[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M717ME8VC3A7Y7HTD1NQ3RNA":3,"skill-same:01M717ME8VC3A7Y7HTD1NQ3RNA":74},{"skillNo":4,"name":5,"nameZh":6,"description":7,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":16,"categories":17,"keywords":26,"publishedAt":47,"linkOnly":48,"markdownContent":49,"markdownBody":50,"frontMatter":51,"refName":54,"commitSha":55,"skillDirectory":56,"forksCount":14,"repositoryHtmlUrl":57,"repositoryHomepageUrl":6,"ownerHtmlUrl":58,"ownerBlogUrl":6,"ownerDescription":6,"updatedAt":59,"install":60,"resources":73},"01M717ME8VC3A7Y7HTD1NQ3RNA","agent-tool-builder",null,"Tools are how AI agents interact with the world. A well-designed tool is the difference between an agent that works and one that hallucinates, fails silently, or costs 10x more tokens than necessary. This skill covers tool design from schema to error handling.","sapsapshen\u002Fai-memory-router","sapsapshen","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F20347959?v=4","01M6XDQEF1DP50DKF37H81J3J5","shenmian","01M6XDQEJPE41Y3DT2763X8ZJN",0,"MIT","https:\u002F\u002Fgithub.com\u002Fsapsapshen\u002Fai-memory-router\u002Fblob\u002Fmain\u002Fskills-store\u002Fglobal\u002Fagent-tool-builder\u002FSKILL.md",[18,23],{"categoryNo":19,"categoryName":20,"taxonomyType":21,"taxonomyTypeDesc":22},"01JBSEEDCAT000000000000015","视觉设计","SCENARIO","使用场景",{"categoryNo":24,"categoryName":25,"taxonomyType":21,"taxonomyTypeDesc":22},"01JBSEEDCAT000000000000023","Skill与智能体开发",[27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43,44,45,46],"agent tool builder","agent tool","tool","builder","error handling","schema","more","tokens","10x","Tools","2026","how","apache","world","ai agents","unknown","tool design","spawner","design","error","2026-11-10 10:12:30",false,"---\nname: agent-tool-builder\ndescription: Tools are how AI agents interact with the world. A well-designed\n  tool is the difference between an agent that works and one that hallucinates,\n  fails silently, or costs 10x more tokens than necessary. This skill covers\n  tool design from schema to error handling.\nrisk: unknown\nsource: vibeship-spawner-skills (Apache 2.0)\ndate_added: 2026-02-27\n---\n\n# Agent Tool Builder\n\nTools are how AI agents interact with the world. A well-designed tool is the\ndifference between an agent that works and one that hallucinates, fails\nsilently, or costs 10x more tokens than necessary.\n\nThis skill covers tool design from schema to error handling. JSON Schema\nbest practices, description writing that actually helps the LLM, validation,\nand the emerging MCP standard that's becoming the lingua franca for AI tools.\n\nKey insight: Tool descriptions are more important than tool implementations.\nThe LLM never sees your code - it only sees the schema and description.\n\n## Principles\n\n- Description quality > implementation quality for LLM accuracy\n- Aim for fewer than 20 tools - more causes confusion\n- Every tool needs explicit error handling - silent failures poison agents\n- Return strings, not objects - LLMs process text\n- Validation gates before execution - reject, fix, or escalate, never silent fail\n- Test tools with the LLM, not just unit tests\n\n## Capabilities\n\n- agent-tools\n- function-calling\n- tool-schema-design\n- mcp-tools\n- tool-validation\n- tool-error-handling\n\n## Scope\n\n- multi-agent-coordination → multi-agent-orchestration\n- agent-memory → agent-memory-systems\n- api-design → api-designer\n- llm-prompting → prompt-engineering\n\n## Tooling\n\n### Standards\n\n- JSON Schema - When: All tool definitions Note: The universal format for tool schemas\n- MCP (Model Context Protocol) - When: Building reusable, cross-platform tools Note: Anthropic's open standard, widely adopted\n\n### Frameworks\n\n- Anthropic SDK - When: Claude-based agents Note: Beta tool runner handles most complexity\n- OpenAI Functions - When: OpenAI-based agents Note: Use strict mode for guaranteed schema compliance\n- Vercel AI SDK - When: Multi-provider tool handling Note: Abstracts differences between providers\n- LangChain Tools - When: LangChain-based agents Note: Converts MCP tools to LangChain format\n\n## Patterns\n\n### Tool Schema Design\n\nCreating clear, unambiguous JSON Schema for tools\n\n**When to use**: Defining any new tool for an agent\n\n# TOOL SCHEMA BEST PRACTICES:\n\n## 1. Detailed Descriptions (Most Important)\n\"\"\"\nBAD - Too vague:\n{\n  \"name\": \"get_stock_price\",\n  \"description\": \"Gets stock price\",\n  \"input_schema\": {\n    \"type\": \"object\",\n    \"properties\": {\n      \"ticker\": {\"type\": \"string\"}\n    }\n  }\n}\n\nGOOD - Comprehensive:\n{\n  \"name\": \"get_stock_price\",\n  \"description\": \"Retrieves the current stock price for a given ticker\n    symbol. The ticker symbol must be a valid symbol for a publicly\n    traded company on a major US stock exchange like NYSE or NASDAQ.\n    Returns the latest trade price in USD. Use when the user asks\n    about current or recent stock prices. Does NOT provide historical\n    data, company info, or predictions.\",\n  \"input_schema\": {\n    \"type\": \"object\",\n    \"properties\": {\n      \"ticker\": {\n        \"type\": \"string\",\n        \"description\": \"The stock ticker symbol, e.g. AAPL for Apple Inc.\"\n      }\n    },\n    \"required\": [\"ticker\"]\n  }\n}\n\"\"\"\n\n## 2. Parameter Descriptions\n\"\"\"\nEvery parameter needs:\n- What it is\n- Format expected\n- Example value\n- Edge cases\u002Flimitations\n\n{\n  \"location\": {\n    \"type\": \"string\",\n    \"description\": \"City and state\u002Fcountry. Format: 'City, State' for US\n      (e.g., 'San Francisco, CA') or 'City, Country' for international\n      (e.g., 'Tokyo, Japan'). Do not use ZIP codes or coordinates.\"\n  },\n  \"unit\": {\n    \"type\": \"string\",\n    \"enum\": [\"celsius\", \"fahrenheit\"],\n    \"description\": \"Temperature unit. Defaults to user's locale if not\n      specified. Use 'fahrenheit' for US users, 'celsius' for others.