# AI handoff prompt

Use this prompt when a current AI agent or chat needs to prepare durable context for a fresh agent, model, session, or human reviewer.

```text
Create a continuation record for the next AI agent or person.

Return Markdown with exactly these sections:
1. Objective and acceptance criteria
2. Current state, separated into completed, in progress, blocked, and unverified
3. Decisions and the rationale for each
4. Evidence, including relevant files, URLs, commands, outputs, and source locations
5. Constraints, including permissions, safety boundaries, deadlines, budgets, and exclusions
6. One bounded next action with an observable expected result
7. Ownership
8. Receiver verification checklist

For every material claim, label it as verified now, from conversation context, or unverified. Re-read authoritative files and current system state before calling anything verified. Preserve exact identifiers and commands when they matter. Do not include credentials, tokens, secrets, private keys, or values that belong in a secret manager. Keep the record compact enough for a fresh agent to inspect before acting.
```

## Receiver prompt

```text
Continue from the supplied handoff record.

Before changing anything:
1. Verify the active identity, workspace, permissions, and current revision.
2. Inspect the cited files and evidence instead of trusting the summary alone.
3. Separate verified facts from conversation claims and unresolved assumptions.
4. Restate the objective, constraints, and one next action.
5. Stop and report the exact blocker if required access or evidence is missing.

After verification, perform the smallest concrete next action and report its observable result. Do not repeat completed work, broaden scope, or expose secrets.
```

Interactive generator: https://handover.sh/tools/ai-handoff-prompt-generator

Open Handoff Continuity Record format: https://handover.sh/protocol

Model-neutral Markdown template: https://handover.sh/templates/agent-handoff