Controlling and Securing OpenAI Agents Execution in JavaScript

Overview

Keep your agents secure, private, and reliable. This course covers safely handling sensitive data, using lifecycle hooks to monitor and customize agent workflows, and applying guardrails to validate and filter all agent inputs and outputs.

Syllabus

• Unit 1: Securely Injecting Sensitive Data into Agents
Securing Sensitive Data with Context
Injecting Context into Agent Runtime
Examining the Secure Conversation Flow
Context Sharing Across Agent Handoffs
Multiple Tools Sharing Secure Context
• Unit 2: Tapping into Agent Workflows with Event Listeners
Building Your First Agent Monitor
Comprehensive Agent Workflow Event Monitoring
Agent Specific Event Listener Implementation
Dynamic Context Injection with Event Listeners
Refactoring Event Listeners for Targeted Monitoring
• Unit 3: Protecting Agents with Input Guardrails
Implementing Guardrail Output Structure
Implementing Input Guardrails for Agent Protection
Building a Content Analysis Agent
Implementing LLM-Based Input Guardrails
Implementing Layered Input Guardrails
• Unit 4: Securing Agent Responses with Output Guardrails
Convert Input Guardrail to Output Guardrail
Converting LLM-Based Input Guardrail to Output Guardrail
Building an Intelligent Output Guardrail for Data Leak Prevention
Attaching Input and Output Guardrails to an Agent

---

Subjects

Computer Science