\"\n  }\n}\n\"\"\"\n\n## 3. Use Enums When Possible\n\"\"\"\nEnums constrain the LLM to valid values:\n\n\"priority\": {\n  \"type\": \"string\",\n  \"enum\": [\"low\", \"medium\", \"high\", \"critical\"],\n  \"description\": \"Task priority level\"\n}\n\n\"action\": {\n  \"type\": \"string\",\n  \"enum\": [\"create\", \"read\", \"update\", \"delete\"],\n  \"description\": \"The CRUD operation to perform\"\n}\n\"\"\"\n\n## 4. Required vs Optional\n\"\"\"\nBe explicit about what's required:\n\n{\n  \"type\": \"object\",\n  \"properties\": {\n    \"query\": {...},      \u002F\u002F Required\n    \"limit\": {...},      \u002F\u002F Optional with default\n    \"offset\": {...}      \u002F\u002F Optional\n  },\n  \"required\": [\"query\"],\n  \"additionalProperties\": false  \u002F\u002F Strict mode\n}\n\"\"\"\n\n### Tool with Input Examples\n\nUsing examples to guide LLM tool usage\n\n**When to use**: Complex tools with nested objects or format-sensitive inputs\n\n# TOOL USE EXAMPLES (Anthropic Beta Feature):\n\n\"\"\"\nExamples show Claude concrete patterns that schemas can't express.\nImproves accuracy from 72% to 90% on complex operations.\n\"\"\"\n\n{\n  \"name\": \"create_calendar_event\",\n  \"description\": \"Creates a calendar event with optional attendees and reminders\",\n  \"input_schema\": {\n    \"type\": \"object\",\n    \"properties\": {\n      \"title\": {\"type\": \"string\", \"description\": \"Event title\"},\n      \"start_time\": {\n        \"type\": \"string\",\n        \"description\": \"ISO 8601 datetime, e.g. 2024-03-15T14:00:00Z\"\n      },\n      \"duration_minutes\": {\"type\": \"integer\", \"description\": \"Event duration\"},\n      \"attendees\": {\n        \"type\": \"array\",\n        \"items\": {\"type\": \"string\"},\n        \"description\": \"Email addresses of attendees\"\n      }\n    },\n    \"required\": [\"title\", \"start_time\", \"duration_minutes\"]\n  },\n  \"input_examples\": [\n    {\n      \"title\": \"Team Standup\",\n      \"start_time\": \"2024-03-15T09:00:00Z\",\n      \"duration_minutes\": 30,\n      \"attendees\": [\"alice@company.com\", \"bob@company.com\"]\n    },\n    {\n      \"title\": \"Quick Chat\",\n      \"start_time\": \"2024-03-15T14:00:00Z\",\n      \"duration_minutes\": 15\n    },\n    {\n      \"title\": \"Project Review\",\n      \"start_time\": \"2024-03-15T16:00:00-05:00\",\n      \"duration_minutes\": 60,\n      \"attendees\": [\"team@company.com\"]\n    }\n  ]\n}\n\n# EXAMPLE DESIGN PRINCIPLES:\n# - Use realistic data, not placeholders\n# - Show minimal, partial, and full specification patterns\n# - Keep concise: 1-5 examples per tool\n# - Focus on ambiguous cases\n\n### Tool Error Handling\n\nReturning errors that help the LLM recover\n\n**When to use**: Any tool that can fail\n\n# ERROR HANDLING BEST PRACTICES:\n\n## Return Informative Errors\n\"\"\"\nBAD:\n{\"error\": \"Failed\"}\n{\"error\": true}\n\nGOOD:\n{\n  \"error\": true,\n  \"error_type\": \"not_found\",\n  \"message\": \"Location 'Atlantis' not found in weather database.\n    Please provide a real city name like 'San Francisco, CA'.\",\n  \"suggestions\": [\"San Francisco, CA\", \"Los Angeles, CA\"]\n}\n\"\"\"\n\n## Anthropic Tool Result with Error\n\"\"\"\n{\n  \"type\": \"tool_result\",\n  \"tool_use_id\": \"toolu_01A09q90qw90lq917835lq9\",\n  \"content\": \"Error: Location 'Atlantis' not found in weather database.\n    Please provide a real city name like 'San Francisco, CA'.\",\n  \"is_error\": true\n}\n\"\"\"\n\n## Error Categories to Handle\n\"\"\"\n1. Input Validation Errors\n   - Missing required parameters\n   - Invalid format\n   - Out of range values\n\n2. External Service Errors\n   - API unavailable\n   - Rate limited\n   - Timeout\n\n3. Business Logic Errors\n   - Resource not found\n   - Permission denied\n   - Conflict\u002Fduplicate\n\n4. Internal Errors\n   - Unexpected exceptions\n   - Data corruption\n\"\"\"\n\n## Implementation Pattern\n\"\"\"\nfrom dataclasses import dataclass\nfrom typing import Union\n\n@dataclass\nclass ToolResult:\n    success: bool\n    content: str\n    error_type: str = None\n    suggestions: list[str] = None\n\n    def to_response(self) -> dict:\n        if self.success:\n            return {\"content\": self.content}\n        return {\n            \"content\": f\"Error ({self.error_type}): {self.content}\",\n            \"is_error\": True\n        }\n\ndef get_weather(location: str) -> ToolResult:\n    # Validate input\n    if not location or len(location) \u003C 2:\n        return ToolResult(\n            success=False,\n            content=\"Location must be at least 2 characters\",\n            error_type=\"validation_error\"\n        )\n\n    try:\n        data = weather_api.fetch(location)\n        return ToolResult(\n            success=True,\n            content=f\"Temperature: {data.temp}°F, Conditions: {data.conditions}\"\n        )\n    except LocationNotFound:\n        return ToolResult(\n            success=False,\n            content=f\"Location '{location}' not found\",\n            error_type=\"not_found\",\n            suggestions=weather_api.suggest_locations(location)\n        )\n    except RateLimitError:\n        return ToolResult(\n            success=False,\n            content=\"Weather service rate limit exceeded. Try again in 60 seconds.\",\n            error_type=\"rate_limit\"\n        )\n    except Exception as e:\n        return ToolResult(\n            success=False,\n            content=f\"Unexpected error: {str(e)}\",\n            error_type=\"internal_error\"\n        )\n\"\"\"\n\n### MCP Tool Pattern\n\nBuilding tools using Model Context Protocol\n\n**When to use**: Creating reusable, cross-platform tools\n\n# MCP TOOL IMPLEMENTATION:\n\n\"\"\"\nMCP (Model Context Protocol) is Anthropic's open standard for\nconnecting AI agents to external systems. Build once, use everywhere.