langgraph

Name: langgraph
Author: hanamizuki

104stars8forks

How to install langgraph?

hanamizuki/solopreneur/langgraph

$npx -y skills add hanamizuki/solopreneur --skill langgraph

Installs into the current project.

Summary

Understand LangGraph v1.0 thoroughly

Files · 1

View on GitHub

SKILL.md

1	# LangGraph Development Principles
2
3	If you are coding with LangGraph, follow these principles and patterns.
4
5	## Critical Structure Requirements
6
7	### MANDATORY FIRST STEP
8	Before creating any files, always search the codebase for existing LangGraph-related files:
9	- Files with names like: `graph.py`, `main.py`, `app.py`, `agent.py`, `workflow.py`
10	- Files containing: `.compile()`, `StateGraph`, `create_react_agent`, `app =`, graph exports
11	- Any existing LangGraph imports or patterns
12
13	If any LangGraph files exist: Follow the existing structure exactly. Do not create new agent.py files.
14
15	Only create agent.py when: Building from completely empty directory with zero existing LangGraph files.
16
17	- When starting from scratch, ensure all of the following:
18	1. `agent.py` at project root with compiled graph exported as `app`
19	2. `langgraph.json` configuration file in the same directory as the graph
20	3. Proper state management defined with `TypedDict` or Pydantic `BaseModel`
21	4. Test small components before building complex graphs
22
23	## Deployment-First Principles
24
25	CRITICAL: All LangGraph agents should be written for DEPLOYMENT unless otherwise specified.
26
27	### Core Requirements:
28	- NEVER ADD A CHECKPOINTER unless explicitly requested by user
29	- Always export compiled graph as `app`
30	- Use prebuilt components when possible
31	- Follow model preference hierarchy: Anthropic > OpenAI > Google
32	- Keep state minimal (MessagesState usually sufficient)
33
34	#### AVOID unless user specifically requests
35	```python
36	# Don't do this unless asked!
37	from langgraph.checkpoint.memory import MemorySaver
38	graph = create_react_agent(model, tools, checkpointer=MemorySaver())
39	```
40
41	#### For existing codebases
42	- Always search for existing graph export patterns first
43	- Work within the established structure rather than imposing new patterns
44	- Do not create `agent.py` if graphs are already exported elsewhere
45
46	### Standard Structure for New Projects:
47	```
48	./agent.py # Main agent file, exports: app
49	./langgraph.json # LangGraph configuration
50	```
51
52	### Export Pattern:
53	```python
54	from langgraph.graph import StateGraph, START, END
55	# ... your state and node definitions ...
56
57	# Build your graph
58	graph_builder = StateGraph(YourState)
59	# ... add nodes and edges ...
60
61	# Export as 'app' for new agents from scratch
62	graph = graph_builder.compile()
63	app = graph # Required for new LangGraph agents
64	```
65
66	## Prefer Prebuilt Components
67
68	Always use prebuilt components when possible - they are deployment-ready and well-tested.
69
70	### Basic Agents - Use create_react_agent:
71	```python
72	from langgraph.prebuilt import create_react_agent
73
74	# Simple, deployment-ready agent
75	graph = create_react_agent(
76	model=model,
77	tools=tools,
78	prompt="Your agent instructions here"
79	)
80	app = graph
81	```
82
83	### Multi-Agent Systems:
84
85	#### Supervisor Pattern (central coordination):
86	```python
87	from langgraph_supervisor import create_supervisor
88
89	supervisor = create_supervisor(
90	agents=[agent1, agent2],
91	model=model,
92	prompt="You coordinate between agents..."
93	)
94	app = supervisor.compile()
95	```
96	Documentation: https://langchain-ai.github.io/langgraph/reference/supervisor/
97
98	#### Swarm Pattern (dynamic handoffs):
99	```python
100	from langgraph_swarm import create_swarm, create_handoff_tool
101
102	alice = create_react_agent(
103	model,
104	[tools, create_handoff_tool(agent_name="Bob")],
105	prompt="You are Alice.",
106	name="Alice",
107	)
108
109	workflow = create_swarm([alice, bob], default_active_agent="Alice")
110	app = workflow.compile()
111	```
112	Documentation: https://langchain-ai.github.io/langgraph/reference/swarm/
113
114	### Only Build Custom StateGraph When:
115	- Prebuilt components don't fit the specific use case
116	- User explicitly asks for custom workflow
117	- Complex branching logic required
118	- Advanced streaming patterns needed
119
120	## Model Preferences
121
122	LLM MODEL PRIORITY (follow this order):
123	```python
124	# 1. PREFER: Anthropic
125	from langchain_anthropic import ChatAnthropic
126	model = ChatAnthropic(model="claude-3-5-sonnet-20241022")
127
128	# 2. SECOND CHOICE: OpenAI
129	from langchain_openai import ChatOpenAI
130	model = ChatOpenAI(model="gpt-4o")
131
132	# 3. THIRD CHOICE: Google
133	from langchain_google_genai import ChatGoogleGenerativeAI
134	model = ChatGoogleGenerativeAI(model="gemini-1.5-pro")
135	```
136
137	NOTE: Assume API keys are available in environment.
138	During development, ignore missing key errors.
139
140	## Message and State Handling
141
142	### CRITICAL: Extract Message Content Properly
143	```python
144	# CORRECT: Extract message content properly
145	result = agent.invoke({"messages": state["messages"]})
146	if result.get("messages"):
147	final_message = result["messages"][-1] # This is a message object
148	content = final_message.content # This is the string content
149
150	# WRONG: Treating message objects as strings
151	content = result["messages"][-1] # This is an object, not a string!
152	if content.startswith("Error"): # Will fail - objects don't have startswith()
153	```
154
155	### State Updates Mu

