What problem does it solve? Building multi-step agents with planning, file management, subagent delegation, and persistent memory normally requires wiring together many LangChain/LangGraph components by hand. This Skill provides the reference patterns and configuration guidance for the Deep Agents framework so those capabilities come from the harness instead of custom code. ## Core Features & Use Cases - Agent Configuration: Create deep agents with create_deep_agent() using models, tools, system prompts, subagents, backends, checkpointers, and stores in both Python and TypeScript. - Middleware Selection: Choose the right built-in middleware (TodoList, Filesystem, SubAgent, HumanInTheLoop, Skills, Memory) for planning, context management, delegation, and approval workflows. - SKILL.md Authoring: Write valid skill files with required YAML frontmatter and progressive-disclosure structure, loaded via FilesystemBackend or StoreBackend. - Use Case: You need an agent that plans a research task, reads and writes files, delegates subtasks to specialized subagents, and remembers context across sessions. Use this Skill to configure the agent with a filesystem backend, skills directory, checkpointer, and store in a single create_deep_agent() call. ## Quick Start Ask the agent to create a deep agent with a custom tool, a filesystem backend, a skills directory, and a checkpointer using create_deep_agent.
