做什么
技能从mcp-create-声明-agent.pot.md转换
技能库 智客分类:Agent 工作流 mcp-create-declarative-agent
技能从mcp-create-声明-agent.pot.md转换
官方网址:skills.sh
先看中文介绍;官方 description 原文单独保留,不改写 SKILL.md。
技能从mcp-create-声明-agent.pot.md转换
每个工具
按 Agent Skills 渐进披露:启动时只加载 name 与 description(约 100 token);任务匹配后才读入整份 SKILL.md 正文;scripts/、references/、assets/ 仅在需要时再读。 本文件正文结构:Create MCP-based Declarative Agent for Microsoft 365 Copilot、Requirements、Project Setup、Key Files Generated、MCP Server Integration、Supported MCP Endpoints。
文件分析:这是一份仅含 SKILL.md 的指令型技能,代理激活后整份正文进入上下文。
Skill converted from mcp-create-declarative-agent.prompt.md
Create MCP-based Declarative Agent for Microsoft 365 CopilotRequirementsProject SetupKey Files GeneratedMCP Server IntegrationSupported MCP EndpointsTool SelectionAuthentication TypesResponse SemanticsDefine Data MappingAdd Adaptive Cards (Optional)Environment Configuration
来源分类:skills.sh agent-skill
namemcp-create-declarative-agentdescription具体调用语法与可用工具以目标 Agent 客户端为准。 查看调用机制说明 ↗
先选择目标 Agent 和安装范围,保留技能包的附属文件,安装后检查客户端能否发现该技能。
复制安装指令给支持 Agent Skills 的代理,确认其中的目标目录与客户端匹配。
把 Agent Skill「mcp-create-declarative-agent」安装到我的项目:SKILL.md 原文与官方 description 见 https://zicq.com/zh/skills/skl-2f2fc73f2a151f4a-Mcp-%E5%88%9B%E5%BB%BA%E5%A3%B0%E6%98%8E%E4%BB%A3%E7%90%86.html 请存为 .cursor/skills/mcp-create-declarative-agent/SKILL.md 或 .claude/skills/mcp-create-declarative-agent/SKILL.md,frontmatter 的 name 与 description 保持原样,不要改写。
需要 Node.js 与 npx。先查看仓库技能列表,确认实际名称。
npx skills add 'https://github.com/github/awesome-copilot' --list
npx skills add 'https://github.com/github/awesome-copilot' --skill 'mcp-create-declarative-agent'
CLI 会交互选择目标 Agent,默认安装到项目;用户级安装使用 -g。先通过查看命令核对仓库内容,再用 npx skills list 检查已安装技能。
---
mode: 'agent'
tools: ['changes', 'search/codebase', 'edit/editFiles', 'problems']
description: 'Create a declarative agent for Microsoft 365 Copilot by integrating an MCP server with authentication, tool selection, and configuration'
model: 'gpt-4.1'
tags: [mcp, m365-copilot, declarative-agent, model-context-protocol, api-plugin]
---
# Create MCP-based Declarative Agent for Microsoft 365 Copilot
Create a complete declarative agent for Microsoft 365 Copilot that integrates with a Model Context Protocol (MCP) server to access external systems and data.
## Requirements
Generate the following project structure using Microsoft 365 Agents Toolkit:
### Project Setup
1. **Scaffold declarative agent** via Agents Toolkit
2. **Add MCP action** pointing to MCP server
3. **Select tools** to import from MCP server
4. **Configure authentication** (OAuth 2.0 or SSO)
5. **Review generated files** (manifest.json, ai-plugin.json, declarativeAgent.json)
### Key Files Generated
**appPackage/manifest.json** - Teams app manifest with plugin reference:
```json
{
"$schema": "https://developer.microsoft.com/json-schemas/teams/vDevPreview/MicrosoftTeams.schema.json",
"manifestVersion": "devPreview",
"version": "1.0.0",
"id": "...",
"developer": {
"name": "...",
"websiteUrl": "...",
"privacyUrl": "...",
"termsOfUseUrl": "..."
