Agent Skills Framework Explorer
/install agent-skills-framework-explorer
\r \r
Agent Skills Framework Explorer\r
\r
Overview\r
\r A comprehensive guide and assistant for navigating the rapidly evolving AI agent skills framework ecosystem. Whether you're exploring addyosmani/agent-skills (40K+ stars), building with Anthropic's Claude Agent SDK, designing multi-agent pipelines with CrewAI/LangGraph, or connecting tools via MCP protocol — this skill helps you understand, compare, and implement the right framework for your use case.\r \r
Triggers\r
\r
- "compare agent frameworks"\r
- "how to use agent-skills"\r
- "build a multi-agent pipeline with [framework]"\r
- "what is MCP protocol"\r
- "CrewAI vs LangGraph vs OpenAI Agents SDK"\r
- "find tools for my agent"\r
- "agent framework comparison"\r
- "MCP server setup"\r
- "AI Agent工具框架对比"\r
- "多智能体框架选型"\r \r
Workflow\r
\r
Step 1: Identify the User's Goal\r
\r Determine the primary use case:\r
- Framework selection: Comparing options for a new project\r
- Tool integration: Adding capabilities to an existing agent\r
- Multi-agent design: Orchestrating multiple specialized agents\r
- Skills marketplace: Finding pre-built agent capabilities\r
- Protocol understanding: Learning MCP, Agent Protocol, or other standards\r
- Migration: Moving from one framework to another\r \r
Step 2: Framework Deep Dive\r
\r Provide structured comparison and guidance for the relevant framework:\r \r
Anthropic agent-skills (addyosmani/agent-skills)\r
- What it is: Shell-based skill framework for CLI agents, Claude Code, and GitHub Actions\r
- Key concepts:
skill.yamlmetadata, bash/python/shell execution, skill chaining\r - Use when: Building CLI tools, automation scripts, developer-facing agents\r
- Ecosystem: 40K+ stars, active community, skill registry at agent-skills.dev\r
- Example skill structure:\r
skill.yaml (name, triggers, description, tools)\r run.sh (main execution script)\r references/ (docs, examples)\r ```\r
\r
OpenAI Agents SDK\r
- What it is: Official Python SDK for building multi-agent systems\r
- Key concepts: Handoffs, tracing, guardrails, function calling\r
- Use when: Building customer-facing agents on OpenAI models\r
- Best for: Product teams needing production-ready agents with built-in observability\r \r
Claude Agent SDK (Anthropic)\r
- What it is: SDK for building Claude-powered agents with tool use\r
- Key concepts: Tools, sessions, context management, computer use\r
- Use when: Deep Claude integration, computer-use agents, complex tool chains\r \r
LangGraph (LangChain)\r
- What it is: Graph-based framework for cyclical agent workflows\r
- Key concepts: Nodes, edges, state machines, human-in-the-loop\r
- Use when: Complex workflows with branching, loops, and multi-agent coordination\r \r
CrewAI\r
- What it is: Role-based multi-agent framework\r
- Key concepts: Agents with roles/goals/backstory, task delegation, crew orchestration\r
- Use when: Team-based AI workflows (e.g., research crew, writing crew)\r
- Best for: Business users who want multi-agent without deep coding\r \r
MCP (Model Context Protocol)\r
- What it is: Open protocol for connecting AI models to external tools\r
- Key concepts: Servers, clients, resources, prompts, tools\r
- Ecosystem: 50+ official and community servers (GitHub, Slack, Postgres, etc.)\r
- Use when: Connecting agents to real-world data and services\r
- Reference: https://modelcontextprotocol.io\r \r
Step 3: Hands-On Implementation\r
\r For each framework, provide starter code and best practices:\r \r
Setting up agent-skills\r
# Clone the framework\r
git clone https://github.com/addyosmani/agent-skills\r
\r
# Register a new skill\r
cd agent-skills/skills\r
npx create-skill my-skill\r
\r
# skill.yaml example\r
name: my-skill\r
triggers:\r
- "run my task"\r
- "execute my workflow"\r
description: "A custom agent skill"\r
tools:\r
- bash\r
- filesystem\r
execute: ./run.sh\r
```\r
\r
#### Building with CrewAI\r
```python\r
from crewai import Agent, Task, Crew\r
\r
researcher = Agent(\r
role="Financial Analyst",\r
goal="Research insurance market trends",\r
backstory="Expert in insurance data analysis",\r
tools=[...],\r
)\r
\r
research_task = Task(\r
description="Analyze Q1 2026 insurance market trends",\r
agent=researcher,\r
)\r
\r
crew = Crew(agents=[researcher], tasks=[research_task])\r
result = crew.kickoff()\r
```\r
\r
#### Connecting via MCP\r
```python\r
from mcp.client import MCPClient\r
\r
client = MCPClient()\r
# Connect to GitHub MCP server\r
await client.connect("github", token=os.getenv("GITHUB_TOKEN"))\r
# Use tools from the server\r
result = await client.call_tool("github", "create_issue", {\r
"repo": "owner/repo",\r
