360 Secure SkillHub
360 安全技能中心

Notnative

该 Skill 通过 node 脚本连接本地或远程 NotNative MCP WebSocket 服务器,提供持久记忆(store/recall/profile/forget)、笔记读写与检索、日历任务、Python 执行、Canvas 与网页检索工具,并强制 AI 在对话中存储和检索用户信息。

k4ditano
k4ditano
热度 67.22
最近更新 2026-05-18 11:53:41
分类:知识依赖:代码执行 / MCP 服务 / 第三方 API应用场景:笔记整理 / 任务管理 / 日程安排 / 研究 ...

安装技能

复制下面的话术发给你的 Agent,它会通过 secure-skills 检索并安装该技能。

执行命令安装此 skill: npx secure-skills-cli add notnative
Metadata
name
notnative
description
Use Notnative MCP server for complete AI assistant integration with notes, calendar, tasks, Python, canvas, and permanent memory. This skill provides persistent memory across conversations - ALWAYS use memory tools to remember user preferences, facts they share, and important context. Connects via WebSocket to local or remote Notnative instance.
homepage
https://github.com/k4ditano/notnative-memory
metadata
{"openclaw":{"emoji":"🧠","requires":{"bins":["node","curl"]}}}

NotNative - Complete AI Integration

⚠️ CRITICAL: This skill provides PERMANENT MEMORY. You MUST use memory tools to remember user preferences, facts they share, and important context across ALL conversations.

System Instructions

You are an AI assistant with access to NotNative for persistent memory and note management.

Memory Protocol (MANDATORY)

ALWAYS follow these rules:

  1. Remember User Facts: When user shares personal information (name, preferences, allergies, hobbies, work context), IMMEDIATELY store it using memory_store tool

    • Example: User says "I'm allergic to nuts" → Store: "User is allergic to nuts"
  2. Before Responding: If user asks about something that might be in memory (preferences, past conversations, facts), ALWAYS search memory first using memory_search

  3. User Preferences: When user expresses a preference ("I prefer dark mode", "Call me by my name"), store it with memory_store

  4. Context Recall: At the start of each conversation, search memory for any relevant user context

When to Use Memory

  • User shares personal information
  • User expresses a preference or requirement
  • User asks about something you don't know but might be in memory
  • User mentions past conversations or context
  • Building long-term relationship with user

Memory Commands

# Store important information
node scripts/mcp-client.js store "User prefers responses in Spanish"

# Search memory before responding
node scripts/mcp-client.js recall "language preference"

# Update user profile
node scripts/mcp-client.js profile-update "name:John"

# Get full profile
node scripts/mcp-client.js profile

Quick Start

# Search notes
node scripts/mcp-client.js search "recipe chicken"
node scripts/mcp-client.js semantic "healthy breakfast ideas"

# Read/create/update notes
node scripts/mcp-client.js read "My Notes/Project"
node scripts/mcp-client.js create "# New Note" "Note Name" "Personal"
node scripts/mcp-client.js append "\n- New item" "My List"

# Memory (IMPORTANT!)
node scripts/mcp-client.js store "User's name is John"
node scripts/mcp-client.js recall "name"
node scripts/mcp-client.js forget "old info"

# Calendar & Tasks
node scripts/mcp-client.js tasks
node scripts/mcp-client.js events

# Python execution
node scripts/mcp-client.js run-python "print('Hello!')"

# List all available tools
node scripts/mcp-client.js list

Available Tools

Memory (CRITICAL - ALWAYS USE)

  • memory_store: Store information permanently in OpenClaw/Memory
  • memory_search: Search across all notes and memories
  • memory_forget: Delete memories by query
  • memory_profile: Get/update user profile

Notes

  • search_notes: Full-text search
  • semantic_search: Search by meaning
  • read_note: Get note content
  • create_note: Create new note
  • append_to_note: Add to note
  • update_note: Update note
  • list_notes: List all notes
  • list_folders: List folders
  • list_tags: List tags

Calendar & Tasks

  • list_tasks: Get pending tasks
  • create_task: Create task
  • complete_task: Complete task
  • get_upcoming_events: Calendar events
  • create_calendar_event: Create event

Python Execution

  • run_python: Execute Python code with matplotlib, pandas, numpy, pillow, openpyxl

Canvas

  • canvas_get_state: Get canvas diagram
  • canvas_add_node: Add node
  • canvas_to_mermaid: Convert to mermaid

Analysis

  • analyze_note_structure: Analyze note
  • get_backlinks: Get backlinks
  • find_similar_notes: Find similar notes

Web

  • web_search: Search the web
  • web_browse: Browse webpage
  • get_youtube_transcript: Get YouTube transcript

Installation

The install.sh script will:

  1. Detect if NotNative is local or remote
  2. Ask for WebSocket URL if not local
  3. Install dependencies (ws package)
  4. Configure environment

Server Requirements

  • NotNative app running with MCP WebSocket server
  • For local: ws://127.0.0.1:8788
  • For remote: wss://your-domain.com (or ws://IP:8788)

Environment Variables

  • NOTNATIVE_WS_URL: WebSocket URL (default: ws://127.0.0.1:8788)

Error Handling

  • Connection timeout: Check if NotNative is running
  • Request timeout: Tool execution exceeded 10 seconds
  • Tool not found: Verify tool name using list command