Agent-skill catalogs already top 700,000 entries — curation hasn't caught up

Notes on A Comprehensive Survey on Agent Skills: Taxonomy, Techniques, and Applications (arXiv:2605.07358) — Yingli Zhou, Wang Shu, Yaodong Su, Wenchuan Du, Yixiang Fang, Xuemin Lin · May 2026

Note published July 25, 2026 · written by SkillFed’s research pipeline from the paper above · how these notes are made

Agent skills now have a formal definition: a skill is the tuple S = (M, R, C) — a root instruction document M, a set of auxiliary resources R (references, templates, scripts), and applicability conditions C that govern when the skill gets retrieved and used. The point of formalizing it is closing the procedural gap — giving an agent a tool through an API or the Model Context Protocol tells it what it can do, not when to do it, how to sequence it with other capabilities, or how to judge whether the result succeeded. Literature tracked from April 2023 to April 2026 sorts into four lifecycle stages — representation, skill acquisition, skill retrieval and selection, and skill evolution — spanning eight application domains, from software engineering and web/GUI automation to robotics, finance, healthcare, and social simulation.

The skill economy this survey maps is already at real scale: SkillsMP indexes over 700,000 skills, SkillNet 300,000+, Skills.sh 90,000+, SkillHub 80,000+, and ClawHub 40,000+. Scale has outrun the discipline needed to manage it. Acquisition — minting new skills from expert write-ups, execution traces, one-off tasks, or scraped corpora — is well studied and getting cheaper with LLM-assisted authoring. Evolution is the laggard: reliably revising, validating, and retiring skills already in a shared library gets far less attention, and current systems are much better at adding artifacts than at safely rewriting or retiring them. Two findings show why that gap matters. SkillsBench finds that even curated skills can carry negative net utility on specific tasks despite looking relevant, and PoisonedSkills shows that third-party skill documentation can hide malicious logic an agent then executes as trusted guidance.