SecureVector icon

SecureVector

AI security: scan prompts and LLM outputs, gate tool calls, track LLM cost and budget. Cloud by default; Local App transport unlocks tool audit, costs, and device ID.

Actions2

Prompt → Scan Prompt

AI-generated

Summary

Scans incoming text for AI security threats using SecureVector’s cloud API or local app, then returns a normalized result for downstream workflow steps. Optionally stops the workflow when selected threat conditions are met.

Inputs

  • Transport — Choose Cloud (default; requires a SecureVector API key) or Local App (requires the desktop app running on loopback; no API key).
  • Local Base URL — In Local App mode, the app URL; it must use http://127.0.0.1 or http://localhost. Defaults to http://127.0.0.1:8741.
  • Prompt to Scan — Text to scan. Defaults to the first available input field among chatInput, prompt, message, content, text, title, input, and output; you can enter an expression or text directly. The runtime requires non-empty text and truncates it to 10,000 characters.
  • Additional Fields — Timeout in seconds (1–300, default 30) and whether to include workflow and execution IDs as metadata (default off).
  • Blocking Options — Block on Threat is off by default. When enabled, choose any combination of API BLOCK verdict, score above the threshold (default 50), and selected risk levels (default critical and high). If a selected condition matches, the node raises an error and stops the workflow unless Continue On Fail is enabled.

Output shape

A single JSON item per input item containing the normalized scan result.

Includes verdict, score, threat_score, riskLevel, threat_level, confidence_score, threats, recommendation, and analysis. Cloud and local responses are mapped to this shape; local scans set recommendation to null. With Continue On Fail enabled, errors are returned as {"error": "..."} items.

Links

Discussion