Run receipt

A run receipt answers the reviewer's first question: "what did you actually do?" It answers with a record instead of a summary. It lists the commands that ran, the checks, their output, and the machine they ran on. And it is attached to the delivery itself, so nobody digs through a log stream.

The line that matters runs between a report and a receipt. A report is the agent's own account of its work, and models write flattering accounts. A receipt comes from the machinery that ran the work. It records failures and skipped steps as faithfully as wins.

For teams reviewing agent output at volume, receipts change the posture. The reviewer no longer re-runs everything to see whether the claims hold. They read what ran, then spend their attention on judgment calls the machinery cannot make.