The Agent Skills Directory

[SAFE]: The skill serves as an authoritative guide for building n8n AI agents, focusing on production-grade patterns such as the shell-and-core architecture and human-in-the-loop (HITL) safety gates.
[PROMPT_INJECTION]: The skill documents the attack surface for indirect prompt injection inherent in AI agent development and provides explicit remediation strategies.
Ingestion points: Processes untrusted data through the n8n Chat Trigger (chatInput) and Webhook nodes.
Boundary markers: Recommends using modular prompt fragments and strict JSON schemas to delimit agent instructions from data.
Capability inventory: The documented agents can perform network operations via HTTP Request tools, execute logic via sub-workflows, and mutate external state in services like Slack, Notion, and Stripe.
Sanitization: Recommends the use of outputParserStructured with autoFix to validate model outputs and the HitlTool suite to enforce human approval for sensitive operations.

n8n-agents