In short
The rule that says what an agent may say when it is done: one line if something needs a human, nothing at all otherwise.
An agent that reports every step is noise. An agent that reports nothing is a risk. The output contract sits between: do the work silently, then say exactly one line if a human needs to act, and say nothing if not.
It is written at the top of every scheduled job. It keeps a phone quiet through a hundred successful runs and makes the one failure impossible to miss.
The contract also governs voice. The line that does get sent is delivered word for word, as the agent, so it has to be written the way the company speaks.
Example
What it looks like.
output-contract.mdplain text
OUTPUT CONTRACT: do all the work first, without narrating. If everything published cleanly, reply exactly [SILENT]. Reply with one line only if something failed or needs a decision.Related
Read next.
- instruction fileThe Markdown file that tells an agent who it is, what it owns, and what it may never do. The agent's job description, readable by the agent.Read the definition
- doctrineThe set of written rules every agent in a company follows. Voice, safety, reporting, git. Written once, applied everywhere.Read the definition
- cronA schedule. The thing that runs an agent at 10:15 every morning whether anyone remembers or not.Read the definition