# AX-SYS-13: Make consequential actions explicit and reversible

> Tools and capabilities · intermediate · AX-SYS-13

Mark irreversible or costly actions, support dry-runs and idempotency keys, and require confirmation before they execute.

## Why

Missing confirmation on consequential actions is how agents wipe data or overspend. The canonical AX framework (agentexperience.ax — Agent Interactivity Patterns) establishes that sensitive flows — purchases, credential changes, destructive actions — are where digital services need explicit hooks to maintain user trust. Anthropic identifies human-in-the-loop checkpoints for irreversible actions as a core production pattern. The 12-Factor Agents framework calls this Factor 6 ("Design Pause Points") and explicitly labels it "not optional." Cross-links to the human anti-pattern AX-HUM-05 (agentic sludge).

## Do

Flag side effects and offer a dry-run.

## Don't

Let a destructive call run on first inference.

## Artifact

```http
POST /v1/refunds HTTP/1.1
Idempotency-Key: 5f3c...-once-per-intent
X-Dry-Run: true
Content-Type: application/json

{ "order_id": "ord_123", "amount": 4200 }
```

```json
{ "dry_run": true, "would_refund": 4200, "currency": "usd", "reversible": false,
  "confirm_with": { "header": "X-Dry-Run", "set_to": "false" } }
```

## Citations

1. [agentexperience.ax — Principles of AX](https://agentexperience.ax/concepts/principles-of-ax/) — Agent Interactivity Patterns: standardized flows for sensitive interactions — purchases, authenticating users, modifying credentials, destructive actions — are where digital services maintain continuity with end customers
2. [Anthropic — Building Effective AI Agents](https://www.anthropic.com/research/building-effective-agents) — Human-in-the-loop checkpoints for irreversible actions are a core pattern for production agents; the absence of confirmation is how agents cause unintended destructive side effects
3. [12-Factor Agents — Factor 6: Design Pause Points](https://github.com/humanlayer/12-factor-agents) — Factor 6 is not optional: agents need suspension points for human review, especially for irreversible operations — this is the most commonly violated agent design principle

---
Source: Agent Experience Principles (axprinciples.com)