Verdicts

Rules that evaluate prompts and return allow/block/warn.

no-secrets

block

Rejects prompts containing AWS keys, GitHub tokens, or API credentials.

no-pii

warn

Checks for SSN, credit card numbers, and other PII before processing.

budget-check

warn

Estimates prompt cost and blocks if it exceeds configured threshold.

language-filter

block

Allows only configured languages, rejects others.

rate-limit

block

Per-user rate limiting based on API key or session.