Towards a Systems Foundation for Agentic Skills: Architecture, Lifecycle, and Security
The paper argues that “agentic skills” need a systems layer: lifecycle, architecture, evaluation, and security rules around reusable execution knowledge.
The authors define skills as portable procedural artifacts that sit between LLM planning and deterministic execution. They map a nine-stage lifecycle, from discovery and authoring through retrieval, orchestration, repair, adaptation, evaluation, and governance. The abstract also flags marketplaces, registries, adversarial threats, runtime verification, and defenses as part of the needed foundation. ArXiv · AI/CL/LG's note
The authors define skills as portable procedural artifacts that sit between LLM planning and deterministic execution. They map a nine-stage lifecycle, from discovery and authoring through retrieval, orchestration, repair, adaptation, evaluation, and governance. The abstract also flags marketplaces, registries, adversarial threats, runtime verification, and defenses as part of the needed foundation. ArXiv · AI/CL/LG's note
score 4