\n\"\"\"\n\n## Basic MCP Server (TypeScript)\n\"\"\"\nimport { Server } from \"@modelcontextprotocol\u002Fsdk\u002Fserver\";\nimport { StdioServerTransport } from \"@modelcontextprotocol\u002Fsdk\u002Fserver\u002Fstdio\";\n\nconst server = new Server({\n  name: \"weather-server\",\n  version: \"1.0.0\"\n});\n\n\u002F\u002F Define tools\nserver.setRequestHandler(\"tools\u002Flist\", async () => ({\n  tools: [\n    {\n      name: \"get_weather\",\n      description: \"Get current weather for a location. Returns\n        temperature, conditions, and humidity. Use for weather\n        queries about specific cities.\",\n      inputSchema: {\n        type: \"object\",\n        properties: {\n          location: {\n            type: \"string\",\n            description: \"City and state, e.g. 'San Francisco, CA'\"\n          },\n          unit: {\n            type: \"string\",\n            enum: [\"celsius\", \"fahrenheit\"],\n            default: \"fahrenheit\"\n          }\n        },\n        required: [\"location\"]\n      }\n    }\n  ]\n}));\n\n\u002F\u002F Handle tool calls\nserver.setRequestHandler(\"tools\u002Fcall\", async (request) => {\n  const { name, arguments: args } = request.params;\n\n  if (name === \"get_weather\") {\n    try {\n      const weather = await fetchWeather(args.location, args.unit);\n      return {\n        content: [\n          {\n            type: \"text\",\n            text: JSON.stringify(weather)\n          }\n        ]\n      };\n    } catch (error) {\n      return {\n        content: [\n          {\n            type: \"text\",\n            text: `Error: ${error.message}`\n          }\n        ],\n        isError: true\n      };\n    }\n  }\n\n  throw new Error(`Unknown tool: ${name}`);\n});\n\n\u002F\u002F Start server\nconst transport = new StdioServerTransport();\nawait server.connect(transport);\n\"\"\"\n\n## MCP Benefits\n\"\"\"\n- Universal compatibility across LLM providers\n- Reusable tool libraries\n- Streaming and SSE transport support\n- Built-in observability\n- Tool access controls\n\"\"\"\n\n### Tool Runner Pattern\n\nUsing SDK tool runners for automatic handling\n\n**When to use**: Building tool loops without manual management\n\n# TOOL RUNNER (Anthropic SDK Beta):\n\n\"\"\"\nThe tool runner handles the tool call loop automatically:\n- Executes tools when Claude calls them\n- Manages conversation state\n- Handles error retries\n- Provides streaming support\n\"\"\"\n\n## Python Example\n\"\"\"\nimport anthropic\nfrom anthropic import beta_tool\n\nclient = anthropic.Anthropic()\n\n@beta_tool\ndef get_weather(location: str, unit: str = \"fahrenheit\") -> str:\n    '''Get the current weather in a given location.\n\n    Args:\n        location: The city and state, e.g. San Francisco, CA\n        unit: Temperature unit, either 'celsius' or 'fahrenheit'\n    '''\n    # Implementation\n    return json.dumps({\"temperature\": \"72°F\", \"conditions\": \"Sunny\"})\n\n@beta_tool\ndef search_web(query: str) -> str:\n    '''Search the web for information.\n\n    Args:\n        query: The search query\n    '''\n    # Implementation\n    return json.dumps({\"results\": [...]})\n\n# Tool runner handles the loop\nrunner = client.beta.messages.tool_runner(\n    model=\"claude-sonnet-4-5\",\n    max_tokens=1024,\n    tools=[get_weather, search_web],\n    messages=[\n        {\"role\": \"user\", \"content\": \"What's the weather in Paris?\"}\n    ]\n)\n\n# Process each message\nfor message in runner:\n    print(message.content[0].text)\n\n# Or just get final result\nfinal = runner.until_done()\n\"\"\"\n\n## TypeScript with Zod\n\"\"\"\nimport { Anthropic } from '@anthropic-ai\u002Fsdk';\nimport { betaZodTool } from '@anthropic-ai\u002Fsdk\u002Fhelpers\u002Fbeta\u002Fzod';\nimport { z } from 'zod';\n\nconst anthropic = new Anthropic();\n\nconst getWeatherTool = betaZodTool({\n  name: 'get_weather',\n  description: 'Get the current weather in a given location',\n  inputSchema: z.object({\n    location: z.string().describe('City and state, e.g. San Francisco, CA'),\n    unit: z.enum(['celsius', 'fahrenheit']).default('fahrenheit')\n  }),\n  run: async (input) => {\n    \u002F\u002F Type-safe input!\n    return JSON.stringify({temperature: '72°F'});\n  }\n});\n\nconst runner = anthropic.beta.messages.toolRunner({\n  model: 'claude-sonnet-4-5',\n  max_tokens: 1024,\n  tools: [getWeatherTool],\n  messages: [{ role: 'user', content: \"What's the weather in Paris?\" }]\n});\n\nfor await (const message of runner) {\n  console.log(message.content[0].text);\n}\n\"\"\"\n\n### Parallel Tool Execution\n\nRunning multiple tools simultaneously\n\n**When to use**: Independent tool calls that can run in parallel\n\n# PARALLEL TOOL EXECUTION:\n\n\"\"\"\nBy default, Claude can call multiple tools in one response.\nThis dramatically reduces latency for independent operations.\n\"\"\"\n\n## Handling Parallel Results\n\"\"\"\n# Claude returns multiple tool_use blocks:\nresponse.content = [\n    {\"type\": \"text\", \"text\": \"I'll check both locations...\"},\n    {\"type\": \"tool_use\", \"id\": \"toolu_01\", \"name\": \"get_weather\",\n     \"input\": {\"location\": \"San Francisco, CA\"}},\n    {\"type\": \"tool_use\", \"id\": \"toolu_02\", \"name\": \"get_weather\",\n     \"input\": {\"location\": \"New York, NY\"}},\n    {\"type\": \"tool_use\", \"id\": \"toolu_03\", \"name\": \"get_time\",\n     \"input\": {\"timezone\": \"America\u002FLos_Angeles\"}},\n    {\"type\": \"tool_use\", \"id\": \"toolu_04\", \"name\": \"get_time\",\n     \"input\": {\"timezone\": \"America\u002FNew_York\"}}\n]\n\n# Execute in parallel\nimport asyncio\n\nasync def execute_tools_parallel(tool_uses):\n    tasks = [execute_tool(t) for t in tool_uses]\n    return await asyncio.gather(*tasks)\n\nresults = await execute_tools_parallel(tool_uses)\n\n# Return ALL results in SINGLE user message (critical!)