Token导航 LogoToken导航

Agent Skill · 研究检索

openclaw-plugin-creator-agent-skill(OpenClaw plugin creator Agent 技能)

openclaw-plugin-creator-agent-skill 用于查找、检索和筛选相关信息,适合在 OpenClaw 中需要根据关键词、任务场景或来源线索快速定位候选结果时使用。可结合来源仓库、安装命令和原始 README 继续核验具体用法。安装前建议确认权限范围、维护状态,以及是否会触发联网、命令执行或文件读写。

openclaw-plugin-creator-agent-skill 是研究检索类 Agent Skill,可用宿主为 OpenClaw,通过ClawHub获取,权限为接触敏感数据;许可证为 MIT-0,第三方扫描有提醒项,按安装量在研究检索类的 73,675 个 Skill 里排第 40,339。

总安装 1,297周安装 53GitHub Stars 公开资料未说明下载 416核验 2026-05-01

安装说明

本站只整理中文说明,不托管安装包,也不代用户安装。
GitHub
许可证
MIT-0
版本
1.0.0

安装方式

通过对话安装

复制提示词发给支持本地命令或 Skills 的 AI 助手,先确认命令和权限,再让它执行。

请帮我安装这个 Agent Skill:openclaw-plugin-creator-agent-skill(OpenClaw plugin creator Agent 技能)
来源仓库:https://github.com/pathanaawej0-dot/openclaw-plugin-creator-agent-skill
安装命令:
openclaw skills install openclaw-plugin-creator-agent-skill
安装前请先检查当前环境是否支持对应 CLI,并向我确认将要执行的命令、安装目录、联网范围和文件读写权限;确认后再执行。

命令行安装

复制命令到本机终端执行。该命令会通过 OpenClaw 从第三方来源获取 Skill;本站只展示命令,不托管安装包,也不自动执行。

ClawHubOpenClaw
openclaw skills install openclaw-plugin-creator-agent-skill

这个 Skill 是做什么的

研究检索类 Skill 主要解决「找资料、核对事实、把来源整理清楚」的问题,通常会让助手先检索再总结。

OpenClaw plugin creator 用于从零构建 OpenClaw 插件。适合在 OpenClaw 中开发工具注册、通道插件或配置模式时使用。提供清单生成、入口点设置和向导式配置支持。通过 clawhub 安装,结合来源仓库和 README 可进一步了解具体用法。

安装前需确认权限范围、维护状态,以及是否涉及联网、命令执行或文件读写。

上手步骤

4 步
  1. 1

    确认自己的宿主工具

    公开资料标注支持 OpenClaw;先确认自己用的是其中之一,再按对应工具的 Skills 目录放置。

  2. 2

    拿到安装命令

    可以直接复制的命令是 openclaw skills install openclaw-plugin-creator-agent-skill;命令里的占位符换成自己的路径或密钥再执行。

  3. 3

    安装后重启宿主

    命令执行完要完全退出宿主工具再打开,让新装的 Skill 被重新扫描到,否则列表里可能看不到。

  4. 4

    按最小权限试用

    它需要接触敏感数据,先在测试目录或测试账号里跑,确认没问题再放到正式项目。

常见问题

6 条

openclaw-plugin-creator-agent-skill 怎么安装?

在终端执行 openclaw skills install openclaw-plugin-creator-agent-skill。装完完全退出宿主工具再打开,让它重新扫描本地 Skill。

openclaw-plugin-creator-agent-skill 支持哪些 AI 工具?

公开资料标注支持 OpenClaw。其他工具只要能读取本地 Skill 目录,一般也能手动放进去,但要自己确认触发方式是否一致。

openclaw-plugin-creator-agent-skill 需要什么权限?

权限级别是接触敏感数据。该 Skill 可能接触密钥、Token、环境变量或敏感配置,应进入高风险复核队列,默认不自动发布。

openclaw-plugin-creator-agent-skill 是免费的吗?

