skills.rest
AhmedElbialy148
AhmedElbialy148COMMUNITY
Explore all skills by this author →
GitHub

mis-langgraph

Standardizes building, structuring, and reviewing LangGraph Python agents in MIS repositories.

Updated Oct 7, 2023
Search Tags:#testing#python#ai agents#langgraph#human-in-the-loop#deepagents#graph architecture
One-click install
npx skills add https://github.com/AhmedElbialy148/Portfolio --skill mis-langgraph-ahmedelbialy148
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: mis-langgraph
Source: https://github.com/AhmedElbialy148/Portfolio/tree/main/.cursor/skills/mis-langgraph
Command: npx skills add https://github.com/AhmedElbialy148/Portfolio --skill mis-langgraph-ahmedelbialy148

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires langgraph, langchain-core, deepagents, pydantic, and includes references (resource) and assets (resource) components.

What problem does it solve? Building LangGraph agents without a shared standard leads to inconsistent project layouts, untyped state, routing logic buried in nodes, and unguarded side-effecting tools. This Skill enforces the MIS contract for every LangGraph repo so agents are structured, testable, and safe to deploy. ## Core Features & Use Cases - Canonical project layout: Two-tier structure with shared services/ packages and per-agent packages containing nodes/, prompts/, schemas/, tools/, and optional eval/ folders, plus drop-in starter templates. - Graph architecture guidance: Typed state with reducers, pure node functions, dedicated routing functions, checkpointer selection, and wiring centralized in graph_builder.py. - Deep Agents harness patterns: When and how to adopt write_todos, task subagents, virtual filesystem, deny-by-default permissions, interrupt_on HITL gates, memory, and skills. - Use Case: When asked to add a new email-assistant agent to an MIS repo, scaffold the package from templates, register it in langgraph.json, wrap send_email with approval, and add unit plus integration tests following the standard. ## Quick Start Ask the agent to bootstrap a new LangGraph repo or add a new agent following the MIS LangGraph standard, and it will scaffold the layout, wire the graph, and apply the review checklist.

Frequently Asked Questions about mis-langgraph

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I structure a LangGraph project with multiple agents?▼

Use a two-tier layout: shared integrations live in src/services/<service>/ while each agent is a self-contained package under src/<agent_name>/ with nodes/, prompts/, schemas/, tools/, and optional eval/ subfolders. Each agent exports one compiled graph registered in langgraph.json.

When should I use the Deep Agents harness instead of a vanilla StateGraph?▼

Adopt the deepagents harness when two or more apply: long multi-turn runs, subtasks needing isolated context, durable artefacts in a virtual filesystem, declarative permissions, inspectable planning, or cross-session memory. If only one applies, a vanilla ReAct loop over a small StateGraph is simpler.

How do I add human approval to LangGraph tools?▼

Wrap side-effecting tools with add_approval at assembly time in tools/base.py for vanilla graphs, or use interrupt_on={"tool_name": True} when building a deep agent. Keep approval off only in dev mode, which is a runtime flag rather than a code branch.

What is the difference between services and tools in LangGraph agents?▼

Services are reusable API clients in src/services/ that know nothing about LangGraph and handle auth, retries, and parsing. Tools are @tool-decorated functions in each agent's tools/ folder that wrap service calls and expose them to the LLM with descriptive docstrings.

How should I test a LangGraph graph?▼

Use three layers: unit tests in tests/unit/ for nodes, tools, and routing with mocked services; integration tests in tests/integration/ running the compiled graph with LLM and network boundaries mocked; and notebooks in tests/notebooks/ for debugging only, never as CI gates.

What Python version and packages does this standard require?▼

It requires Python 3.11 or later, langgraph >= 0.2, and langchain-core. The Deep Agents sections additionally require the deepagents package, and the recommended dev stack includes pytest, pytest-asyncio, ruff, and mypy.

Related Skills

Explore More Software Engineering Skills→
wrightchinwrightchin

add-slack

Configure Slack channel integration for NanoClaw via the Chat SDK.

Community
Intermediate
christian-bromannchristian-bromann
3

langgraph-graph-api

Execute compiled LangGraph graphs with invoke, stream, and batch patterns.

Community
Advanced
artgaurav16420-ossartgaurav16420-oss

langsmith-fetch

Fetch and analyze recent LangSmith traces to debug LangChain and LangGraph agents.

Community
Intermediate
marvelousempiremarvelousempire

agent-orchestration-frameworks

Guide framework selection via decision trees for agent orchestration projects.

Community
Advanced