AWS explains the architecture of agents and skills in AWS Quick
AWS describes how an agent, skills, and connected data work together in AWS Quick – from the default Quick agent through custom agents and system skills to connections with spaces, knowledge bases, and connectors (Slack, Jira, Salesforce, etc.).
AWS explains in its blog how agents, skills, and connected data work together in AWS Quick. AWS Quick is an AI tool for automating tasks, analyzing data, building web applications, and research, with which the user communicates in natural language. An agent is a configurable AI with its own instructions, knowledge sources (spaces), and connectors; the default built-in "Quick" agent has access to all the resources the user has access to, so there's no need to determine in advance what data a task needs. Alongside it, custom agents can be created that are restricted to specific spaces, connectors, and skills, for example an HR assistant for onboarding, which, once tested, can be shared with other users or groups within the account.
A skill is a reusable set of instructions that teaches the agent a specific recurring task. The agent runs in a controlled loop based on the "reasoning and acting" (ReAct) principle, similar to frameworks such as the Strands Agents SDK. Quick's default behavior largely comes from system skills – these include generating Word, PowerPoint, Excel, and PDF documents, browsing the web, running code, deep research, transcribing audio and video, generating images, managing AWS Quick Flows and routines, handing off work to the Kiro coding agent, or managing memory and to-do tasks. System skills can be turned on and off in Quick, but according to the source, they cannot be edited or duplicated.
Agents connect to data through spaces (a curated set of documents and resources for an individual, team, or group), knowledge bases (indexed content synced from sources such as Google Drive, Microsoft OneDrive, Confluence, SharePoint, Amazon S3, a web crawler, or Amazon Bedrock Managed Knowledge Bases), and datasets/dashboards from AWS QuickSight. Connectors then allow agents to read data and take actions in external applications – cloud connectors cover, among others, Outlook, Slack, Jira, Asana, Salesforce, and remote MCP servers, while local connectors only work on the machine where the user hosts them themself.
The article covers only the part described here; further details on creating and sharing custom skills can be found in the source article.
Why it matters
The architecture described allows companies and individuals to automate recurring work tasks without changing access rights – the agent always works only with the data and tools the user already has permission to use. For teams, it's key that custom agents can be restricted to specific resources and shared across the organization, which makes it possible to standardize processes (e.g. onboarding) without having to build their own AI infrastructure.
Two audiences, two different impacts
What this means
For individuals
An AWS Quick user can write down recurring tasks (weekly status updates, release notes, creating tickets) once as a skill and then run them using natural language, instead of explaining them to the agent again each time.
For a business
Companies can create custom agents in AWS Quick that are restricted to specific spaces, connectors, and skills (e.g. an HR onboarding assistant), standardizing recurring team processes, while the agent never has broader access to data and tools than the user themself has.
ProductivityCheck the original
Event sources
only one source so far · 1 publisher, 0 independent. We count feeds from the same owner only once.