IndyKite Intent Agent - from a prompt to AuthZEN evaluations

The Intent Agent turns a sentence into an authorization question. Given "Can I retrieve the quote stock_quote?" and the user it is asked for, it answers with what the sentence asks, in the shape the AuthZEN API decides on:

{
  "subject": { "type": "User", "id": "roy" },
  "action": { "name": "CAN_RETRIEVE" },
  "resource": { "type": "Quote", "id": "stock_quote" },
  "context": { "parameters": {} }
}

It is a self-hosted A2A service, shipped as the indykite/intent-agent image, and it speaks only the vocabulary of your project: node types from the data schema and actions that ACTIVE KBAC policies grant. It does not decide anything - the decision stays with AuthZEN - and it never takes the subject from the prompt: every evaluation about "I" or "me" gets the subject_type / subject_id the caller supplied, so prompt text cannot speak for someone else. Unclear, out-of-scope, and manipulative prompts are refused with an explanation the caller can show to the user.

When to use

Activate this skill when the user wants to: