Now that your session has done real work, there is something to inspect. Confirm what the agent is pointed at, then let it drive the quiz in the integrated browser and report what actually happened rather than what it intended.
In this lesson, you will:
- read the session details panel.
- check the session’s project, path, branch, changes, and context usage.
- run a browser-level smoke test and fix any failures.
Read the session details
The session details tell you exactly what the agent is working on. You do not need to watch this panel constantly, but everything in it matters when a result surprises you.
The panel shows where the work is happening, what has changed, and how full the context window is. There is no model row, because you choose the model for each request in the composer.
- Confirm that the project, path, and branch are the ones you think you are editing.
- Read Changes to see whether this session has touched anything yet.
- Check context usage. As it climbs, the agent has less room for your actual task, and that is your cue to start a fresh session.
Test before Git writes anything
The integrated browser is a real browser, so the agent can drive the quiz and verify its behavior. Send the following prompt:
Run a browser-level smoke test for the quiz in the integrated browser. Check keyboard navigation, score updates, correct and incorrect feedback, and the results screen. Fix any failures, then report what passed.
- Watch the integrated browser as the agent runs through the questions.
- If anything fails, let the agent fix it and rerun the test until everything passes.
- Move on only when the build and test are green.
Nothing has been written to Git yet. The next step is /init, which reads the project as it stands, so it is worth making sure the project works first.
Summary and next steps
You confirmed what the session is working on and verified the quiz with a browser-level smoke test. Continue to Lesson 4: Capture project instructions.