Completion checks
- The command or agent run has ended.
- Expected files or configuration exist.
- Relevant tests, lint, or build commands ran.
- Any failures are explained, not hidden.
- If deployment happened, the live application works in the browser.
Ask for evidence
A useful final report should say what changed, which commands ran, what passed, what was not tested, and what the customer should manually check.
Do not trust wording alone. An agent can write "done" even if the build failed or the live environment was not checked. Verification must be concrete.
Useful checks
- Open the relevant page and test the user flow.
- Confirm the status or health endpoint responds.
- Check that the service or container is running.
- Compare the live environment with the expected Git version.