
mcp-builder
by Lordsisodia
SISO Agency Internal Platform - Task management and day tracking system with LifeLock integration
SKILL.md
name: mcp-builder category: integration-connectivity/mcp-integrations version: 1.0.0 description: Build custom Model Context Protocol servers to extend Claude's capabilities author: blackbox5/mcp verified: true tags: [mcp, integration, development, api]
MCP Builder
When to Use:
- Building custom tool integrations for Claude
- Connecting Claude to internal APIs or databases
- Creating reusable Claude capabilities
- Extending Claude with external data sources
Each tool should have a single responsibility, clear JSON Schema validation, and helpful error messages.
<available_skills> <skill_group name="Server Setup"> Create a basic MCP server Help me create an MCP server for [API/service] Set up stdio or SSE transport Configure server to use stdio transport Implement tool listing capability Add listTools request handler </skill_group>
<skill_group name="Tool Development"> Create a new tool with schema Add a tool to my MCP server that [does X] Define input validation schema Create JSON Schema for tool inputs Write tool handler function Implement the tool logic </skill_group>
<skill_group name="Integration Patterns"> Connect to external API Integrate with REST API Query database from MCP Add database query capability Read/write files through MCP Add file system access </skill_group>
<skill_group name="Best Practices"> Implement proper error responses Add error handling to tool Validate input parameters Add input validation Document tool usage Add tool documentation </skill_group> </available_skills>
<best_practices> Use TypeScript for type safety Give tools single responsibility Use JSON Schema for validation Return helpful error messages Make tools idempotent when possible Close connections properly Use timeouts for external calls Handle rate limiting gracefully Log important operations Mix multiple responsibilities in one tool Skip input validation Return cryptic error messages Forget error handling Hardcode credentials Log sensitive data Ignore rate limits </best_practices>
<error_handling> Tool not found Verify tool is registered in server Check tool name spelling Ensure listTools includes the tool Invalid input Return clear validation error Specify which field is invalid Provide example of valid input API connection failed Return actionable error message Suggest checking API credentials Recommend verifying permissions Rate limit exceeded Implement exponential backoff Return helpful retry message Consider caching responses </error_handling>
<integration_notes>
Claude Desktop Config
json { "mcpServers": { "my-server": { "command": "node", "args": ["/path/to/server/build/index.js"] } } }
</integration_notes>
<security_considerations> Input Validation Validate all input parameters with JSON Schema Secrets Management Use environment variables for sensitive data Rate Limiting Implement rate limiting for external API calls Logging Never log sensitive data (passwords, tokens) Authentication Use authentication for protected resources Sanitization Sanitize data before external API calls </security_considerations>
<advanced_features> Resources Expose data sources that Claude can read Prompts Provide reusable prompt templates Tools Executable functions Claude can call </advanced_features>
スコア
総合スコア
リポジトリの品質指標に基づく評価
SKILL.mdファイルが含まれている
ライセンスが設定されている
100文字以上の説明がある
GitHub Stars 100以上
3ヶ月以内に更新がある
10回以上フォークされている
オープンIssueが50未満
プログラミング言語が設定されている
1つ以上のタグが設定されている
レビュー
レビュー機能は近日公開予定です