"title": "Agent-generated task",\r
})\r
```\r
\r
### Step 4: Framework Comparison Matrix\r
\r
| Criteria | agent-skills | OpenAI Agents SDK | Claude Agent SDK | LangGraph | CrewAI | MCP |\r
|----------|-------------|-------------------|-----------------|-----------|--------|-----|\r
| Multi-agent | ⭐⭐ | ⭐⭐⭐⭐ | ⭐⭐⭐ | ⭐⭐⭐⭐⭐ | ⭐⭐⭐⭐⭐ | ⭐⭐⭐ |\r
| Ease of use | ⭐⭐⭐⭐ | ⭐⭐⭐⭐ | ⭐⭐⭐ | ⭐⭐⭐ | ⭐⭐⭐⭐⭐ | ⭐⭐⭐ |\r
| Production-ready | ⭐⭐⭐ | ⭐⭐⭐⭐⭐ | ⭐⭐⭐⭐ | ⭐⭐⭐ | ⭐⭐⭐ | ⭐⭐⭐⭐ |\r
| Customization | ⭐⭐⭐⭐ | ⭐⭐ | ⭐⭐⭐ | ⭐⭐⭐⭐⭐ | ⭐⭐⭐ | ⭐⭐⭐⭐ |\r
| Tool ecosystem | ⭐⭐⭐ | ⭐⭐⭐⭐ | ⭐⭐⭐ | ⭐⭐⭐⭐ | ⭐⭐⭐ | ⭐⭐⭐⭐⭐ |\r
| Best for | Dev tools, CLI | Product agents | Claude-native | Complex workflows | Business teams | Tool integration |\r
\r
### Step 5: Skill Discovery\r
\r
Help users find pre-built skills for their needs:\r
1. Search agent-skills.dev registry by category\r
2. Find MCP servers for specific tools (database, API, etc.)\r
3. Discover CrewAI templates and Crew templates\r
4. Identify gaps where a custom skill should be built\r
\r
## Input/Output Examples\r
\r
### Example 1: Framework Selection\r
\r
**Input:**\r
```\r
I want to build an AI agent that researches stock trends and writes a weekly report.\r
Which framework should I use?\r
```\r
\r
**Output:**\r
```\r
Recommendation: CrewAI + MCP combination\r
\r
Rationale:\r
- Research + writing = multi-role workflow → CrewAI's agent delegation shines\r
- Stock data requires external API tools → MCP for data connections\r
- Weekly automation → agent-skills for scheduling\r
\r
Suggested stack:\r
1. CrewAI crew: Researcher agent (MCP-connected) + Writer agent\r
2. MCP servers: Financial data API, web search\r
3. agent-skills: Weekly trigger, report delivery\r
\r
Implementation path:\r
Week 1: Set up CrewAI with 2 agents\r
Week 2: Add MCP financial data tools\r
Week 3: Deploy via agent-skills scheduler\r
```\r
\r
### Example 2: MCP Server Selection\r
\r
**Input:**\r
```\r
I need to connect my agent to PostgreSQL and Slack. Which MCP servers should I use?\r
```\r
\r
**Output:**\r
```\r
Recommended MCP Servers:\r
\r
1. PostgreSQL MCP Server\r
- npm install @modelcontextprotocol/server-postgres\r
- Capabilities: query, schema inspection, backup\r
- Config: DATABASE_URL env var\r
\r
2. Slack MCP Server\r
- npm install @modelcontextprotocol/server-slack\r
- Capabilities: send messages, read channels, manage threads\r
- Config: SLACK_BOT_TOKEN, SLACK_TEAM_ID\r
\r
Quick setup:\r
```bash\r
npx mcp install @modelcontextprotocol/server-postgres\r
npx mcp install @modelcontextprotocol/server-slack\r
```\r
\r
Security note: Store tokens in environment variables, never hardcode.\r
```\r
\r
## Key Frameworks Reference\r
\r
| Framework | Repository | Stars | Primary Language | License |\r
|-----------|-----------|-------|-----------------|---------|\r
| agent-skills | addyosmani/agent-skills | 40K+ | Shell | MIT |\r
| OpenAI Agents SDK | openai/openai-agents-python | Growing | Python | Apache 2.0 |\r
| Claude Agent SDK | anthropics/anthropic-cookbook | - | Python | - |\r
| LangGraph | langchain-ai/langgraph | - | Python | MIT |\r
| CrewAI | crewAI同事/crewai | - | Python | MIT |\r
| MCP | modelcontextprotocol/spec | - | Multi | Apache 2.0 |\r
\r
## Best Practices\r
\r
1. **Start with the right abstraction level** — agent-skills for CLI tools, CrewAI for business workflows, LangGraph for complex state machines\r
2. **Use MCP for all external integrations** — it provides a standardized, swappable interface\r
3. **Combine frameworks** — use CrewAI for orchestration + MCP for tools + agent-skills for deployment\r
4. **Monitor agent behavior** — all major frameworks support tracing (OpenAI, LangSmith, etc.)\r
5. **Design for failure** — agents can hallucinate; add guardrails and human-in-the-loop for critical actions\r
- 确保已安装 OpenClaw(本地或 Docker 部署)
- 在对话框中输入安装命令:
/install agent-skills-framework-explorer - 安装完成后,直接呼叫该 Skill 的名称或使用
/agent-skills-framework-explorer触发 - 根据 Skill 的参数说明提供必要输入,即可获得结构化输出
Agent Skills Framework Explorer 是什么?
AI-powered assistant for exploring, understanding, and building with AI agent skills frameworks — covers Anthropic agent-skills, OpenAI agent SDKs, LangChain... 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件,目前累计下载 106 次。
如何安装 Agent Skills Framework Explorer?
在 OpenClaw 或 Claude Code 对话框中运行命令「/install agent-skills-framework-explorer」即可一键安装,无需额外配置。
Agent Skills Framework Explorer 是免费的吗?
是的,Agent Skills Framework Explorer 完全免费,采用 MIT-0 许可证,可自由下载、安装和使用。
Agent Skills Framework Explorer 支持哪些平台?
Agent Skills Framework Explorer 跨平台运行,可在任意部署了 OpenClaw / Claude Code 的环境中使用(cross-platform)。
谁开发了 Agent Skills Framework Explorer?
由 lingfeng-19(@gechengling)开发并维护,当前版本 v1.0.0。