\ntool_results = [\n    {\"type\": \"tool_result\", \"tool_use_id\": \"toolu_01\", \"content\": \"72°F, Sunny\"},\n    {\"type\": \"tool_result\", \"tool_use_id\": \"toolu_02\", \"content\": \"45°F, Cloudy\"},\n    {\"type\": \"tool_result\", \"tool_use_id\": \"toolu_03\", \"content\": \"2:30 PM PST\"},\n    {\"type\": \"tool_result\", \"tool_use_id\": \"toolu_04\", \"content\": \"5:30 PM EST\"}\n]\n\n# CORRECT: All results in one message\nmessages.append({\"role\": \"user\", \"content\": tool_results})\n\n# WRONG: Separate messages (breaks parallel execution pattern)\n# messages.append({\"role\": \"user\", \"content\": [tool_results[0]]})\n# messages.append({\"role\": \"user\", \"content\": [tool_results[1]]})\n\"\"\"\n\n## Encouraging Parallel Tool Use\n\"\"\"\nAdd to system prompt:\n\"For maximum efficiency, whenever you need to perform multiple\nindependent operations, invoke all relevant tools simultaneously\nrather than sequentially.\"\n\"\"\"\n\n## Disabling Parallel (When Needed)\n\"\"\"\nresponse = client.messages.create(\n    model=\"claude-sonnet-4-5\",\n    tools=tools,\n    tool_choice={\"type\": \"auto\", \"disable_parallel_tool_use\": True},\n    messages=messages\n)\n\"\"\"\n\n## Validation Checks\n\n### Tool Description Must Be Comprehensive\n\nSeverity: WARNING\n\nTool descriptions should be at least 100 characters\n\nMessage: Tool description is too short. Add details about when to use it, parameters, and return values.\n\n### Parameter Descriptions Required\n\nSeverity: WARNING\n\nEvery parameter should have a description\n\nMessage: Parameter missing description. Describe what it is and the expected format.\n\n### Schema Should Specify Required Fields\n\nSeverity: INFO\n\nExplicitly define which fields are required\n\nMessage: Schema doesn't specify required fields. Add 'required' array.\n\n### Tool Implementation Needs Error Handling\n\nSeverity: ERROR\n\nTool functions should handle exceptions\n\nMessage: Tool function without try\u002Fexcept block. Add error handling.\n\n### Error Results Need is_error Flag\n\nSeverity: WARNING\n\nWhen returning errors, set is_error to true\n\nMessage: Error result without is_error flag. Add 'is_error': true.\n\n### Tools Should Return Strings\n\nSeverity: WARNING\n\nReturn JSON string, not dict\u002Fobject\n\nMessage: Returning dict instead of string. Use json.dumps() or JSON.stringify().\n\n### Tools Should Validate Inputs\n\nSeverity: WARNING\n\nValidate LLM-provided inputs before execution\n\nMessage: Tool function without visible input validation. Validate before execution.\n\n### SQL Queries Must Use Parameterization\n\nSeverity: ERROR\n\nNever concatenate user input into SQL\n\nMessage: SQL query appears to use string concatenation. Use parameterized queries.\n\n### External Calls Need Timeouts\n\nSeverity: WARNING\n\nHTTP requests and external calls should have timeouts\n\nMessage: External API call without timeout. Add timeout parameter.\n\n### MCP Tools Must Have Input Schema\n\nSeverity: ERROR\n\nAll MCP tools require inputSchema\n\nMessage: MCP tool definition missing inputSchema.\n\n## Collaboration\n\n### Delegation Triggers\n\n- user needs to coordinate multiple tools -> multi-agent-orchestration (Tool orchestration across agents)\n- user needs persistent memory between tool calls -> agent-memory-systems (State management for tools)\n- user building voice agent tools -> voice-agents (Audio\u002Fvoice-specific tool requirements)\n- user needs computer control tools -> computer-use-agents (Desktop automation tools)\n- user wants to test their tools -> agent-evaluation (Tool testing and evaluation)\n\n## Related Skills\n\nWorks well with: `multi-agent-orchestration`, `api-designer`, `llm-architect`, `backend`\n\n## When to Use\n\n- User mentions or implies: agent tool\n- User mentions or implies: function calling\n- User mentions or implies: tool schema\n- User mentions or implies: tool design\n- User mentions or implies: mcp server\n- User mentions or implies: mcp tool\n- User mentions or implies: tool use\n- User mentions or implies: build tool for agent\n- User mentions or implies: define function\n- User mentions or implies: input_schema\n- User mentions or implies: tool_use\n- User mentions or implies: tool_result\n","# Agent Tool Builder\n\nTools are how AI agents interact with the world. A well-designed tool is the\ndifference between an agent that works and one that hallucinates, fails\nsilently, or costs 10x more tokens than necessary.\n\nThis skill covers tool design from schema to error handling. JSON Schema\nbest practices, description writing that actually helps the LLM, validation,\nand the emerging MCP standard that's becoming the lingua franca for AI tools.\n\nKey insight: Tool descriptions are more important than tool implementations.\nThe LLM never sees your code - it only sees the schema and description.\n\n## Principles\n\n- Description quality > implementation quality for LLM accuracy\n- Aim for fewer than 20 tools - more causes confusion\n- Every tool needs explicit error handling - silent failures poison agents\n- Return strings, not objects - LLMs process text\n- Validation gates before execution - reject, fix, or escalate, never silent fail\n- Test tools with the LLM, not just unit tests\n\n## Capabilities\n\n- agent-tools\n- function-calling\n- tool-schema-design\n- mcp-tools\n- tool-validation\n- tool-error-handling\n\n## Scope\n\n- multi-agent-coordination → multi-agent-orchestration\n- agent-memory → agent-memory-systems\n- api-design → api-designer\n- llm-prompting → prompt-engineering\n\n## Tooling\n\n### Standards\n\n- JSON Schema - When: All tool definitions Note: The universal format for tool schemas\n- MCP (Model Context Protocol) - When: Building reusable, cross-platform tools Note: Anthropic's open standard, widely