Type: url
Original: https://arxiv.org/abs/2602.12430
Status: generated
Summary
The arXiv abstract page for a survey of agent skills - composable packages of instructions, code and resources that LLM agents load on demand - organized along four axes: architectural foundations, skill acquisition, deployment at scale, and security. Citing analyses that found vulnerabilities in 26.1% of community-contributed skills, it proposes a four-tier, gate-based Skill Trust and Lifecycle Governance Framework and identifies seven open challenges for trustworthy skill ecosystems.