Understand this story
On November 7, 2026, Thomas Delteil published an article explaining the taxonomy of Amazon Quick agent primitives. This taxonomy breaks down agent work using four questions—When, What, Who, and How—corresponding to three layers: triggers, agents, and capabilities. Among these, triggers are divided into four categories: human-initiated, agent-delegated, scheduled, and event-driven; agents are defined as persistent learning executors composed of instructions, capabilities, and acquired context, with the underlying LLM being merely a replaceable parameter; capabilities encompass tools, skills, and workflows, with skills loaded via progressive disclosure and workflows fixing step order in code. Currently, this article has been published by AWS Machine Learning, and there have been no subsequent developments.
AI-generated from public reports
Latest development · Report
Amazon Quick Agent Primitive TaxonomyDevelopments
1 confirmed developments
Thomas Delteil published an article explaining the Amazon Quick agent primitive taxonomy
Source:AWS Machine Learning
Reports and discussion timeline
Browse reports and discussion by time. Their evidence levels remain distinct.
Nov 7
AWS Machine LearningFirst-party
Amazon Quick Agent Primitive Taxonomy
The Amazon Quick team proposed a taxonomy of agent primitives that breaks down agent work using four questions—When, What, Who, and How—corresponding to three layers: triggers, agents, and capabilities. Triggers fall into four categories: human-initiated, agent-delegated, scheduled, and event-driven. An agent is defined as a persistent, learning executor composed of instructions, capabilities, and acquired context, with the underlying LLM serving merely as a replaceable parameter. Capabilities encompass tools, skills, and workflows; skills are loaded via progressive disclosure, and workflows fix step order in code.