adopted\n\n### Frameworks\n\n- Anthropic SDK - When: Claude-based agents Note: Beta tool runner handles most complexity\n- OpenAI Functions - When: OpenAI-based agents Note: Use strict mode for guaranteed schema compliance\n- Vercel AI SDK - When: Multi-provider tool handling Note: Abstracts differences between providers\n- LangChain Tools - When: LangChain-based agents Note: Converts MCP tools to LangChain format\n\n## Patterns\n\n### Tool Schema Design\n\nCreating clear, unambiguous JSON Schema for tools\n\n**When to use**: Defining any new tool for an agent\n\n# TOOL SCHEMA BEST PRACTICES:\n\n## 1. Detailed Descriptions (Most Important)\n\"\"\"\nBAD - Too vague:\n{\n  \"name\": \"get_stock_price\",\n  \"description\": \"Gets stock price\",\n  \"input_schema\": {\n    \"type\": \"object\",\n    \"properties\": {\n      \"ticker\": {\"type\": \"string\"}\n    }\n  }\n}\n\nGOOD - Comprehensive:\n{\n  \"name\": \"get_stock_price\",\n  \"description\": \"Retrieves the current stock price for a given ticker\n    symbol. The ticker symbol must be a valid symbol for a publicly\n    traded company on a major US stock exchange like NYSE or NASDAQ.\n    Returns the latest trade price in USD. Use when the user asks\n    about current or recent stock prices. Does NOT provide historical\n    data, company info, or predictions.\",\n  \"input_schema\": {\n    \"type\": \"object\",\n    \"properties\": {\n      \"ticker\": {\n        \"type\": \"string\",\n        \"description\": \"The stock ticker symbol, e.g. AAPL for Apple Inc.\"\n      }\n    },\n    \"required\": [\"ticker\"]\n  }\n}\n\"\"\"\n\n## 2. Parameter Descriptions\n\"\"\"\nEvery parameter needs:\n- What it is\n- Format expected\n- Example value\n- Edge cases\u002Flimitations\n\n{\n  \"location\": {\n    \"type\": \"string\",\n    \"description\": \"City and state\u002Fcountry. Format: 'City, State' for US\n      (e.g., 'San Francisco, CA') or 'City, Country' for international\n      (e.g., 'Tokyo, Japan'). Do not use ZIP codes or coordinates.\"\n  },\n  \"unit\": {\n    \"type\": \"string\",\n    \"enum\": [\"celsius\", \"fahrenheit\"],\n    \"description\": \"Temperature unit. Defaults to user's locale if not\n      specified. Use 'fahrenheit' for US users, 'celsius' for others.\"\n  }\n}\n\"\"\"\n\n## 3. Use Enums When Possible\n\"\"\"\nEnums constrain the LLM to valid values:\n\n\"priority\": {\n  \"type\": \"string\",\n  \"enum\": [\"low\", \"medium\", \"high\", \"critical\"],\n  \"description\": \"Task priority level\"\n}\n\n\"action\": {\n  \"type\": \"string\",\n  \"enum\": [\"create\", \"read\", \"update\", \"delete\"],\n  \"description\": \"The CRUD operation to perform\"\n}\n\"\"\"\n\n## 4. Required vs Optional\n\"\"\"\nBe explicit about what's required:\n\n{\n  \"type\": \"object\",\n  \"properties\": {\n    \"query\": {...},      \u002F\u002F Required\n    \"limit\": {...},      \u002F\u002F Optional with default\n    \"offset\": {...}      \u002F\u002F Optional\n  },\n  \"required\": [\"query\"],\n  \"additionalProperties\": false  \u002F\u002F Strict mode\n}\n\"\"\"\n\n### Tool with Input Examples\n\nUsing examples to guide LLM tool usage\n\n**When to use**: Complex tools with nested objects or format-sensitive inputs\n\n# TOOL USE EXAMPLES (Anthropic Beta Feature):\n\n\"\"\"\nExamples show Claude concrete patterns that schemas can't express.\nImproves accuracy from 72% to 90% on complex operations.\n\"\"\"\n\n{\n  \"name\": \"create_calendar_event\",\n  \"description\": \"Creates a calendar event with optional attendees and reminders\",\n  \"input_schema\": {\n    \"type\": \"object\",\n    \"properties\": {\n      \"title\": {\"type\": \"string\", \"description\": \"Event title\"},\n      \"start_time\": {\n        \"type\": \"string\",\n        \"description\": \"ISO 8601 datetime, e.g. 2024-03-15T14:00:00Z\"\n      },\n      \"duration_minutes\": {\"type\": \"integer\", \"description\": \"Event duration\"},\n      \"attendees\": {\n        \"type\": \"array\",\n        \"items\": {\"type\": \"string\"},\n        \"description\": \"Email addresses of attendees\"\n      }\n    },\n    \"required\": [\"title\", \"start_time\", \"duration_minutes\"]\n  },\n  \"input_examples\": [\n    {\n      \"title\": \"Team Standup\",\n      \"start_time\": \"2024-03-15T09:00:00Z\",\n      \"duration_minutes\": 30,\n      \"attendees\": [\"alice@company.com\", \"bob@company.com\"]\n    },\n    {\n      \"title\": \"Quick Chat\",\n      \"start_time\": \"2024-03-15T14:00:00Z\",\n      \"duration_minutes\": 15\n    },\n    {\n      \"title\": \"Project Review\",\n      \"start_time\": \"2024-03-15T16:00:00-05:00\",\n      \"duration_minutes\": 60,\n      \"attendees\": [\"team@company.com\"]\n    }\n  ]\n}\n\n# EXAMPLE DESIGN PRINCIPLES:\n# - Use realistic data, not placeholders\n# - Show minimal, partial, and full specification patterns\n# - Keep concise: 1-5 examples per tool\n# - Focus on ambiguous cases\n\n### Tool Error Handling\n\nReturning errors that help the LLM recover\n\n**When to use**: Any tool that can fail\n\n# ERROR HANDLING BEST PRACTICES:\n\n## Return Informative Errors\n\"\"\"\nBAD:\n{\"error\": \"Failed\"}\n{\"error\": true}\n\nGOOD:\n{\n  \"error\": true,\n  \"error_type\": \"not_found\",\n  \"message\": \"Location 'Atlantis' not found in weather database.\n    Please provide a real city name like 'San Francisco, CA'.\",\n  \"suggestions\": [\"San Francisco, CA\", \"Los Angeles, CA\"]\n}\n\"\"\"\n\n## Anthropic Tool Result with Error\n\"\"\"\n{\n  \"type\": \"tool_result\",\n  \"tool_use_id\": \"toolu_01A09q90qw90lq917835lq9\",\n  \"content\": \"Error: Location 'Atlantis' not found in weather database.\n    Please provide a real city name like 'San Francisco, CA'.