来源标注的许可证是 MIT-0。本站不代收任何费用,也不托管安装包;能否商用、要不要署名,以许可证原文为准。

openclaw-plugin-creator-agent-skill 用的人多吗?

公开数据是累计安装 1,297 次,近一周安装 53 次,下载 416 次。这些数字来自第三方目录统计,只能当作热度参考,不代表质量或安全性。

openclaw-plugin-creator-agent-skill 安全吗?

第三方扫描有提醒项,说明存在需要人工确认的行为。安装前请打开源码看清它读了什么、联网到哪里。

值得选的地方和要先想清楚的地方

值得选的地方

  • 专注 OpenClaw 一种宿主,触发方式和使用习惯不用来回切换。
  • 许可证明确为 MIT-0,商用和分发边界好判断。

先想清楚的地方

  • 需要接触敏感数据,装到常用环境前要先评估影响面。
  • 第三方扫描存在提醒项或未通过项,需要人工逐条确认。
  • 第三方目录整理,版本更新和行为细节以来源仓库为准。

适不适合你

更适合

  • 已经在用 OpenClaw,想直接扩展它能力的人。
  • 需要处理研究检索类任务、又不想每次都手写提示词的人。

不太适合

  • 环境里有大量密钥、凭证,不方便给第三方 Skill 读取的人。
  • 需要厂商级技术支持和 SLA 的生产系统。

研究检索类 Skill 对比

6 个同类
Skill总安装GitHub 星权限许可证扫描
find-skills

寻找技能 · Codex、Claude

31,199,81316,486操作浏览器MIT有提醒
self-improving-agent

自我改进 Agent · OpenClaw

16,464,0023,396接触敏感数据MIT-0通过
soultrace

灵魂追踪 · Codex、Claude

7,588,94412需要联网未明确有提醒
self-improving

自我完善 · OpenClaw

6,829,2841,046需要联网MIT-0通过
microsoft-foundry

微软代工厂 · Codex、Claude

6,614,144764调用外部服务未明确有提醒
openclaw-plugin-creator-agent-skill

OpenClaw plugin creator Agent 技能 · 本页

1,297—接触敏感数据MIT-0有提醒

按安装量取同分类前几名;数字来自第三方目录统计,只作热度参考,不代表质量结论。

SKILL.md 原文

name
openclaw-plugin-creator-skill
description
Build OpenClaw plugins from scratch — including the manifest, entry point, tool registration, channel plugins, provider plugins, config schema, setup wizards, runtime helpers, and publishing steps. Use this skill whenever the user wants to create, extend, debug, or publish any OpenClaw plugin, skill, or hook.
version
1.0.0
homepage
https://github.com/pathanaawej0-dot/openclaw-plugin-creation-skill
metadata
clawdbot
true

OpenClaw Plugin Development — Expert Reference

🛡️ Security, Privacy & Trust

This skill is purely instructional and does not execute remote code, access external endpoints (other than documentation), or modify system files without explicit user direction. All generated code should be reviewed before execution.

  • Data Privacy: This skill does not collect or transmit user data.
  • External Calls: No hidden background network requests.
  • Trust Statement: This skill follows OpenClaw's security best practices for plugin development.

What can a plugin do?

Plugins run in-process with the Gateway (via jiti, TypeScript loaded at runtime). They extend OpenClaw with:

CapabilityRegistration method
Messaging channelapi.registerChannel(...)
LLM / model providerapi.registerProvider(...)
Agent tool (LLM-callable)api.registerTool(tool, opts?)
Text-to-speech / STTapi.registerSpeechProvider(...)
Image generationapi.registerImageGenerationProvider(...)
Media understandingapi.registerMediaUnderstandingProvider(...)
Web searchapi.registerWebSearchProvider(...)
Custom CLI commandapi.registerCommand(def)
Event hookapi.registerHook(events, handler, opts?)
Gateway HTTP routeapi.registerHttpRoute(params)
Gateway RPC methodapi.registerGatewayMethod(name, handler)
CLI subcommandapi.registerCli(registrar, opts?)
Background serviceapi.registerService(service)
Interactive handlerapi.registerInteractiveHandler(registration)
Context engine (exclusive)api.registerContextEngine(id, factory)
Memory prompt sectionapi.registerMemoryPromptSection(builder)
Treat plugins as trusted code — they run with full Gateway access.

