Duplicate Detection
The duplicate detection guardrail catches repeated emails sent to the same recipients with the same content. This prevents common AI agent failure modes like infinite retry loops, accidental re-sends, and stuck automation workflows.
| Default | Enabled |
| Analysis method | Content hashing compared against recent emails |
| Risk levels | Green / Orange |
How it works
When an email is sent through Sluice, a fingerprint is computed from the email body (text and HTML content) and the sorted recipient list. If the same fingerprint was seen in a previous email within the detection window, the email is flagged as a duplicate.
Configuration
| Setting | Default | Description |
|---|---|---|
| Detection window | 24 hours | How far back to check for duplicates (1 hour to 1 year) |
| Threshold | 1 | Number of matching prior emails before flagging |
Use cases
Preventing agent loops — A bug in your AI agent causes it to repeatedly send the same email every few seconds. Duplicate detection flags the second email and prevents the loop from flooding your customer's inbox.
Retry safety — Your agent retries on transient SMTP errors. If the first attempt actually succeeded, the retry would send a duplicate. Duplicate detection catches it.
Scheduled email workflows — Your automation sends a weekly report. If a workflow bug triggers it twice, the second send is flagged.
Allowing controlled repetition — If your use case legitimately sends similar emails (e.g., weekly newsletters), increase the threshold so the guardrail only flags after multiple duplicates.