\",\n  \"is_error\": true\n}\n\"\"\"\n\n## Error Categories to Handle\n\"\"\"\n1. Input Validation Errors\n   - Missing required parameters\n   - Invalid format\n   - Out of range values\n\n2. External Service Errors\n   - API unavailable\n   - Rate limited\n   - Timeout\n\n3. Business Logic Errors\n   - Resource not found\n   - Permission denied\n   - Conflict\u002Fduplicate\n\n4. Internal Errors\n   - Unexpected exceptions\n   - Data corruption\n\"\"\"\n\n## Implementation Pattern\n\"\"\"\nfrom dataclasses import dataclass\nfrom typing import Union\n\n@dataclass\nclass ToolResult:\n    success: bool\n    content: str\n    error_type: str = None\n    suggestions: list[str] = None\n\n    def to_response(self) -> dict:\n        if self.success:\n            return {\"content\": self.content}\n        return {\n            \"content\": f\"Error ({self.error_type}): {self.content}\",\n            \"is_error\": True\n        }\n\ndef get_weather(location: str) -> ToolResult:\n    # Validate input\n    if not location or len(location) \u003C 2:\n        return ToolResult(\n            success=False,\n            content=\"Location must be at least 2 characters\",\n            error_type=\"validation_error\"\n        )\n\n    try:\n        data = weather_api.fetch(location)\n        return ToolResult(\n            success=True,\n            content=f\"Temperature: {data.temp}°F, Conditions: {data.conditions}\"\n        )\n    except LocationNotFound:\n        return ToolResult(\n            success=False,\n            content=f\"Location '{location}' not found\",\n            error_type=\"not_found\",\n            suggestions=weather_api.suggest_locations(location)\n        )\n    except RateLimitError:\n        return ToolResult(\n            success=False,\n            content=\"Weather service rate limit exceeded. Try again in 60 seconds.\",\n            error_type=\"rate_limit\"\n        )\n    except Exception as e:\n        return ToolResult(\n            success=False,\n            content=f\"Unexpected error: {str(e)}\",\n            error_type=\"internal_error\"\n        )\n\"\"\"\n\n### MCP Tool Pattern\n\nBuilding tools using Model Context Protocol\n\n**When to use**: Creating reusable, cross-platform tools\n\n# MCP TOOL IMPLEMENTATION:\n\n\"\"\"\nMCP (Model Context Protocol) is Anthropic's open standard for\nconnecting AI agents to external systems. Build once, use everywhere.\n\"\"\"\n\n## Basic MCP Server (TypeScript)\n\"\"\"\nimport { Server } from \"@modelcontextprotocol\u002Fsdk\u002Fserver\";\nimport { StdioServerTransport } from \"@modelcontextprotocol\u002Fsdk\u002Fserver\u002Fstdio\";\n\nconst server = new Server({\n  name: \"weather-server\",\n  version: \"1.0.0\"\n});\n\n\u002F\u002F Define tools\nserver.setRequestHandler(\"tools\u002Flist\", async () => ({\n  tools: [\n    {\n      name: \"get_weather\",\n      description: \"Get current weather for a location. Returns\n        temperature, conditions, and humidity. Use for weather\n        queries about specific cities.\",\n      inputSchema: {\n        type: \"object\",\n        properties: {\n          location: {\n            type: \"string\",\n            description: \"City and state, e.g. 'San Francisco, CA'\"\n          },\n          unit: {\n            type: \"string\",\n            enum: [\"celsius\", \"fahrenheit\"],\n            default: \"fahrenheit\"\n          }\n        },\n        required: [\"location\"]\n      }\n    }\n  ]\n}));\n\n\u002F\u002F Handle tool calls\nserver.setRequestHandler(\"tools\u002Fcall\", async (request) => {\n  const { name, arguments: args } = request.params;\n\n  if (name === \"get_weather\") {\n    try {\n      const weather = await fetchWeather(args.location, args.unit);\n      return {\n        content: [\n          {\n            type: \"text\",\n            text: JSON.stringify(weather)\n          }\n        ]\n      };\n    } catch (error) {\n      return {\n        content: [\n          {\n            type: \"text\",\n            text: `Error: ${error.message}`\n          }\n        ],\n        isError: true\n      };\n    }\n  }\n\n  throw new Error(`Unknown tool: ${name}`);\n});\n\n\u002F\u002F Start server\nconst transport = new StdioServerTransport();\nawait server.connect(transport);\n\"\"\"\n\n## MCP Benefits\n\"\"\"\n- Universal compatibility across LLM providers\n- Reusable tool libraries\n- Streaming and SSE transport support\n- Built-in observability\n- Tool access controls\n\"\"\"\n\n### Tool Runner Pattern\n\nUsing SDK tool runners for automatic handling\n\n**When to use**: Building tool loops without manual management\n\n# TOOL RUNNER (Anthropic SDK Beta):\n\n\"\"\"\nThe tool runner handles the tool call loop automatically:\n- Executes tools when Claude calls them\n- Manages conversation state\n- Handles error retries\n- Provides streaming support\n\"\"\"\n\n## Python Example\n\"\"\"\nimport anthropic\nfrom anthropic import beta_tool\n\nclient = anthropic.Anthropic()\n\n@beta_tool\ndef get_weather(location: str, unit: str = \"fahrenheit\") -> str:\n    '''Get the current weather in a given location.\n\n    Args:\n        location: The city and state, e.g. San Francisco, CA\n        unit: Temperature unit, either 'celsius' or 'fahrenheit'\n    '''\n    # Implementation\n    return json.dumps({\"temperature\": \"72°F\", \"conditions\": \"Sunny\"})\n\n@beta_tool\ndef search_web(query: str) -> str:\n    '''Search the web for information.\n\n    Args:\n        query: The search query\n    '''\n    # Implementation\n    return json.dumps({\"results\": [...]})\n\n# Tool runner handles the loop\nrunner = client.beta.messages.tool_runner(\n    model=\"claude-sonnet-4-5\",\n    max_tokens=1024,\n    tools=[get_weather, search_web],\n    messages=[\n        {\"role\": \"user\", \"content\": \"What's the weather in Paris?