Plugin formats

FormatHow it works
Nativeopenclaw.plugin.json + TypeScript module — official plugins, npm packages
BundleCodex/Claude/Cursor-compatible layout (.codex-plugin/, .claude-plugin/, .cursor-plugin/)

This skill covers native plugins. For bundles see the Bundles docs.


Step 1 — File structure

my-plugin/
├── openclaw.plugin.json   ← REQUIRED manifest
├── package.json           ← npm package + openclaw metadata
├── index.ts               ← main entry point
├── setup-entry.ts         ← lightweight setup-only entry (optional, channels)
└── src/
    ├── channel.ts         ← channel plugin object (if channel plugin)
    ├── runtime.ts         ← createPluginRuntimeStore (if needed)
    ├── tools.ts
    └── provider.test.ts

Internal barrel convention:

api.ts          ← public exports for external consumers
runtime-api.ts  ← internal-only runtime exports

Never import your own plugin through openclaw/plugin-sdk/<your-plugin>. Route internal imports through ./api.ts.


Step 2 — package.json

Tool / provider plugin:

{
  "name": "@myorg/openclaw-my-plugin",
  "version": "1.0.0",
  "type": "module",
  "openclaw": {
    "extensions": ["./index.ts"],
    "providers": ["my-provider"]
  }
}

Channel plugin:

{
  "name": "@myorg/openclaw-my-channel",
  "version": "1.0.0",
  "type": "module",
  "openclaw": {
    "extensions": ["./index.ts"],
    "setupEntry": "./setup-entry.ts",
    "channel": {
      "id": "my-channel",
      "label": "My Channel",
      "blurb": "Short description shown in onboarding."
    },
    "startup": {
      "deferConfiguredChannelFullLoadUntilAfterListen": true
    }
  }
}

openclaw field reference:

FieldTypeDescription
extensionsstring[]Entry point files (relative paths)
setupEntrystringLightweight setup-only entry (optional)
channelobjectChannel metadata: id, label, blurb, selectionLabel, docsPath, order, aliases
providersstring[]Provider ids registered by this plugin
installobjectInstall hints: npmSpec, localPath, defaultChoice
startupobjectStartup behavior flags

Step 3 — openclaw.plugin.json (Manifest)

Every native plugin MUST have this file. OpenClaw validates config from it WITHOUT executing plugin code.

Minimal (no config):

{
  "id": "my-plugin",
  "configSchema": {
    "type": "object",
    "additionalProperties": false,
    "properties": {}
  }
}

Full example (provider with auth):

{
  "id": "my-provider",
  "name": "My Provider",
  "description": "Adds My Provider LLM to OpenClaw",
  "version": "1.0.0",
  "providers": ["my-provider"],
  "providerAuthEnvVars": {
    "my-provider": ["MY_PROVIDER_API_KEY"]
  },
  "providerAuthChoices": [
    {
      "provider": "my-provider",
      "method": "api-key",
      "choiceId": "my-provider-api-key",
      "choiceLabel": "My Provider API key",
      "optionKey": "myProviderApiKey",
      "cliFlag": "--my-provider-api-key",
      "cliOption": "--my-provider-api-key <key>",
      "cliDescription": "My Provider API key",
      "onboardingScopes": ["text-inference"]
    }
  ],
  "uiHints": {
    "apiKey": {
      "label": "API Key",
      "placeholder": "sk-...",
      "sensitive": true,
      "help": "From your My Provider dashboard"
    }
  },
  "configSchema": {
    "type": "object",
    "additionalProperties": false,
    "properties": {
      "apiKey": { "type": "string" }
    },
    "required": ["apiKey"]
  }
}