People who liked this also viewed

6 picks

Type

find-skillsHelps users discover and install agent skills when they ask questions like "how do I do X", "find a skill for X", "is there a skill that can...", or express interest in extending capabilities. This skSkillsJul 20262.6M27k
grill-meA relentless interview to sharpen a plan or design.SkillsJul 2026633k182k
vercel-react-best-practicesVercel's official collection of agent skillsSkillsJul 2026573k29k
improve-codebase-architectureScan a codebase for deepening opportunities, present them as a visual HTML report, then grill through whichever one you pick.SkillsJul 2026522k182k
setup-matt-pocock-skillsConfigure this repo for the engineering skills — set up its issue tracker, triage label vocabulary, and domain doc layout. Run once before first use of the other engineering skills.SkillsJul 2026454k182k
triageMove issues and external PRs through a state machine of triage roles — categorise, verify, grill if needed, and write agent-ready briefs.SkillsJul 2026424k182k

langgraph

104stars8forks

View on GitHub

How to install langgraph?

hanamizuki/solopreneur/langgraph

$npx -y skills add hanamizuki/solopreneur --skill langgraph

Installs into the current project.

Summary

Understand LangGraph v1.0 thoroughly

Files · 1

View on GitHub

SKILL.md

1	# LangGraph Development Principles
2
3	If you are coding with LangGraph, follow these principles and patterns.
4
5	## Critical Structure Requirements
6
7	### MANDATORY FIRST STEP
8	Before creating any files, always search the codebase for existing LangGraph-related files:
9	- Files with names like: `graph.py`, `main.py`, `app.py`, `agent.py`, `workflow.py`
10	- Files containing: `.compile()`, `StateGraph`, `create_react_agent`, `app =`, graph exports
11	- Any existing LangGraph imports or patterns
12
13	If any LangGraph files exist: Follow the existing structure exactly. Do not create new agent.py files.
14
15	Only create agent.py when: Building from completely empty directory with zero existing LangGraph files.
16
17	- When starting from scratch, ensure all of the following:
18	1. `agent.py` at project root with compiled graph exported as `app`
19	2. `langgraph.json` configuration file in the same directory as the graph
20	3. Proper state management defined with `TypedDict` or Pydantic `BaseModel`
21	4. Test small components before building complex graphs
22
23	## Deployment-First Principles
24
25	CRITICAL: All LangGraph agents should be written for DEPLOYMENT unless otherwise specified.
26
27	### Core Requirements:
28	- NEVER ADD A CHECKPOINTER unless explicitly requested by user
29	- Always export compiled graph as `app`
30	- Use prebuilt components when possible
31	- Follow model preference hierarchy: Anthropic > OpenAI > Google
32	- Keep state minimal (MessagesState usually sufficient)
33
34	#### AVOID unless user specifically requests
35	```python
36	# Don't do this unless asked!
37	from langgraph.checkpoint.memory import MemorySaver
38	graph = create_react_agent(model, tools, checkpointer=MemorySaver())
39	```
40
41	#### For existing codebases
42	- Always search for existing graph export patterns first
43	- Work within the established structure rather than imposing new patterns
44	- Do not create `agent.py` if graphs are already exported elsewhere
45
