.07296/) (2026)
- [Skill is Not One-Size-Fits-All: Model-Aware Skill Alignment for LLM Agents](https://mlsanity.com/paper/2605.30723/) (2026)

## Built on by (most cited first, 6 of 6)

- [COBRA-Skills: Contextual Bandit-Guided Evolution for Agent Skill Optimization](https://mlsanity.com/paper/2609.11682/) (2026)
- [SkillPoison: Progressive Skill Poisoning via Successful Experiences](https://mlsanity.com/paper/2610.07645/) (2026)
- [LexAgentHallu: A Hierarchical Benchmark for Profiling Hallucinations in Legal Agents](https://mlsanity.com/paper/2609.09754/) (2026)
- [EVISKILL: Grounding Skill Evolution in Replayable Evidence](https://mlsanity.com/paper/2610.05030/) (2026)
- [From Existing Methods to New Research Ideas: Two Perspectives on Recursive Self-Improvement](https://mlsanity.com/paper/2610.14625/) (2026)
- [EvoQuantLab: Training-Free Dual-Loop Evolution for Auditable Quantitative Research](https://mlsanity.com/paper/2610.17130/) (2026)

## Abstract (arXiv)

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.