\"}\n    ]\n)\n\n# Process each message\nfor message in runner:\n    print(message.content[0].text)\n\n# Or just get final result\nfinal = runner.until_done()\n\"\"\"\n\n## TypeScript with Zod\n\"\"\"\nimport { Anthropic } from '@anthropic-ai\u002Fsdk';\nimport { betaZodTool } from '@anthropic-ai\u002Fsdk\u002Fhelpers\u002Fbeta\u002Fzod';\nimport { z } from 'zod';\n\nconst anthropic = new Anthropic();\n\nconst getWeatherTool = betaZodTool({\n  name: 'get_weather',\n  description: 'Get the current weather in a given location',\n  inputSchema: z.object({\n    location: z.string().describe('City and state, e.g. San Francisco, CA'),\n    unit: z.enum(['celsius', 'fahrenheit']).default('fahrenheit')\n  }),\n  run: async (input) => {\n    \u002F\u002F Type-safe input!\n    return JSON.stringify({temperature: '72°F'});\n  }\n});\n\nconst runner = anthropic.beta.messages.toolRunner({\n  model: 'claude-sonnet-4-5',\n  max_tokens: 1024,\n  tools: [getWeatherTool],\n  messages: [{ role: 'user', content: \"What's the weather in Paris?\" }]\n});\n\nfor await (const message of runner) {\n  console.log(message.content[0].text);\n}\n\"\"\"\n\n### Parallel Tool Execution\n\nRunning multiple tools simultaneously\n\n**When to use**: Independent tool calls that can run in parallel\n\n# PARALLEL TOOL EXECUTION:\n\n\"\"\"\nBy default, Claude can call multiple tools in one response.\nThis dramatically reduces latency for independent operations.\n\"\"\"\n\n## Handling Parallel Results\n\"\"\"\n# Claude returns multiple tool_use blocks:\nresponse.content = [\n    {\"type\": \"text\", \"text\": \"I'll check both locations...\"},\n    {\"type\": \"tool_use\", \"id\": \"toolu_01\", \"name\": \"get_weather\",\n     \"input\": {\"location\": \"San Francisco, CA\"}},\n    {\"type\": \"tool_use\", \"id\": \"toolu_02\", \"name\": \"get_weather\",\n     \"input\": {\"location\": \"New York, NY\"}},\n    {\"type\": \"tool_use\", \"id\": \"toolu_03\", \"name\": \"get_time\",\n     \"input\": {\"timezone\": \"America\u002FLos_Angeles\"}},\n    {\"type\": \"tool_use\", \"id\": \"toolu_04\", \"name\": \"get_time\",\n     \"input\": {\"timezone\": \"America\u002FNew_York\"}}\n]\n\n# Execute in parallel\nimport asyncio\n\nasync def execute_tools_parallel(tool_uses):\n    tasks = [execute_tool(t) for t in tool_uses]\n    return await asyncio.gather(*tasks)\n\nresults = await execute_tools_parallel(tool_uses)\n\n# Return ALL results in SINGLE user message (critical!)\ntool_results = [\n    {\"type\": \"tool_result\", \"tool_use_id\": \"toolu_01\", \"content\": \"72°F, Sunny\"},\n    {\"type\": \"tool_result\", \"tool_use_id\": \"toolu_02\", \"content\": \"45°F, Cloudy\"},\n    {\"type\": \"tool_result\", \"tool_use_id\": \"toolu_03\", \"content\": \"2:30 PM PST\"},\n    {\"type\": \"tool_result\", \"tool_use_id\": \"toolu_04\", \"content\": \"5:30 PM EST\"}\n]\n\n# CORRECT: All results in one message\nmessages.append({\"role\": \"user\", \"content\": tool_results})\n\n# WRONG: Separate messages (breaks parallel execution pattern)\n# messages.append({\"role\": \"user\", \"content\": [tool_results[0]]})\n# messages.append({\"role\": \"user\", \"content\": [tool_results[1]]})\n\"\"\"\n\n## Encouraging Parallel Tool Use\n\"\"\"\nAdd to system prompt:\n\"For maximum efficiency, whenever you need to perform multiple\nindependent operations, invoke all relevant tools simultaneously\nrather than sequentially.\"\n\"\"\"\n\n## Disabling Parallel (When Needed)\n\"\"\"\nresponse = client.messages.create(\n    model=\"claude-sonnet-4-5\",\n    tools=tools,\n    tool_choice={\"type\": \"auto\", \"disable_parallel_tool_use\": True},\n    messages=messages\n)\n\"\"\"\n\n## Validation Checks\n\n### Tool Description Must Be Comprehensive\n\nSeverity: WARNING\n\nTool descriptions should be at least 100 characters\n\nMessage: Tool description is too short. Add details about when to use it, parameters, and return values.\n\n### Parameter Descriptions Required\n\nSeverity: WARNING\n\nEvery parameter should have a description\n\nMessage: Parameter missing description. Describe what it is and the expected format.\n\n### Schema Should Specify Required Fields\n\nSeverity: INFO\n\nExplicitly define which fields are required\n\nMessage: Schema doesn't specify required fields. Add 'required' array.\n\n### Tool Implementation Needs Error Handling\n\nSeverity: ERROR\n\nTool functions should handle exceptions\n\nMessage: Tool function without try\u002Fexcept block. Add error handling.\n\n### Error Results Need is_error Flag\n\nSeverity: WARNING\n\nWhen returning errors, set is_error to true\n\nMessage: Error result without is_error flag. Add 'is_error': true.\n\n### Tools Should Return Strings\n\nSeverity: WARNING\n\nReturn JSON string, not dict\u002Fobject\n\nMessage: Returning dict instead of string. Use json.dumps() or JSON.stringify().\n\n### Tools Should Validate Inputs\n\nSeverity: WARNING\n\nValidate LLM-provided inputs before execution\n\nMessage: Tool function without visible input validation. Validate before execution.\n\n### SQL Queries Must Use Parameterization\n\nSeverity: ERROR\n\nNever concatenate user input into SQL\n\nMessage: SQL query appears to use string concatenation. Use parameterized queries.\n\n### External Calls Need Timeouts\n\nSeverity: WARNING\n\nHTTP requests and external calls should have timeouts\n\nMessage: External API call without timeout. Add timeout parameter.\n\n### MCP Tools Must Have Input Schema\n\nSeverity: ERROR\n\nAll MCP tools require inputSchema\n\nMessage: MCP tool definition missing inputSchema.