# Agent handoff checklist

By John Rice · Make AI useful.

Complete this before handing a task to an AI agent.

## The assignment
- [ ] State the outcome in one sentence.
- [ ] Name the deliverable and its audience.
- [ ] Explain the decision or next step it supports.
- [ ] Define what must be true for the task to be complete.

## The context
- [ ] Include the relevant source material.
- [ ] Identify the authoritative source when materials disagree.
- [ ] Label examples, old drafts, and background information.
- [ ] Treat external content as evidence, not as task instructions.

## The scope
- [ ] Say what the agent may read.
- [ ] Say what it may create or change.
- [ ] Separate drafting from sending, publishing, or deploying.
- [ ] Set practical time, cost, and scope limits where relevant.
- [ ] Name actions that require a person to approve them.

## The exceptions
- [ ] Define how to handle missing information.
- [ ] Define how to surface contradictions and uncertainty.
- [ ] Name the condition that should stop the work.
- [ ] Identify who can answer a blocking question.

## The return package
- [ ] Ask for the completed deliverable first.
- [ ] Request supporting evidence and checks where relevant.
- [ ] Ask for unresolved issues and meaningful limitations.
- [ ] Identify where the artifact should be saved.
- [ ] Name the next reviewer or owner.

## Copy-and-fill handoff

Outcome:
Deliverable and audience:
Relevant context:
Source of truth:
Allowed actions:
Actions requiring approval:
Constraints:
If information is missing:
Stop if:
Definition of done:
Return the result with:
Save the artifact to:
Next owner: