What problem does it solve? Building LLM-powered workflows that need parallel execution, human approval gates, or multi-step coordination is hard to get right with plain async code. This Skill provides proven patterns for orchestrating AI agents on Trigger.dev with durable execution, typed batch results, and real-time streaming. ## Core Features & Use Cases - Agent Patterns: Implement prompt chaining, routing, parallelization, orchestrator-workers, and evaluator-optimizer loops using Trigger.dev tasks and the Vercel AI SDK. - Human-in-the-Loop: Pause tasks with waitpoints for Slack approvals, webhook callbacks, or frontend confirmation before continuing execution. - Realtime Streaming: Pipe AI completions and progress updates from tasks to React frontends using typed streams and hooks. - LLM Tool Calling: Convert schemaTask definitions into AI SDK tools so models can autonomously call your tasks. - Use Case: Build a fact-checking pipeline that extracts claims from an article, verifies each claim in parallel with batch.triggerByTaskAndWait, waits for human approval, and streams progress to a dashboard. ## Quick Start Ask the agent to build a Trigger.dev task that runs multiple LLM calls in parallel and aggregates the typed results.
