# Who Maintains Agent Skills? A Longitudinal Study of Human-Governed, AI-Assisted Skill Maintenance

arXiv: https://arxiv.org/abs/2609.05677 · HTML: https://mlsanity.com/paper/2609.05677/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Placed in Agent skill learning and evolution, one of 16 directions in LLM agents and tool use.
- Builds on 31 of its 36 references that are in the map.
- First posted on arXiv 4 Sep 2026.

## Placed in

- LLM agents and tool use / [Agent skill learning and evolution](https://mlsanity.com/agents/agent-skill-learning-and-evolution/)

## Builds on (most cited first, 30 of 31)

- [Reflexion: Language Agents with Verbal Reinforcement Learning](https://mlsanity.com/paper/2303.11366/) (2023)
- [Voyager: An Open-Ended Embodied Agent with Large Language Models](https://mlsanity.com/paper/2305.16291/) (2023)
- [ExpeL: LLM Agents Are Experiential Learners](https://mlsanity.com/paper/2308.10144/) (2023)
- [SkillsBench: Benchmarking How Well Agent Skills Work Across Diverse Tasks](https://mlsanity.com/paper/2602.12670/) (2026)
- [Trace2Skill: Distill Trajectory-Local Lessons into Transferable Agent Skills](https://mlsanity.com/paper/2603.25158/) (2026)
- [SkillOpt: Executive Strategy for Self-Evolving Agent Skills](https://mlsanity.com/paper/2605.23904/) (2026)
- [The BiGGen Bench: A Principled Benchmark for Fine-grained Evaluation of Language Models with Language Models](https://mlsanity.com/paper/2406.05761/) (2024)
- [SWE-Skills-Bench: Do Agent Skills Actually Help in Real-World Software Engineering?](https://mlsanity.com/paper/2603.15401/) (2026)
- [Agent Skills: A Data-Driven Analysis of Claude Skills for Extending Large Language Model Functionality](https://mlsanity.com/paper/2602.08004/) (2026)
- [How Well Do Agentic Skills Work in the Wild: Benchmarking LLM Skill Usage in Realistic Settings](https://mlsanity.com/paper/2604.04323/) (2026)
- [SkillLearnBench: Benchmarking Continual Learning Methods for Agent Skill Generation on Real-World Tasks](https://mlsanity.com/paper/2604.20087/) (2026)
- [SKILLFOUNDRY: Building Self-Evolving Agent Skill Libraries from Heterogeneous Scientific Resources](https://mlsanity.com/paper/2604.03964/) (2026)
- [MUSE-Autoskill: Self-Evolving Agents via Skill Creation, Memory, Management, and Evaluation](https://mlsanity.com/paper/2605.27366/) (2026)
- [Agent READMEs: An Empirical Study of Context Files for Agentic Coding](https://mlsanity.com/paper/2511.12884/) (2025)
- [Prompting in the Wild: An Empirical Study of Prompt Evolution in Software Repositories](https://mlsanity.com/paper/2412.17298/) (2024)
- [SkillGraph: Skill-Augmented Reinforcement Learning for Agents via Evolving Skill Graphs](https://mlsanity.com/paper/2605.12039/) (2026)
- [Context Engineering for AI Agents in Open-Source Software](https://mlsanity.com/paper/2510.21413/) (2025)
- [More Skills, Worse Agents? Skill Shadowing Degrades Performance When Expanding Skill Libraries](https://mlsanity.com/paper/2605.24050/) (2026)
- [SkillAdaptor: Self-Adapting Skills for LLM Agents from Trajectories](https://mlsanity.com/paper/2606.01311/) (2026)
- [SkillHone: A Harness for Continual Agent Skill Evolution Through Persistent Decision History](https://mlsanity.com/paper/2606.08671/) (2026)
- [Learning from Supervision with Semantic and Episodic Memory: A Reflective Approach to Agent Adaptation](https://mlsanity.com/paper/2510.19897/) (2025)
- [Promises, Perils, and (Timely) Heuristics for Mining Coding Agent Activity](https://mlsanity.com/paper/2601.18345/) (2026)
- [WINELL: Wikipedia Never-Ending Updating with LLM Agents](https://mlsanity.com/paper/2508.03728/) (2025)
- [An Empirical Study of the Evolution of GitHub Actions Workflows](https://mlsanity.com/paper/2602.14572/) (2026)
- [A Dataset of Agentic AI Coding Tool Configurations](https://mlsanity.com/paper/2605.08435/) (2026)
- [ExpGraph: Model-Agnostic Experience Learning with Graph-Structured Memory for LLM Agents](https://mlsanity.com/paper/2605.30712/) (2026)
- [The AI Attribution Paradox: Transparency as Social Strategy in Open-Source Software Development](https://mlsanity.com/paper/2512.00867/) (2025)
- [Rule Taxonomy and Evolution in AI IDEs: A Mining and Survey Study](https://mlsanity.com/paper/2606.12231/) (2026)
- [Mining Architectural Quality Under Agentic AI Adoption: A Causal Study of Java Repositories](https://mlsanity.com/paper/2606.13298/) (2026)
- [Beyond Simpson's Paradox: A Cascade of Confounders in AI Agent Pull-Request Co-Authorship](https://mlsanity.com/paper/2606.22711/) (2026)

## Abstract (arXiv)

Lifelong LLM agents increasingly rely on external skill artifacts as one element for preserving and reusing capabilities over time. These skills (usually portable Markdown files such as SKILL.md) describe when and how to apply a capability and must be corrected, expanded, and consolidated as tools and usage patterns shift over deployment. Recent work seeks to automate skill curation, but it largely evaluates against automated baselines and treats human maintenance as an unmeasured bottleneck. We study that missing process directly. We mine the full commit histories of five public AI-skill repositories, a purposive sample of AI-tooling organizations, covering 873 commits, 143 skill files, and 254 substantive post-creation edits from October 2025 to June 2026. We code each edit with pre-registered governance, operation, and trigger-evidence codebooks. Three findings emerge. First, every substantive edit is authored or merged through a named human account, while 62% carry an AI co-author trailer, with large repository-level variation. Second, these edits are genuine curation: an audited sample shows that most change skill content, and the coded operations are dominated by additions and corrections. Third, a pre-registered rule-likeness axis fails its reliability gate; reliably coding rule-likeness from commit artifacts remains an open measurement problem. We release the corpus, codebooks, mining scripts, and a replay protocol for automated skill curators. For self-evolving agents, public skill maintenance currently looks less like an autonomous pipeline than a human-governed, AI-assisted loop that future curators must measure against and operate within.