# My first useful loop

## Six lines
- User and value: [who outside the team; what result they need]
- Trigger and routine: [when it runs; input -> make -> check -> deliver]
- Check: [specific pass/fail conditions]
- Human control: [allowed actions, budget, retry limit, stop/ask path]
- Feedback: [what to measure; one change to test]
- First test: [five reachable users; what will count as useful]

## Build brief for an AI coding agent
Goal: [one small useful outcome].
Context: [real examples, approved sources, current project].
Constraints: [budget, permitted data/actions, delivery channel, human approvals].
Done when: one input produces a checked result, failures stop safely, and the saved test cases pass.

Build one complete slice first. Show the working result and test evidence. Mark assumptions. Never claim an untested system is ready. Ask before exceeding the agreed permissions or budget.

## First five users
[Keep private names and contact details in your team's private workspace, not in a public repository.]

## Next smallest step
[What you will test today]