Runtime Decision (beforeAction)
AI-generatedSummary
Ask the Vorim API for a live runtime policy decision on whether an agent may perform an action requiring a given scope, and return the API's decision response for each input item.
Inputs
- Agent ID (required) — Vorim agent identifier (e.g. agid_acme_a1b2c3d4) whose action is being evaluated.
- Scope — Required scope for the action: agent:read, write, execute, transact, communicate, delegate, or elevate. Defaults to agent:execute.
- Action Target — Name of the action or tool being gated (e.g. sendEmail). Defaults to an empty string.
Output shape
One output item per input item, containing the decision object returned by POST /v1/runtime/decisions (the response's data field if present, otherwise the full response).
The exact decision fields come from the Vorim API and aren't defined in the node. The node doesn't branch on the result, so use a downstream IF/Switch node to act on it. The request always sends action_type 'tool_call'. Credentials (vorimApi) supply the API key and optional base URL (default https://api.vorim.ai). An API error fails the node unless Continue On Fail is handled, which this code doesn't do explicitly. Items are paired to their inputs. The decision ID can be passed to Emit Audit Event for traceability.
Examples
Example 1: Gate an email-sending step before it runs
Agent ID = agid_acme_a1b2c3d4, Scope = agent:communicate, Action Target = sendEmail; follow with an IF node on the decision result.
Example 2: Check a payment tool call against policy
Agent ID from an expression, Scope = agent:transact, Action Target = createInvoice.