Use this page when you want to check a setup detail, adapt an example, or prepare a session for colleagues. It collects the official documentation behind the agent setup guide and notes what each source helps explain.
Source review date: September 5, 2026. We retrieved the relevant pages to check modes, tool access, paths, configuration fields, instruction discovery, and documented capabilities. The examples use those formats with tasks and file paths from this repository.
A feature described in documentation still needs a hands-on check in the client you plan to use. We distinguish those two kinds of evidence throughout the guide. Installation and sign-in links lead to the provider’s current instructions because account and operating-system requirements can change.
Distinguishing agent-workflow checks from ordinary code tests
University policy and lab data
These sources support lab data and university policy. We reviewed the public guidance; the detailed Sensitive Data Guide service listings require U-M sign-in and were not inspected. Check those listings for the exact service and data type before a real study task.
A short rehearsal is useful before teaching a session or publishing a release. Focus on the clients you will actually demonstrate:
Reopen the pages for installation, mode selection, discovery paths, configuration examples, permissions, and feature availability. A working link is only part of the check; make sure its content still supports the instructions.
Try orientation, a plan-to-implementation handoff, skill discovery, and one small review in each client you intend to show. Record the version, model, selected modes and permissions, whether you used an editor, CLI, or hosted service, the operating system, and the result. These details will help a colleague troubleshoot a different outcome.
Update the review date for the material you actually checked. If a feature is unavailable or its documentation no longer supports an example, explain the limitation and remove the unsupported instructions.
Keep the exercise’s agreed behavior the same while updating client setup files. After changing filenames or navigation, recheck the examples and links.
The agent workflow discussion brings these documented features together and suggests ways to explore them in this repository. The sources establish what the tools describe as available. Claims about productivity, model rankings, scientific validity, or how much human review a task needs would require additional evidence.
