Browser automation becomes more dependable when it is treated as a sequence of observed state changes, not a sequence of clicks. A button press is an action; evidence that the intended result occurred is a separate requirement.
1. Start with a fresh view
Read the current page before choosing a control. Labels, dialogs, and form fields may change between visits. Prefer visible labels and roles over fragile coordinates, and refresh the view after navigation.
2. Make each action small
Enter one logical value at a time when dealing with sensitive fields. Use short, unique request identifiers so a retry can be distinguished from an earlier attempt. Avoid repeating a submit action just because the response is slow.
3. Separate action from confirmation
A click reported as successful only tells you that the automation performed the click. It does not prove that a server accepted a form or that a post became public. Look for a new page, a success message, or another observable change before deciding what happened.
4. Respect human checkpoints
CAPTCHAs, verification codes, and account challenges exist for a reason. When one appears, pause automated interaction and let the account owner complete it. Resume from a fresh page view instead of guessing at hidden state.
5. Plan for partial progress
Keep track of what has already succeeded. If a workflow stops after saving a draft or sending a verification email, resuming should not create duplicates. Where possible, design operations to be safe when retried.
6. Verify the final outcome
For a published article, check its public page without relying on an authenticated editor session. Match the intended title and report precisely what was observed. If independent verification has not succeeded, say so.
Reliable automation is less about moving quickly through a website and more about knowing when to proceed, when to retry, and when to stop.