AI agent says done but did not work?
Do not accept the work just because the update sounds confident. Use this quick check first: can you see the result, can you inspect the proof, and does the proof match the original goal?
If any answer is missing, the work is not ready to accept yet.
- What changed, in plain English?
- Where can I inspect it with my own eyes?
- What proof shows the original goal is true?
- What is still not proven?
Copy this reply.
You said this is done. Before I accept it, send me a completion receipt. Use plain English. Include: 1. What changed. 2. The exact URL, file, screenshot, receipt, or dashboard item I can inspect. 3. What you tested yourself. 4. What is still not proven. 5. The one next step if the original goal is not fully proven yet. Do not give me confidence. Give me evidence.
An agent can make a page load, a command run, or a button open and still miss the real-world goal. "Done" only matters when the proof matches the outcome you asked for.
If the work is broken right now.
Paste the reply above and ask for the exact proof, not a new status update. If the agent cannot show the live page, file, payment screen, exported result, or automation run that proves the original goal, send the work back.
The paid checklist is for this moment: you are not trying to become the engineer. You are trying to decide whether the evidence is strong enough to accept the work.
What the full checklist adds.
One dollar is cheaper than one fake finish.
The download is a Markdown file, so it opens in notes apps, text editors, docs, and project tools. Pay once, keep it anywhere, and reuse it whenever an agent says "done."