# Autonomous Agent Transaction Trust Checklist

Use this before allowing an AI agent to spend, hire, delegate, purchase, sign, reserve scarce resources, or commit an organization.

## Authority
- Can the agent prove who the accountable principal is?
- Is the agent actually authorized to perform this action?
- If authority was delegated through other agents, is the chain continuous?
- Did any child agent gain broader scope than its parent?
- Are spending limits explicit?
- Are counterparty restrictions explicit?
- Does authority expire?

## Transaction
- Is the exact action inside the mandate?
- Is the amount inside the inherited ceiling?
- Is the currency permitted?
- Is the counterparty permitted?
- Is the transaction inside the valid time window?

## Evidence
- What evidence did the agent rely on?
- Is that evidence independently retrievable or cryptographically fingerprinted?
- Is additional proof required above certain risk levels?

## Human control
- At what threshold must a human approve?
- Can a sub-agent bypass that threshold? It should not.
- Is the approval reference preserved?

## Recourse
- Above what exposure is recourse required?
- Does available coverage match the full transaction amount?
- What happens if the counterparty fails?

## Permanent record
- Can you prove which policy version made the decision?
- Can you prove which authority chain was evaluated?
- Can you reproduce the transaction record after today's model, wallet, cloud or blockchain is gone?

## AgentProof implementation
AgentProof Autonomous Trust Resolution v0.2 answers these questions before action. Agent Transaction Passport™ preserves the transaction afterward.

Free sandbox: https://agentproof.hatchable.site/sandbox/
Production resolver: https://agentproof.hatchable.site/api/paid/resolve_transaction
Enterprise: https://agentproof.hatchable.site/enterprise/