46	### Standard Structure for New Projects:
47	```
48	./agent.py # Main agent file, exports: app
49	./langgraph.json # LangGraph configuration
50	```
51
52	### Export Pattern:
53	```python
54	from langgraph.graph import StateGraph, START, END
55	# ... your state and node definitions ...
56
57	# Build your graph
58	graph_builder = StateGraph(YourState)
59	# ... add nodes and edges ...
60
61	# Export as 'app' for new agents from scratch
62	graph = graph_builder.compile()
63	app = graph # Required for new LangGraph agents
64	```
65
66	## Prefer Prebuilt Components
67
68	Always use prebuilt components when possible - they are deployment-ready and well-tested.
69
70	### Basic Agents - Use create_react_agent:
71	```python
72	from langgraph.prebuilt import create_react_agent
73
74	# Simple, deployment-ready agent
75	graph = create_react_agent(
76	model=model,
77	tools=tools,
78	prompt="Your agent instructions here"
79	)
80	app = graph
81	```
82
83	### Multi-Agent Systems:
84
85	#### Supervisor Pattern (central coordination):
86	```python
87	from langgraph_supervisor import create_supervisor
88
89	supervisor = create_supervisor(
90	agents=[agent1, agent2],
91	model=model,
92	prompt="You coordinate between agents..."
93	)
94	app = supervisor.compile()
95	```
96	Documentation: https://langchain-ai.github.io/langgraph/reference/supervisor/
97
98	#### Swarm Pattern (dynamic handoffs):
99	```python
100	from langgraph_swarm import create_swarm, create_handoff_tool
101
102	alice = create_react_agent(
103	model,
104	[tools, create_handoff_tool(agent_name="Bob")],
105	prompt="You are Alice.",
106	name="Alice",
107	)
108
109	workflow = create_swarm([alice, bob], default_active_agent="Alice")
110	app = workflow.compile()
111	```
112	Documentation: https://langchain-ai.github.io/langgraph/reference/swarm/
113
114	### Only Build Custom StateGraph When:
115	- Prebuilt components don't fit the specific use case
116	- User explicitly asks for custom workflow
117	- Complex branching logic required
118	- Advanced streaming patterns needed
119
120	## Model Preferences
121
122	LLM MODEL PRIORITY (follow this order):
123	```python
124	# 1. PREFER: Anthropic
125	from langchain_anthropic import ChatAnthropic
126	model = ChatAnthropic(model="claude-3-5-sonnet-20241022")
127
128	# 2. SECOND CHOICE: OpenAI
129	from langchain_openai import ChatOpenAI
130	model = ChatOpenAI(model="gpt-4o")
131
132	# 3. THIRD CHOICE: Google
133	from langchain_google_genai import ChatGoogleGenerativeAI
134	model = ChatGoogleGenerativeAI(model="gemini-1.5-pro")
135	```
136
137	NOTE: Assume API keys are available in environment.
138	During development, ignore missing key errors.
139
140	## Message and State Handling
141
142	### CRITICAL: Extract Message Content Properly
143	```python
144	# CORRECT: Extract message content properly
145	result = agent.invoke({"messages": state["messages"]})
146	if result.get("messages"):
147	final_message = result["messages"][-1] # This is a message object
148	content = final_message.content # This is the string content
149
150	# WRONG: Treating message objects as strings
151	content = result["messages"][-1] # This is an object, not a string!
152	if content.startswith("Error"): # Will fail - objects don't have startswith()
153	```
154
155	### State Updates Mu

People who liked this also viewed

6 picks

Type