},
"name": {
"short": "Agent Name",
"full": "Full Agent Name"
},
"description": {
"short": "Short description",
"full": "Full description"
},
"copilotAgents": {
"declarativeAgents": [
{
"id": "declarativeAgent",
"file": "declarativeAgent.json"
}
]
}
}
```
**appPackage/declarativeAgent.json** - Agent definition:
```json
{
"$schema": "https://aka.ms/json-schemas/copilot/declarative-agent/v1.0/schema.json",
"version": "v1.0",
"name": "Agent Name",
"description": "Agent description",
"instructions": "You are an assistant that helps with [specific domain]. Use the available tools to [capabilities].",
"capabilities": [
{
"name": "WebSearch",
"websites": [
{
"url": "https://learn.microsoft.com"
}
]
},
{
"name": "MCP",
"file": "ai-plugin.json"
}
]
}
```
**appPackage/ai-plugin.json** - MCP plugin manifest:
```json
{
"schema_version": "v2.1",
"name_for_human": "Service Name",
"description_for_human": "Description for users",
"description_for_model": "Description for AI model",
"contact_email": "[email protected]",
"namespace": "serviceName",
"capabilities": {
"conversation_starters": [
{
"text": "Example query 1"
}
]
},
"functions": [
{
"name": "functionName",
"description": "Function description",
"capabilities": {
"response_semantics": {
"data_path": "$",
"properties": {
"title": "$.title",
"subtitle": "$.description"
}
}
}
}
],
"runtimes": [
{
"type": "MCP",
"spec": {
"url": "https://api.service.com/mcp/"
},
"run_for_functions": ["functionName"],
"auth": {
"type": "OAuthPluginVault",
"reference_id": "${{OAUTH_REFERENCE_ID}}"
}
}
]
}
```
**/.vscode/mcp.json** - MCP server configuration:
```json
{
"serverUrl": "https://api.service.com/mcp/",
"pluginFilePath": "appPackage/ai-plugin.json"
}
```
## MCP Server Integration
### Supported MCP Endpoints
The MCP server must provide:
- **Server metadata** endpoint
- **Tools listing** endpoint (exposes available functions)
- **Tool execution** endpoint (handles function calls)
### Tool Selection
When importing from MCP:
1. Fetch available tools from server
2. Select specific tools to include (for security/simplicity)
3. Tool definitions are auto-generated in ai-plugin.json
### Authentication Types
**OAuth 2.0 (Static Registration)**
```json
"auth": {
"type": "OAuthPluginVault",
"reference_id": "${{OAUTH_REFERENCE_ID}}",
"authorization_url": "https://auth.service.com/authorize",
"client_id": "${{CLIENT_ID}}",
"client_secret": "${{CLIENT_SECRET}}",
"scope": "read write"
}
```
**Single Sign-On (SSO)**
```json
"auth": {
"type": "SSO"
}
```
## Response Semantics
### Define Data Mapping
Use `response_semantics` to extract relevant fields from API responses:
```json
"capabilities": {
"response_semantics": {
"data_path": "$.results",
"properties": {
"title": "$.name",
"subtitle": "$.description",
"url": "$.link"
}
}
}
```
### Add Adaptive Cards (Optional)
See the `mcp-create-adaptive-cards` prompt for adding visual card templates.
