# SkillGrad: Optimizing Agent Skills Like Gradient Descent

Source: [SkillGrad: Optimizing Agent Skills Like Gradient Descent](<https://www.sota2.com/research/paper/6a17cc27dbd7150506f8139c>)

Collection index: [llms.txt](<https://www.sota2.com/research/llms.txt>)

Data updated: 2026-05-31T06:00:01.495Z

Authors: Hanyu Wang, Yifan Lan, Bochuan Cao, Lu Lin, Jinghui Chen

Published: 2026-05-26T23:18:43.000Z

## Abstract

Agent skills provide a lightweight way to adapt LLM agents to specialized domains by storing reusable procedural knowledge in structured files. However, whether downloaded from third parties or self-generated, these skills are often unreliable, incomplete, or outdated. Existing skill-evolution methods often address these deficiencies through heuristic reflections without an explicit optimization formulation. In this paper, we propose SkillGrad, a gradient-descent-inspired framework for optimizing agent skills. SkillGrad treats the skill package as a structured parameter to optimize in a gradient descent fashion: task executions provide trajectory-level loss evidence, automatic diagnoses then provide text-based gradients that indicate the correction directions. To stabilize optimization across iterations, a momentum agent accumulates recurring diagnostic patterns into a persistent memory overlay. Finally, an LLM-based patcher executes the parameter update by applying layer-aware edits to the skill package. Evaluated on SpreadsheetBench Verified and WikiTableQuestions, SkillGrad consistently outperforms training-based skill evolution baselines across two backbone LLMs, improving over the strongest training-based baseline by $6.7$ percentage points on average. Ablations further show that momentum and contrastive diagnosis both contribute to the final skill quality.

## Recorded benchmark results

A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage.

- [Spreadsheet Reasoning on SpreadsheetBench Verified](<https://www.sota2.com/research/sota/spreadsheet-reasoning-on-spreadsheetbench-verified.md>): Spreadsheet Reasoning; SpreadsheetBench Verified; Accuracy: 71.11.

- [Table Question Answering on WikiTableQuestions OOD](<https://www.sota2.com/research/sota/table-question-answering-on-wikitablequestions-ood.md>): Table Question Answering; WikiTableQuestions OOD; Accuracy: 83.34.

## Original sources and implementation

[Original paper](<https://arxiv.org/abs/2605.27760>)

[PDF](<https://arxiv.org/pdf/2605.27760.pdf>)

[Implementation](<https://github.com/wwwhy725/SkillGrad>)