Guardrail templates Preview
The starting points offered when you create a guardrail rule or a guardrail profile, and what each one turns on.
After reading this page you will be able to:
- Pick the right template for a risk
- Know what to adjust after starting from a template
Preview
These are the templates offered by + Create guardrail rule on Guardrail rules. Guardrail profiles offers its own set when you create a profile: Blank, Content safety, PII protection, Prompt attack defense and ITSM safe-reply.
| Template | Turns on | Typical adjustments |
|---|---|---|
| Start blank | Nothing | Build your own |
| PII protection | Redact PII in both directions, masking emails, phone numbers, cards, IBANs, national IDs and tokens | Choose Mask, Tokenise or Block per type |
| Prompt attack defense | Block prompt injection on input | Monitor first, then tune for your tools |
| Secrets and credentials | Custom patterns in both directions for cloud keys, private keys and platform tokens: masked, with private keys blocked | Block more of them |
| Profanity | Word filter with the managed profanity list, masking replies | Add your own words |
| Content safety | Harm categories scored in both directions, starting in monitor mode. Self-harm and harassment escalate to a person, and the rest block | Tune each category's threshold on live scores |
| Confidential HR topics | Deny topics: Salary and pay, Redundancy and job cuts, HR case (an HR case is flagged for a person on the way in) | Escalate to HR instead of blocking |
Tip
Templates are only a starting point. Everything they set can be changed before and after you save.