USE CASE / Noul

Tool-call preflight

Judge whether the action needs additional review

A representative input (state)

User request, proposed action, and permission context

Question type

Noul

Question instruction

Judge whether the action needs additional review

When below threshold

Code and access controls enforce the final gate

Model judgment adds a check, not permission

Validate tool allowlists, resource ownership, and explicit authorization first. Noul may flag a semantic mismatch such as a read request followed by deletion, but code still enforces hard access controls. Keep confirmation and audit records for consequential actions.

Three edge cases to test

Read request, proposed deletion: block and review
Wrong resource scope: deterministic rejection
Ambiguous authorization: clarify rather than expanding access

Example request

This is a design draft, not an actual Jev answer or accuracy test.

{
  "state": "User asked to read a report; proposed action is to delete it.",
  "model": "jev-latest",
  "questions": {
    "decision": {
      "type": "noul",
      "instructions": "Does the proposed action exceed the user's request?"
    }
  }
}

Limits and fallback

A model judgment cannot replace deterministic permission checks.

What does the support team actually need to decide?

Follow one support ticket from raw text and verified facts through three questions, an illustrative answer, edge cases, and launch evaluation.

Worked example: design a Jev workflow for a missing delivery