integrate-atlas-chat

Adds a streaming Atlas Agent chat UI to a Cognite Flows app, handling the installation, component structure, and hook wiring so that integration code for useAtlasChat is not written by hand. It is meant to be used whenever building a chat interface backed by Atlas agents, triggered by terms such as useAtlasChat, atlas chat, streaming chat, agent chat, or chat UI. The skill accepts an agent external ID as its argument, identifying which Atlas agent the chat should connect to.

The documentation is intentionally minimal and directs the work through the official Cognite guide on AI agent integration for Flows rather than restating the implementation inline. Its declared tools include reading, globbing, grepping, editing, writing, and running Bash, which fit a task that installs packages and modifies the app's source to wire up the chat component and the useAtlasChat hook. Because the SKILL.md itself does not enumerate component code, parameters, or package names, the concrete steps are expected to follow the linked guide.

The skill also positions itself within a larger sequence for building a full chat application. For a complete chat app, the recommended order is to run integrate-atlas-chat first, then create-client-tool once per tool the chat needs, and then setup-python-tools if Python tools are required. This makes it the entry point and foundation step, establishing the streaming chat surface before tool-specific skills layer additional capabilities on top.