← Pause and Practice

PART 4 · RECORD AND PLAYBACK · TWO PROMPTS

Record once.
Replay intelligently.

Use a browser-capable AI coding agent. Let it perform a local TestBucks browsing flow, record what it actually did as intent, then replay that intent—not a fixed sequence of clicks.

1

Perform the flow. Save the intent.

Test TESTBUCKS_URL by finding a vanilla latte and inspecting its available size and customization information. As you work, observe and record the flow you actually execute, including actions, visible results, and any issues.

Then save that completed flow as an OpenTest.ai otai/testcase v1 intent-based test in testbucks-latte-intent.json. Use a title and goal; add optional steps with description, action, target, value, and expected only where useful. Preserve the outcome to verify, not brittle selectors or a fixed click path. Keep the execution evidence separate from the reusable test definition. If blocked or unable to finish, report that and label the intent as unverified—do not invent a successful flow.

Use the fictional local TestBucks fixture. Start at Home with the customization dialog closed. Inspect options without adding to the cart. Do not submit orders or payments.
Observed flow → saved intent → fresh execution
2

Replay the saved intent.

Read testbucks-latte-intent.json and replay that intent against the current TestBucks interface. Use the goal and expected outcomes to decide how to execute it now. Inspect the live page and choose the current controls dynamically, even if the layout or route has changed. Do not reuse old selectors blindly or weaken the expected result.

Compare this run with the recorded run. Report what changed, whether the same intent was achieved, and the evidence supporting PASS, FAIL, or BLOCKED. Do not call a missing feature or blocked flow a pass. Keep the saved intent unchanged and save the new execution evidence separately.

Start fresh at Home with the customization dialog closed. Inspect options without adding to the cart. Do not submit orders or payments.

Use the same agent workspace for both prompts, or attach the saved JSON to the second conversation. If the interface has not changed, you are practicing replay—not demonstrating adaptation to a change. Use an approved modified TestBucks copy to deliberately change the interface.

No live AI or browser automation runs on this lesson page. The exercise stops before a real purchase.

OpenTest.ai test-case format ↗