Full manifest field reference → see references/manifest.md


Step 4 — Entry Points

Import rules (CRITICAL)

// ✅ ALWAYS use focused subpaths
import { definePluginEntry } from "openclaw/plugin-sdk/plugin-entry";
import { defineChannelPluginEntry } from "openclaw/plugin-sdk/core";
import { createPluginRuntimeStore } from "openclaw/plugin-sdk/runtime-store";

// ❌ NEVER use monolithic root (deprecated, will be removed)
import { ... } from "openclaw/plugin-sdk";

Full subpath reference → see references/sdk-subpaths.md

Tool / Hook / Provider plugin (definePluginEntry)

import { definePluginEntry, type OpenClawPluginApi } from "openclaw/plugin-sdk/plugin-entry";
import { Type } from "@sinclair/typebox";

export default definePluginEntry({
  id: "my-plugin",
  name: "My Plugin",
  description: "Human-readable description",

  register(api: OpenClawPluginApi) {
    // Access plugin config (from plugins.entries.my-plugin.config)
    const config = api.pluginConfig as { apiKey: string };

    // Required tool — always available to LLM
    api.registerTool({
      name: "my_tool",                      // must not clash with core tools
      description: "Does something useful",
      parameters: Type.Object({
        input: Type.String({ description: "The input" }),
        count: Type.Optional(Type.Number()),
      }),
      async execute(_callId, params) {
        return {
          content: [{ type: "text", text: `Result: ${params.input}` }],
        };
      },
    });

    // Optional tool — user must add to tools.allow
    api.registerTool(
      {
        name: "advanced_tool",
        description: "Side-effect tool (opt-in)",
        parameters: Type.Object({ pipeline: Type.String() }),
        async execute(_callId, params) {
          return { content: [{ type: "text", text: params.pipeline }] };
        },
      },
      { optional: true },
    );

    // Command — runs without going through LLM
    api.registerCommand({
      name: "my-plugin-status",
      description: "Show plugin status",
      handler: async () => ({ text: "Plugin is running" }),
    });

    // Event hook
    api.registerHook(["message.received"], async (event) => {
      api.logger.info(`Received message: ${event.type}`);
    });
  },
});

Channel plugin (defineChannelPluginEntry)

import { defineChannelPluginEntry } from "openclaw/plugin-sdk/core";
import { channelPlugin } from "./src/channel.js";
import { setMyRuntime } from "./src/runtime.js";

export default defineChannelPluginEntry({
  id: "my-channel",
  name: "My Channel",
  description: "Connects OpenClaw to MyPlatform",
  plugin: channelPlugin,
  setRuntime: setMyRuntime,

  // registerFull: only for tools/routes that should NOT load in setup-only mode
  registerFull(api) {
    api.registerTool({
      name: "my_channel_status",
      description: "Check channel connection",
      parameters: { type: "object", properties: {} },
      async execute() {
        return { content: [{ type: "text", text: "connected" }] };
      },
    });
  },
});

Setup entry (channels only)

// setup-entry.ts — lightweight, no heavy imports
import { defineSetupPluginEntry } from "openclaw/plugin-sdk/core";
import { channelPlugin } from "./src/channel.js";

export default defineSetupPluginEntry(channelPlugin);

When OpenClaw uses setupEntry: channel is disabled/unconfigured, or deferred loading is on. Must include: channel registration, HTTP routes needed before listen, gateway methods for startup. Must NOT include: CLI registrations, background services, heavy runtime imports.

