Oriel — AI & technology worth noticing, every day.

Taxonomy of agentic primitives in Amazon Quick

Web page · https://aws.amazon.com/blogs/machine-learning/taxonomy-of-agentic-primitives-in-amazon-quick

As we built Amazon Quick, its agents grew from answering questions to taking actions, and the same primitives kept getting rebuilt under different names. This post shares the taxonomy we converged on — triggers, agents, capabilities, and learned context — found by decomposing agentic work into four plain questions: When, What, Who, and How.

First seen by Oriel on .

AI summary

This post presents Amazon Quick’s taxonomy of agentic primitives, organizing work around triggers, persistent learning agents, and capabilities governed by different confidence contracts.

  • Triggers combine activation and task instructions, covering human-initiated, agent-delegated, scheduled, and event-driven execution.
  • An agent is a persistent, learning actor defined by standing instructions, dynamically available capabilities, and accumulated learned context.
  • Tools provide typed atomic operations, skills provide progressive-disclosure guidance, and workflows coordinate deterministic steps with coded control flow and error handling.
  • Skills capture and share institutional expertise, while workflows provide auditability and deterministic coordination; either can evolve into the other through crystallization or dissolution.
  • A Slack digest request may use a simple scheduled instruction, a skill, or a workflow, depending on the user’s required confidence and consistency.

#agentic ai #amazon quick #workflows #model context protocol

Generated by gpt-5.6-sol. Check the original for details.

Aliases

SourceExternal IDURL
AWS Machine Learning Blog755c7d8248913b9c1698eceff615a3e12637e5e9https://aws.amazon.com/blogs/machine-learning/taxonomy-of-agentic-primitives-in-amazon-quick/ Read discussion