Implementation
Pre-trade protection checklist for AI agents
A deployable checklist for teams giving an autonomous AI agent trading or wallet authority.
Published by Kondux · Updated July 24, 2026
Direct answer
Before an AI agent can trade, define its owner, mandate, permitted assets, contracts, methods, networks, amounts, slippage, counterparties, and expiration. Before every trade, decode the final payload, compare it with intent, simulate the outcome, screen the route and counterparties, and issue an allow, block, or review verdict.
Before granting wallet authority
Start with least privilege. The agent should receive only the authority needed for a defined strategy and time period.
- Record the owner, agent identity, software version, wallet, chain, and approved strategy.
- Allow only required assets, contracts, routers, methods, and destinations.
- Set per-trade, daily, and cumulative value limits.
- Set maximum slippage, fees, gas, price impact, and approval amounts.
- Use expiring, revocable permissions or session keys where supported.
- Define which conditions require human or multi-agent review.
- Test revocation, emergency stop, and recovery before funding the agent.
Before every trade
Inspect the final executable action, not only the agent’s plan or natural-language explanation.
- Bind the proposed trade to the original user or strategy intent.
- Decode contract calls, nested calls, approvals, recipients, and value transfers.
- Verify token identity, router and pool integrity, liquidity, fees, and expected outputs.
- Check amount, slippage, price impact, gas, deadline, and allowance against policy.
- Screen relevant counterparties and addresses using current risk data.
- Simulate the state change and compare it with the expected result.
- Return allow, block, or review with stable reason codes before signing.
After the decision
Preserve the mandate, normalized intent, decoded payload, simulation, risk inputs, policy version, verdict, reviewer or override, and transaction identifier. Monitoring without this context shows what happened but not why it was permitted.
Seraph’s product direction combines these pre-execution checks and decision evidence at the point where an agent hands an action to the wallet or transaction tool.