Now liveThe Skillselion MCP - thousands of ranked skills, loaded into your agent mid-task. No install.Get it →
jhleee avatar

Jhleee Layered Agent Architecture

  • Updated February 8, 2026
  • jhleee/layered-agent-architecture

jhleee-layered-agent-architecture is a Claude Code plugin that scaffolds seven-layer LangGraph-style agent projects with subgraphs, tools, and architecture linting.

About

jhleee-layered-agent-architecture is a community Claude Code plugin bundle focused on structuring multi-step AI agents with a disciplined layered model rather than a single monolithic prompt file. Developers shipping LangGraph or similar graph-based agents use it to scaffold projects, separate subgraphs from parent graphs, and keep tools, prompts, and nodes in predictable layers. The catalog metadata highlights architecture, dependency control, and linting—signals that the skill pushes you toward maintainable agent repos you can extend without entangling every node. It fits Developers who already chose an agent-shaped product and need a repeatable layout before integration work piles up. It is less about LLM API wiring alone and more about how the agent project is organized on disk and in the graph. Register the plugin in Claude Code, then invoke it when starting a new agent repo or refactoring a tangled graph into subgraphs with clear boundaries.

  • Seven-layer agent architecture pattern with explicit dependency boundaries between prompts, tools, nodes, and graphs
  • LangGraph-oriented scaffolding for StateGraph and subgraphs so solo builders split orchestration from leaf capabilities
  • Project scaffolding and versioning hooks aligned with long-lived agent codebases
  • Linting and dependency rules to catch circular tool or node imports early
  • Keywords span architect, layered, nodes, skill, and stategraph for discovery in agent-heavy repos

Jhleee Layered Agent Architecture by the numbers

  • Data as of Jul 7, 2026 (Skillselion catalog sync)
/plugin install jhleee-layered-agent-architecture@jhleee/layered-agent-architecture

Add your badge

Show developers this plugin is listed on Skillselion. Paste this into your README.

Listed on Skillselion
Last updatedFebruary 8, 2026
Repositoryjhleee/layered-agent-architecture

What it does

Install this Claude Code plugin when you want a seven-layer agent architecture scaffold with LangGraph state graphs, subgraphs, tool nodes, and dependency linting instead of ad-hoc agent folders.

Who is it for?

Best when you're starting or refactoring a LangGraph agent and want an opinionated folder and graph structure before adding more tools.

Skip if: Skip if you only need a single-shot prompt skill with no graph, or teams that already enforce your own internal agent framework.

What you get

After install, you get a layered project pattern and guardrails so new LangGraph nodes and subgraphs land in consistent folders with dependency checks before you scale features.

  • Layered agent directory layout aligned to the seven-layer pattern
  • Clear separation of prompts, tools, nodes, and parent or child graphs
  • Dependency and structure guidance suitable for iterative agent expansion

Recommended Plugins

How it compares

Agent architecture scaffolding plugin, not a generic refactor or leadership workflow pack.

FAQ

Who is Jhleee Layered Agent Architecture for?

It is for developers building layered Claude or LangGraph agents who need scaffolding, subgraph boundaries, and dependency discipline in one plugin.

When should I use Jhleee Layered Agent Architecture?

Use it at the start of a new agent repo or when an existing StateGraph has grown messy and you want a seven-layer layout and linting-oriented structure.

How do I add Jhleee Layered Agent Architecture to my agent?

Add the jhleee/layered-agent-architecture marketplace entry to Claude Code, install the bundle, then invoke its architecture skill from your project when scaffolding or restructuring graphs.

AI Agentsagentsautomation

This week in AI coding

Five minutes, every Monday - the tools, releases and tactics for developers.

unsubscribe anytime.