engineer-agents

What problem does it solve?

Creating, configuring, and operating Braze AI agents involves many moving parts—prompt and schema design, Liquid bindings, model and thinking-level selection, Canvas vs. catalog deployment differences, and production monitoring. This Skill consolidates guidance across agent creation, deployment, and runtime troubleshooting so engineers can avoid mismatches, runtime failures, and costly misconfigurations.

Core Features & Use Cases

• Agent design & prompts: Advice on instruction structure, example-driven prompts, and output schema alignment to reduce parsing errors.
• Deployment patterns: Best practices for Canvas step agents (real-time, per-user) and catalog field agents (batch, per-row), including input size and circular-reference constraints.
• Production readiness: Guidance on fallback values, rate limits, logging, monitoring, and rollback strategies to keep agents safe in production.
• Use Case: Create a Canvas agent that generates personalized email subject lines at send time while ensuring output schema matches Canvas routing rules and providing fallback text for timeouts.

Quick Start

Use the engineer-agents skill to create and test a Canvas agent that generates personalized subject lines for a high-value user segment and deploy it with a defined output schema and fallback.