Hardened support agent

★ 7.3 · n8n · medium · 8 nodes

Hardened support agent is an n8n automation template that pairs two AI support agents — a hardened version and a deliberately unsafe one — alongside a regression-testing scenario to compare their behaviour under adversarial inputs. Reach for it when you need to verify that a support chatbot resists prompt injection before going live, when your team updates an agent's system prompt and wants a reproducible way to catch regressions, or when you're auditing an existing AI agent for security gaps. The template stores session or conversation data in Postgres and is triggered via webhook: an incoming request routes to whichever agent variant you're testing. It imports directly into self-hosted or cloud n8n as a standard workflow. Trigger: Webhook. Integrations: Postgres. The workflow is about stress-testing agent defences, not running a production support desk — for live customer-facing flows you'd need to extract and harden the relevant agent branch separately. Use only on systems you own or are authorised to test. Best suited for teams building or reviewing AI agents; not a drop-in replacement for a fully configured production support agent.