AARI icon

AARI

AARI Execution Firewall — risk-score and gate actions before they run

Actions2

Gate

AI-generated

Summary

Evaluates an action's risk score against AARI execution firewall policies before running it, routing items to either the Allowed/Warn or Blocked node output.

Inputs

  • Action Type (required) — The category of action being performed (e.g. 'db.delete', 'payment.charge', 'http.post'). Specific types allow accurate risk scoring.
  • Resource (required) — Target identifier for the action (e.g. database table, URL, or file path) used for path-based risk evaluation.
  • Mode — Select 'Enforced' to route high-risk BLOCK decisions to the Blocked output, or 'Advisory' to score and log actions without blocking.
  • Environment — Execution environment ('prod', 'staging', or 'dev'). Production adds a risk score penalty.
  • Agent ID — Identifier for this workflow agent in the AARI dashboard. Defaults to 'n8n-agent'.
  • Parameters (JSON) — JSON object of action parameters evaluated by risk rules (e.g. amount_cents for payment scoring).
  • Run ID — Group ID for workflow execution actions. Auto-generates a UUID if left blank.

Output shape

Dual main output split into 'Allowed / Warn' (Index 0) and 'Blocked' (Index 1).

Original item JSON is preserved and enriched with AARI metadata (aariActionId, aariDecision, aariScore, aariPolicyHits, aariMessage, aariDegraded). Items blocked in enforced mode are sent to output index 1 with an additional aariBlockReason field, and automatically trigger a SKIPPED completion report to AARI.

Examples

Example 1: Enforce high-risk database deletions in production

Action Type: 'db.delete', Resource: 'postgres:prod:users', Mode: 'Enforced', Environment: 'Production'

Example 2: Advisory risk scoring for external HTTP POST calls

Action Type: 'http.post', Resource: 'https://api.example.com/pay', Mode: 'Advisory', Parameters: '{"amount_cents": 5000}'

Links

Discussion