Browse AI Agents & MCP SECTION 8: Safety, Guardrails & Human-in-the-Loop

The Approval Gate Designer

5 fill-in slots · from The AI Prompt Handbook for AI Agents & MCP

Your details

0/5 filled

The Approval Gate Designer

Design approval gates for this agent.

Agent: [DESCRIBE]
Actions with consequences: [LIST]
Who approves: [DESCRIBE THE ROLE AND THEIR CONTEXT]
Volume: [HOW MANY APPROVALS PER DAY]

For each action requiring approval:
1. What triggers the gate — a rule, not a judgement call, where
   possible
2. What the approver sees: what the agent proposes, why, what it
   is based on, what happens if approved, what happens if not
3. Options available: approve, modify, reject, ask for more
4. What the agent does with a rejection — the rejection reason
   should be usable
5. Timeout behaviour
6. What is logged

Then address the real problem:
7. Approval fatigue. At [VOLUME] approvals per day, will the
   approver actually read them? What proportion get rubber-stamped?
8. Which gates could be replaced by a constraint that makes the
   dangerous action impossible instead of merely gated?
9. Which gates could be batched, sampled, or made post-hoc rather
   than blocking?
10. What is the right number of gates for this system, and am I
    over or under?

A gate that is always approved is not a control. Tell me which of
mine are theatre.

Highlighted [slots] are still empty. Downloads are plain .txt — paste into any AI assistant.