agent-orchestration-patterns

Multi-agent workflows execute sequentially by chaining agents through an orchestrator that passes each agent's output as input to the next. This Skill provides orchestration patterns that standardize tool schemas with Pydantic and manage agent state across turns, ensuring reliable handoffs between research, synthesis, and writing agents or similar multi-step processes.

Pydantic-based tool schemas enforce strict input and output validation, preventing malformed data from reaching external systems. This Skill standardizes tool interface design so agents reliably interact with APIs, databases, and services, reducing runtime errors in tool-calling pipelines built with LangChain, LlamaIndex, or Anthropic tools.

Agent state management patterns preserve conversation history and context across interaction turns, enabling coherent multi-step reasoning. This Skill provides structured state management using Pydantic to ensure agents recall prior decisions and information, supporting complex workflows where later agents depend on earlier results.

Yes. This Skill implements comprehensive error handling, timeouts, and logging within tool execution, making agent interactions with external systems reliable and debuggable. It covers sequential, parallel, router-based, and hierarchical orchestration patterns so you can choose the right execution model for your workflow.

Sequential orchestration chains agents linearly, passing outputs forward; hierarchical orchestration uses a parent agent to delegate subtasks to child agents and aggregate results. This Skill covers both patterns plus parallel and router-based approaches, helping you select the orchestration strategy that matches your workflow complexity and dependencies.

Pydantic is not strictly required, but this Skill enforces it to prevent tool interface inconsistencies and validation failures at scale. Unstructured data leads to unpredictable agent behavior; Pydantic schemas ensure every tool call is validated before execution, reducing debugging and making orchestration patterns reliable across agent frameworks.