trigger-agents

AI agent patterns with Trigger.dev - orchestration, parallelization, routing, evaluator-optimizer, and human-in-the-loop. Use when building LLM-powered tasks that need parallel workers, approval gates, tool calling, or multi-step agent workflows.

27stars5forksUpdated 6/16/2026

Web Development #orchestration #trigger-dev #vercel-ai-sdk #ai-agents #llm #typescript

Security Assessment

Medium Risk(60/100)

Detected risks:

Secret Exposure([references/waitpoints.md] token =, [references/waitpoints.md] Token =, [references/waitpoints.md] token=)

Sensitive File Access([references/waitpoints.md] .env)

Security Score60/100

About trigger-agents

The trigger-agents skill provides production-ready AI agent patterns built on Trigger.dev's durable execution framework. It solves the challenge of building reliable, scalable LLM-powered workflows that require complex orchestration, parallel processing, conditional routing, and quality validation gates. Instead of writing custom coordination logic, developers can leverage proven patterns for common agent architectures.

This skill includes five core patterns: Prompt Chaining for sequential LLM calls with validation gates, Routing for intelligent model selection based on task complexity, Parallelization for concurrent processing of independent tasks, Orchestrator-Workers for coordinating multiple specialized agents, and Evaluator-Optimizer for iterative self-improvement until quality thresholds are met. Each pattern integrates with the Vercel AI SDK and supports streaming, human-in-the-loop approval gates, and tool calling capabilities.

Ideal for developers building AI-powered applications that need reliability beyond simple API calls. Use cases include content generation pipelines with quality checks, customer support systems that route to appropriate models, data processing workflows that run analyses in parallel, multi-agent systems coordinating specialized tasks, and any LLM application requiring human approval steps or iterative refinement. The patterns work with major LLM providers through the AI SDK and leverage Trigger.dev's infrastructure for retries, monitoring, and long-running workflows.

FAQ

Which LLM providers are supported?

The skill works with any provider supported by the Vercel AI SDK, including OpenAI, Anthropic, Google, and others. Examples in the documentation use OpenAI's models but can be adapted to any provider.

When should I use Parallelization vs Orchestrator-Workers?

Use Parallelization when you have independent tasks that can run simultaneously and you need all results (like analyzing sentiment, summarizing, and moderating content in parallel). Use Orchestrator-Workers when you have a complex workflow where a coordinator agent needs to dynamically assign subtasks to specialized worker agents.

Can I combine multiple patterns in one workflow?

Yes, patterns are designed to compose together. For example, you can use Routing to select a model, then Prompt Chaining with validation gates, and include Human-in-the-Loop waitpoints for approval before final steps.

What makes these patterns 'production-ready'?

They leverage Trigger.dev's durable execution which provides automatic retries, error handling, long-running workflow support, monitoring, and the ability to pause/resume for human approval. This eliminates common failure modes in LLM applications.

Do I need to know Trigger.dev to use these patterns?

Basic familiarity with Trigger.dev's task API is helpful, but the patterns are self-contained examples you can adapt. The skill shows how to use batch operations, task definitions, and the integration with the AI SDK.