How to Stop an AI Agent From Emailing 400 Customers at 3AM: Budget Caps, Idempotency & a Kill Switch in n8n

说明与用途

在 n8n 中实现保护模式，包括幂等键、月度预算、速率限制和杀死开关，防止了 400 封错误邮件的批量发送并控制了 LLM 成本。

问题

An unattended AI agent sent 400 incorrect emails at 3 AM, causing operational chaos.

结果

The guard pattern stopped unintended mass emailing, enforced budget limits, and added a kill switch, reducing operational risk.

工作流

• 1. Generate SHA‑256 idempotency key from client, action type, recipient, payload.
• 2. Insert key into guard_actions ledger in Postgres.
• 3. Query ledger for monthly cost per client and compare to cap.
• 4. Count actions in last hour and compare to rate cap.
• 5. If any check fails, send Slack alert, stop workflow with Error node; otherwise proceed with LLM call and outbound action.

证据

put one guard step in front of every LLM call and every outbound action

来源

SHA-256 of client, action type, recipient, payload

来源

SUM(cost_usd) by client and month vs cap

来源

COUNT(*) of actions in last hour vs cap

来源

Slack alert, then Stop and Error node

来源