The wrong file to the wrong person

A test we ran on purpose, on Gmail with demo accounts and an agent called proposals-agent. On the first run it did its job and still got it wrong. On the second run, the same mistake was held for a person.

Protect my first agentWatch your agent’s calls live, no signup

  1. Day three, it emailed our pricing to a client

    We gave an agent a sandbox Drive and inbox on purpose, with one instruction: send the proposal to the client. For two days it ran that job fine.

    On the third day it picked the file with the nearest name, the one with our margins in it, and sent it to the contact with the nearest name. The right file and the wrong file were one word apart. So were the two contacts.

    What got out on the first run. Illustration with example data.

  2. It believed it had double-checked

    It read the brief, found a proposal, found a client and sent, which is exactly what it’s for. We’d written “double-check the recipient” into the prompt. It believed it had.

  3. Second run: the same email, stopped in the queue

    Same wrong file, same wrong client, because nothing about the agent had changed. This time the email stopped in the queue.

    The approval screen showed the recipient, the attachment name and the agent that asked. The pricing file was sitting next to the client’s name, so it was rejected. Thirty seconds, and nothing left the building.

    The send that didn’t leave.Illustration with example data.

  4. What caught it: every send waits for a person

    Sending email is an approval rule. Each send stops in a queue that shows the agent that asked, the recipient and the attachment name. You approve or reject it, and your answer goes on the record.

    Approval rules work on every plan, with a passkey or an email link. Push notifications to your phone are on Team and above.

What did your agent send last week that nobody read first?

If you don’t know, start by watching. Observe Mode shows every call one agent makes, live, with no signup. It records and never blocks, so nothing about your setup changes.

Watch your agent’s calls live, no signupProtect my first agent

The other tests

Every one ran on demo accounts, and every agent was doing its job.

Take the keys back this afternoon.

Put one agent through AgentValet on the free plan and watch what it does for a day. If the approvals and the record don’t satisfy you, turn it off. Nothing else about your setup changes.

Protect my first agentWatch your agent’s calls live, no signup

npx @agentvalet/register