Skip to main content

danube-tools

Use Danube's 100+ API tools (Gmail, GitHub, Notion, etc.) through MCP. Search for tools, check authentication, execute with parameters, and handle errors gracefully.

설치로 이동

소스 정보

저장소
Kernel8901/ai-agent-skills-classification
최근 소스 활동
2026년 4월 4일 15:26
감지된 SKILL.md 언어
영어
스타
5
포크
1

설치 방법

기본적으로 소스를 먼저 확인하는 Prompt가 선택됩니다. 직접 명령으로 전환하거나 로컬 사본을 다운로드할 수도 있습니다.

소스 파일 검토

설치 여부를 결정하기 전에 SKILL.md와 SkillsMP에 표시된 보조 파일을 읽어 보세요.

파일 탐색기
3 개 파일

SKILL.md 표시 중

SKILL.md
소스 지침 · 읽기 전용 미리보기
name
danube-tools
description
Use Danube's 100+ API tools (Gmail, GitHub, Notion, etc.) through MCP. Search for tools, check authentication, execute with parameters, and handle errors gracefully.
license
MIT
compatibility
openclaw
metadata
{"author":"danube","version":"1.0.0","tags":["danube","mcp","apis","tools"]}
# Using Danube Tools Access 100+ API tools for Gmail, GitHub, Notion, Google Calendar, and more through Danube's MCP integration. **Setup:** If not configured yet, run `bash scripts/setup.sh` to add Danube MCP to OpenClaw. ## When to Use Use Danube when users want to: - Send emails, create issues, manage calendars - Interact with cloud services (GitHub, Notion, Google Drive) - Execute any external API action **Don't use for:** Local file operations, calculations, or non-API tasks. ## Core Workflow Every tool interaction follows this pattern: ### 1. Search for Tools Use `search_tools()` with natural language: ```python search_tools("send email") # → Gmail, Outlook search_tools("create github issue") # → GitHub - Create Issue search_tools("calendar events") # → Google Calendar ``` ### 2. Check Authentication If tool requires credentials, guide user to connect: ``` "To use Gmail, you need to connect your account first. Visit: https://danubeai.com/dashboard 1. Go to Tools section 2. Find Gmail and click 'Connect' 3. Follow the OAuth flow Let me know when you're ready!" ``` **Always check auth BEFORE attempting execution.** ### 3. Gather Parameters Ask for missing required parameters: ``` User: "Send an email" You: "I can help! I need: - Who should I send it to? - What's the subject? - What should the message say?" ``` ### 4. Execute Tool ```python execute_tool( tool_id="gmail-send-email-uuid", parameters={ "to": "user@example.com", "subject": "Meeting", "body": "Confirming our 2pm meeting." } ) ``` ### 5. Handle Response **Success:** ``` "✅ Email sent successfully to user@example.com!" ``` **Auth Error:** ``` "🔐 Authentication failed. Reconnect Gmail at: https://danubeai.com/dashboard → Tools → Gmail" ``` **Other Error:** ``` "⚠️ Failed: [error]. Let me help troubleshoot..." ``` ## Common Patterns ### Email Tools ``` User: "Email john@example.com about the project" 1. search_tools("send email") → Find Gmail 2. Check Gmail authentication 3. Extract: to="john@example.com", subject="Project" 4. Ask: "What should the message say?" 5. Confirm: "I'll send email to john@example.com. Proceed?" 6. execute_tool() 7. Report: "✅ Email sent!" ``` ### GitHub Tools ``` User: "Create issue about the login bug" 1. search_tools("github create issue") 2. Check GitHub authentication 3. Ask: "Which repository?" 4. Ask: "Describe the bug?" 5. execute_tool() 6. Report: "✅ Issue created: [link]" ``` ### Calendar Tools ``` User: "What's on my calendar today?" 1. search_tools("calendar events") 2. Check authentication 3. execute_tool(date=today) 4. Format results: "Here's your schedule: • 9:00 AM - Team standup • 2:00 PM - Client meeting" ``` ## Best Practices ### ✅ Do: - **Search first** - Always use `search_tools()`, don't assume tool IDs - **Check auth** - Verify credentials before execution - **Confirm actions** - Get user approval for emails, issues, etc. - **Be specific** - "Email sent to john@example.com" not just "Done" - **Handle errors** - Provide solutions, not just error messages ### ❌ Don't: - Assume tool IDs without searching - Auto-execute without confirmation - Give vague responses like "Error" or "Done" - Skip authentication checks ## Available Tools | Tool | Purpose | |------|---------| | `list_services` | Browse available services | | `search_tools` | Find tools by query | | `get_service_tools` | List tools for a service | | `execute_tool` | Run tool with parameters | | `search_contacts` | Find user's contacts | ## Error Handling **Authentication (401):** ``` "🔐 [Service] requires authentication. Visit https://danubeai.com/dashboard → Tools → [Service] → Connect" ``` **Missing Parameters:** ``` "I need: • [param1]: [description] • [param2]: [description]" ``` **Rate Limit:** ``` "⚠️ Hit rate limit for [Service]. • Try again in a few minutes • Use alternative service • Break into smaller batches" ``` ## Multi-Step Workflows Some tasks need multiple tools: ``` User: "Email me a summary of today's GitHub commits" 1. search_tools("github commits") → Get tool 2. execute_tool() → Fetch commits 3. Format into summary 4. search_tools("send email") → Get Gmail 5. execute_tool() → Send email 6. Report: "✅ Sent summary of 5 commits!" ``` ## Communication Templates **Request Auth:** ``` "To use [Service], connect your account: 1. Visit https://danubeai.com/dashboard 2. Tools → [Service] → Connect 3. Come back when ready!" ``` **Confirm Execution:** ``` "I'll [action] using [Tool]. Parameters: [list] Proceed?" ``` **Report Success:** ``` "✅ Done! [Specific result] [Link if applicable]" ``` ## Quick Reference **Workflow:** ``` 1. User requests action 2. search_tools() → Find tool 3. Check authentication → Guide if needed 4. Gather parameters → Ask for missing info 5. Confirm → Get approval 6. execute_tool() → Run it 7. Report → Success or error ``` **Key URLs:** - **MCP Server:** https://mcp.danubeai.com/mcp - **Dashboard:** https://danubeai.com/dashboard - **Connect Services:** https://danubeai.com/dashboard → Tools **Debug:** - Check API key: `cat ~/.openclaw/.env | grep DANUBE` - Restart: `openclaw gateway restart` - Check errors: `openclaw doctor`
GitHub에서 보기