## Environment Configuration
Create `.env.local` or `.env.dev` for credentials:
```env
OAUTH_REFERENCE_ID=your-oauth-reference-id
CLIENT_ID=your-client-id
CLIENT_SECRET=your-client-secret
```
## Testing & Deployment
### Local Testing
1. **Provision** agent in Agents Toolkit
2. **Start debugging** to sideload in Teams
3. Test in Microsoft 365 Copilot at https://m365.cloud.microsoft/chat
4. Authenticate when prompted
5. Query the agent using natural language
### Validation
- Verify tool imports in ai-plugin.json
- Check authentication configuration
- Test each exposed function
- Validate response data mapping
## Best Practices
### Tool Design
- **Focused functions**: Each tool should do one thing well
- **Clear descriptions**: Help the model understand when to use each tool
- **Minimal scoping**: Only import tools the agent needs
- **Descriptive names**: Use action-oriented function names
### Security
- **Use OAuth 2.0** for production scenarios
- **Store secrets** in environment variables
- **Validate inputs** on the MCP server side
- **Limit scopes** to minimum required permissions
- **Use reference IDs** for OAuth registration
### Instructions
- **Be specific** about the agent's purpose and capabilities
- **Define behavior** for both successful and error scenarios
- **Reference tools** explicitly in instructions when applicable
- **Set expectations** for users about what the agent can/cannot do
### Performance
- **Cache responses** when appropriate on MCP server
- **Batch operations** where possible
- **Set timeouts** for long-running operations
- **Paginate results** for large datasets
## Common MCP Server Examples
### GitHub MCP Server
```
URL: https://api.githubcopilot.com/mcp/
Tools: search_repositories, search_users, get_repository
Auth: OAuth 2.0
```
### Jira MCP Server
```
URL: https://your-domain.atlassian.net/mcp/
Tools: search_issues, create_issue, update_issue
Auth: OAuth 2.0
```
### Custom Service
```
URL: https://api.your-service.com/mcp/
Tools: Custom tools exposed by your service
Auth: OAuth 2.0 or SSO
```
## Workflow
Ask the user:
1. What MCP server are you integrating with (URL)?
2. What tools should be exposed to Copilot?
3. What authentication method does the server support?
4. What should the agent's primary purpose be?
5. Do you need response semantics or Adaptive Cards?
Then generate:
- Complete appPackage/ structure (manifest.json, declarativeAgent.json, ai-plugin.json)
- mcp.json configuration
- .env.local template
- Provisioning and testing instructions
## Troubleshooting
### MCP Server Not Responding
- Verify server URL is correct
- Check network connectivity
- Validate MCP server implements required endpoints
### Authentication Fails
- Verify OAuth credentials are correct
- Check reference ID matches registration
- Confirm scopes are requested properly
- Test OAuth flow independently
### Tools Not Appearing
- Ensure mcp.json points to correct server
- Verify tools were selected during import
- Check ai-plugin.json has correct function definitions
- Re-fetch actions from MCP if server changed
### Agent Not Understanding Queries
- Review instructions in declarativeAgent.json
- Check function descriptions are clear
- Verify response_semantics extract correct data
- Test with more specific queries
Agent 工作流
创造有效技能指南。 当用户想创造出新的技能(或更新现有的技能),以专业知识,工作流程,或工具集成来扩展克洛德的能力时,应该使用这种技能.
Agent 工作流
使用ClawdHub CLI搜索,安装,更新并发布从taladhub.com的代理技能. 需要获取苍蝇上的新技能时使用,将安装的技能同步到最新版本或特定版本,或者发布 npm-instainddhub CLI 的新/更新的技能文件夹.
Agent 工作流
管弦乐团多代理团队,任务设定周期,交接协议,审查工作流程. 使用时间: (1)建立2+特派员队伍,具有不同专业,(2)确定任务路线和生命周期(收录框_ spec_建设_审查_完成),(3)在特派员之间制定交接协议,(4)建立审查和质量关口,(5)管理特派员之间的交流和文物共享.
Agent 工作流
Spec-first,TDD,子代理驱动的软件开发工作流程. 当:(1)构建任何新功能或应用——触发脑暴_计划_子代理执行回路,(2)调试出一个bug或测试失败——触发系统性的根起过程,(3)用户说"让我们构建","帮助我计划","我想添加X",或"这个被打破",(4)完成一个功…