Table of Contents

Domain intent templates

Why you're reading this page: This page provides intent names and actor:action (signal) sets for fraud/security and customer intent. It is the right place if you are asking "How do I define intents for my domain?" in a repeatable way.

Templates for defining intent names and actor:action (signal) sets for two domains beyond greenwashing: fraud / security and customer intent. Use these to answer "How do I define intents for my domain?" in a repeatable way.


Fraud / security

Intent names (suggested): SuspiciousAccess, CredentialStuffing, AccountRecovery, NormalLogin, HighRiskSession.

Suggested actor:action (signals):

Actor Action (examples) Meaning
user login.failed Failed login attempt
user login.retry Retry after failure
user ip.changed IP change in session
user password.reset Password reset flow
user login.success Successful login
user device.verified Device verification
user captcha.passed CAPTCHA solved
system velocity.high Many requests in short time

Policy idea: Block when confidence high and signal count ≥ N (e.g. many failed logins + IP change); Observe when medium risk; Allow when low. See examples/fraud-intent.


Customer intent

Intent names (suggested): PurchaseIntent, InfoGathering, SupportRequest, BrowsingOnly.

Suggested actor:action (signals):

Actor Action (examples) Meaning
user browse.category Category browse
user cart.add Add to cart
user checkout.start Start checkout
user payment.submit Submit payment
user search.product Product search
user view.product View product
user compare.product Compare products
user view.faq View FAQ
user contact.click Click contact
user ticket.create Create support ticket
user chat.start Start chat

Policy idea: Allow high confidence; Observe medium; route by intent (purchase → checkout, support → human/chat). See examples/customer-intent.


How to use

  1. Choose a domain (fraud, customer, or your own).
  2. Define intent names and map them to confidence thresholds or rules.
  3. Record events with space.Observe(actor, action) using the suggested actions (or your own).
  4. Run your intent model (e.g. LlmIntentModel or a rule-based model) and policy; use the intent name + confidence for routing or blocking.

For greenwashing, see Greenwashing detection how-to and Greenwashing metrics.

Next step: When you're done with this page → Designing intent models or Scenarios.