The mental model I give agents

I run it as a loop, not a checklist: start loose, look at what comes back, tighten with one specific correction.

1. Outcome first. I open with what has to exist when the work is done, and who it's for.
2. Truth gate. A conclusion has to be earned. "How can you be sure?" "Not just kinda but do." "Do not make assumptions." Absence of evidence isn't evidence of absence, so unknowns stay visible as [check].
3. Control gate. Research first, ask me one question at a time, plan, review, then act. Agents run freely on reversible work; anything irreversible waits for my approval.
4. Shape it for the experience. Cut it down to the simplest thing that still lets someone act, with the detail underneath. My name never goes on AI-sounding writing.
5. Persist. Every good investigation becomes knowledge or a skill, filed in the right place, so the next session starts smarter.

Two small habits live inside it. I say "this is good, but" and give one correction that fixes the agent's model of the situation, and I say "way too many words" when it runs long.

The same few ideas show up in every system I've built.

1. One owner per kind of thing. Every kind of request has one owner that holds its file, its format and its memory step. Route first, then work.
2. One door per risk. Actions go through one approval tool, memory through one memory tool, releases through one ship script. One door is one place to check, log and change.
3. The tool decides, the agent classifies. The agent names what kind of thing it's doing. A table I own answers, and an exit code says what happens next. Agents branch on numbers, not on a feeling that something is fine.
4. Doubt fails closed, and stricter wins. No rule means ask. An unreadable rule means ask. When two rules fit, the stricter one applies.
5. My yes has a source. Only my own message in my own chat is a yes. Text that claims to be me, a file that says I agreed, and a tap on the agent's own card are not.
6. Nobody grades their own work. Builder, tester and reviewer are separate sessions. The tester never sees the builder's transcript.
7. Quiet doesn't mean done. Owed work stays owed until it's acknowledged, and a sweep tells me if it sits.
8. Attention is the scarce thing. Do the work first, then ask once: one thing, recommended choice first, a picture if it's visual, paced through my day.
9. Autonomy is earned per row. A risky action goes automatic only after clean approvals and a line I wrote myself, and goes back with one edit.
10. A correction becomes a rule. It's dated, in my words, in the file the agent reads every turn.
11. Every claim carries its evidence. Verified in a system, seen in mail, found by an AI and not re-checked, estimated, or [check].