SKILL.md
AI Agent Basics
Build production-grade AI agents with modern architectures and patterns.
When to Use This Skill
Invoke this skill when:
- Designing new AI agent systems
- Implementing ReAct or Plan-and-Execute patterns
- Building autonomous task-solving agents
- Integrating cognitive loops into applications
Parameter Schema
| Parameter | Type | Required | Description | Default |
|---|---|---|---|---|
task |
string | Yes | What agent capability to build | - |
architecture |
enum | No | single, multi, hybrid |
single |
framework |
enum | No | langchain, langgraph, custom |
langgraph |
complexity |
enum | No | basic, intermediate, advanced |
intermediate |
Quick Start
# Basic ReAct Agent
from langgraph.prebuilt import create_react_agent
from langchain_anthropic import ChatAnthropic
llm = ChatAnthropic(model="claude-sonnet-4-20250514")
agent = create_react_agent(llm, tools=[search, calculator])
result = await agent.ainvoke({"messages": [("user", "What is 25 * 4?")]})Core Patterns
1. ReAct Agent
# Thought → Action → Observation loop
graph = StateGraph(AgentState)
graph.add_node("think", reason_node)
graph.add_node("act", action_node)
graph.add_node("observe", observation_node)2. Plan-and-Execute
# Create plan → Execute steps → Verify
planner = create_planner(llm)
executor = create_executor(llm, tools)3. Reflexion
# Execute → Reflect → Improve
agent_with_reflection = add_reflection_layer(base_agent)Troubleshooting
| Issue | Solution |
|---|---|
| Agent loops forever | Add max_iterations limit |
| Wrong tool selected | Improve tool descriptions |
| Context too large | Implement summarization |
| Slow responses | Use streaming |
Best Practices
- Start with simple single-agent before multi-agent
- Always add circuit breakers (max iterations)
- Use verbose mode for debugging
- Implement human-in-the-loop for critical decisions
Related Skills
llm-integration- LLM API configurationtool-calling- Function calling implementationagent-memory- Memory systems