\n\n## Collaboration\n\n### Delegation Triggers\n\n- user needs to coordinate multiple tools -> multi-agent-orchestration (Tool orchestration across agents)\n- user needs persistent memory between tool calls -> agent-memory-systems (State management for tools)\n- user building voice agent tools -> voice-agents (Audio\u002Fvoice-specific tool requirements)\n- user needs computer control tools -> computer-use-agents (Desktop automation tools)\n- user wants to test their tools -> agent-evaluation (Tool testing and evaluation)\n\n## Related Skills\n\nWorks well with: `multi-agent-orchestration`, `api-designer`, `llm-architect`, `backend`\n\n## When to Use\n\n- User mentions or implies: agent tool\n- User mentions or implies: function calling\n- User mentions or implies: tool schema\n- User mentions or implies: tool design\n- User mentions or implies: mcp server\n- User mentions or implies: mcp tool\n- User mentions or implies: tool use\n- User mentions or implies: build tool for agent\n- User mentions or implies: define function\n- User mentions or implies: input_schema\n- User mentions or implies: tool_use\n- User mentions or implies: tool_result\n",{"name":5,"risk":42,"source":52,"date_added":53,"description":7},"vibeship-spawner-skills (Apache 2.0)",1772150400000,"main","7a38d8e7073e61c16eda85681cb32fb68a8657a5","skills-store\u002Fglobal\u002Fagent-tool-builder","https:\u002F\u002Fgithub.com\u002Fsapsapshen\u002Fai-memory-router","https:\u002F\u002Fgithub.com\u002Fsapsapshen","2026-04-14 21:12:12",{"source":8,"skillName":5,"commands":61,"prompt":71,"zipAvailable":72},[62,65,68],{"tool":63,"command":64},"npx","npx skills add sapsapshen\u002Fai-memory-router --skill agent-tool-builder",{"tool":66,"command":67},"bunx","bunx skills add sapsapshen\u002Fai-memory-router --skill agent-tool-builder",{"tool":69,"command":70},"pnpm","pnpm dlx skills add sapsapshen\u002Fai-memory-router --skill agent-tool-builder","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002Fsapsapshen\u002Fai-memory-router\u002Fblob\u002Fmain\u002Fskills-store\u002Fglobal\u002Fagent-tool-builder\u002FSKILL.md",true,[],[75,98,124,134,147,164],{"skillNo":76,"name":77,"nameZh":6,"description":78,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":79,"categories":80,"keywords":86,"publishedAt":47},"01M71ATQRKG48SH06FQ697T5G9","wechat-search","Search WeChat Official Account articles using OpenClaw's web search and fetch capabilities with compliance-focused design.","https:\u002F\u002Fgithub.com\u002Fsapsapshen\u002Fai-memory-router\u002Fblob\u002Fmain\u002Fskills-store\u002Fopenclaw\u002Fwechat-search-release\u002FSKILL.md",[81],{"categoryNo":82,"categoryName":83,"taxonomyType":84,"taxonomyTypeDesc":85},"01JBSEEDCAT000000000000037","律师与法务","PROFESSION","适用职业",[87,88,89,90,91,45,92,93,94,95,96,97],"wechat search","wechat","search","capabilities","openclaw","web search","compliance","web","official","fetch","with",{"skillNo":99,"name":100,"nameZh":6,"description":101,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":102,"categories":103,"keywords":104,"publishedAt":47},"01M71ATNAVGZB35PFYM2GC5N00","trend-tap","Real-time trending topics aggregator across 7 platforms (X\u002FTwitter, Reddit, Google Trends, Hacker News, Zhihu, Bilibili, Weibo). Trigger: when user says 'trend', 'trends', 'trend tap', '热点', '热搜', or asks what is trending\u002Fhot. Do NOT use web_search — this skill already fetches live data from all platforms with zero API keys.","https:\u002F\u002Fgithub.com\u002Fsapsapshen\u002Fai-memory-router\u002Fblob\u002Fmain\u002Fskills-store\u002Fopenclaw\u002Ftrend-tap\u002FSKILL.md",[],[105,106,107,108,109,110,89,111,112,113,114,115,116,117,118,119,120,121,122,123],"trend","trend tap","tap","api keys","Bash","time","trending","hacker","aggregator","do","when","brew","live","data","google","install","read","across","api",{"skillNo":125,"name":126,"nameZh":6,"description":127,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":128,"categories":129,"keywords":130,"publishedAt":47},"01M71ATJ9AM5KM15AA8PS302EB","trend-radar","Real-time trending topics aggregator across 7 platforms (X\u002FTwitter, Reddit, Google Trends, Hacker News, Zhihu, Bilibili, Weibo). Trigger: when user says 'trend', 'trends', 'trend pulse', '热点', '热搜', or asks what is trending\u002Fhot. Do NOT use web_search — this skill already fetches live data from all platforms with zero API keys.","https:\u002F\u002Fgithub.com\u002Fsapsapshen\u002Fai-memory-router\u002Fblob\u002Fmain\u002Fskills-store\u002Fopenclaw\u002Ftrend-radar\u002FSKILL.md",[],[131,132,105,116,117,118,119,109,89,121,133,122,123,115,113,112,111,120,110,108],"trend radar","radar","pulse",{"skillNo":135,"name":136,"nameZh":6,"description":137,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":138,"categories":139,"keywords":140,"publishedAt":47},"01M71ATGF095N3EJM9G8SD0BB5","tech-news-daily","每日科技新闻汇——从全球多个 RSS 源抓取最新科技新闻，翻译（英→中）、去重、排版成每日简报。当用户询问科技新闻、科技新鲜事、每日科技、技术动态、Daily Tech News 或类似表达时触发此技能。","https:\u002F\u002Fgithub.com\u002Fsapsapshen\u002Fai-memory-router\u002Fblob\u002Fmain\u002Fskills-store\u002Fopenclaw\u002Ftech-news-daily\u002FSKILL.md",[],[141,142,143,144,145,146],"news","tech","tech news daily","daily","rss","翻译",{"skillNo":148,"name":149,"nameZh":6,"description":150,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":151,"categories":152,"keywords":156,"publishedAt":47},"01M71ATFAEAMQ17M70NQNY7SR8","summarize","Summarize URLs or files with the summarize CLI (web, PDFs, images, audio, YouTube).","https:\u002F\u002Fgithub.com\u002Fsapsapshen\u002Fai-memory-router\u002Fblob\u002Fmain\u002Fskills-store\u002Fopenclaw\u002Fsummarize\u002FSKILL.md",[153],{"categoryNo":154,"categoryName":155,"taxonomyType":21,"taxonomyTypeDesc":22},"01JBSEEDCAT000000000000021","资料检索与研究",[149,94,157,158,159,97,116,160,161,162,107,163,120],"Audio","cli","youtube","images","https","files","pdfs",{"skillNo":165,"name":166,"nameZh":6,"description":167,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":168,"categories":169,"keywords":170,"publishedAt":47},"01M71ATDNS3R0M4H936PACQAWZ","sonoscli","Control Sonos speakers (discover\u002Fstatus\u002Fplay\u002Fvolume\u002Fgroup).","https:\u002F\u002Fgithub.com\u002Fsapsapshen\u002Fai-memory-router\u002Fblob\u002Fmain\u002Fskills-store\u002Fopenclaw\u002Fsonoscli\u002FSKILL.md",[],[166,171,172,173,120,174,161,175,176,177,178,179],"status","play","go","latest","github","discover","control","cmd","volume"]