One plugin, many capabilities

export default definePluginEntry({
  id: "my-hybrid",
  name: "My Hybrid Plugin",
  register(api) {
    api.registerProvider({ id: "my-provider", /* ... */ });
    api.registerSpeechProvider({ id: "my-speech", /* ... */ });
    api.registerTool({ name: "my_tool", /* ... */ });
    api.registerCommand({ name: "my-cmd", /* ... */ });
  },
});

Step 5 — Runtime helpers (api.runtime)

// Read/write config
const cfg = api.runtime.config.loadConfig();
await api.runtime.config.writeConfigFile({ ...cfg, talk: { enabled: true } });

// TTS (telephony)
const result = await api.runtime.tts.textToSpeechTelephony({
  text: "Hello from OpenClaw",
  cfg: api.config,
});
// Returns PCM audio buffer + sample rate. Edge TTS NOT supported for telephony.

// List voices
const voices = await api.runtime.tts.listVoices({ provider: "openai", cfg });

// Web search
const results = await api.runtime.webSearch.search({ query: "...", maxResults: 5 });

// Spawn subagent
const session = await api.runtime.subagent.spawn({ /* ... */ });

Full runtime namespace reference → see references/runtime.md

createPluginRuntimeStore

For shared mutable state across module imports (the correct pattern):

import { createPluginRuntimeStore } from "openclaw/plugin-sdk/runtime-store";

const runtimeStore = createPluginRuntimeStore<{
  client: MyPlatformClient;
  logger: Logger;
}>("My Channel runtime not initialized");

export const setMyRuntime = (rt: { client: MyPlatformClient; logger: Logger }) =>
  runtimeStore.setRuntime(rt);
export const getMyRuntime = () => runtimeStore.getRuntime();
// Also: tryGetRuntime(), clearRuntime()

Step 6 — openclaw.yml config

plugins:
  enabled: true
  allow:
    - my-plugin
  deny: []
  load:
    paths:
      - ~/Projects/my-plugin   # for dev/local plugins
  entries:
    my-plugin:
      enabled: true
      config:
        apiKey: "sk-..."
        timeout: 60

# Enable optional tools
tools:
  allow:
    - advanced_tool      # specific tool
    - my-plugin          # all tools from plugin

# Plugin slots (exclusive categories)
plugins:
  slots:
    memory: memory-lancedb       # or "memory-core" or "none"
    contextEngine: my-engine     # or "legacy"

# Channel config (for channel plugins)
channels:
  my-channel:
    enabled: true
    token: "bot-token"
    allowFrom:
      - "user-id-1"

Step 7 — Discovery precedence (first match wins)

  1. plugins.load.paths — explicit paths in config
  2. <workspace>/.openclaw/<plugin-root>/ — workspace-local (disabled by default, must explicitly enable)
  3. ~/.openclaw/<plugin-root>/ — global user extensions
  4. Bundled plugins — shipped with OpenClaw (many enabled by default)

Conflict rule: If multiple plugins try to own the same channel/tool id, first match wins, others are skipped with a warning.


Step 8 — CLI commands

# Discovery
openclaw plugins list
openclaw plugins list --enabled --verbose
openclaw plugins list --json
openclaw plugins inspect <id>
openclaw plugins inspect <id> --json
openclaw plugins doctor

# Install
openclaw plugins install @myorg/my-plugin          # ClawHub first, npm fallback
openclaw plugins install clawhub:@myorg/my-plugin  # ClawHub only
openclaw plugins install ./my-plugin               # local path
openclaw plugins install -l ./my-plugin            # link (dev, no copy)
openclaw plugins install <spec> --force            # overwrite existing
openclaw plugins install <spec> --pin              # record exact npm spec
openclaw plugins install <spec> --marketplace https://github.com/<owner>/<repo>

# Manage
openclaw plugins enable <id>
openclaw plugins disable <id>
openclaw plugins update <id>
openclaw plugins update --all
openclaw plugins uninstall <id>
openclaw plugins uninstall <id> --keep-files
openclaw plugins registry --refresh               # refresh metadata after changes

# Fix
openclaw doctor --fix                             # auto-remove stale entries
openclaw gateway restart                          # REQUIRED after any plugin change

Step 9 — Testing

# Run plugin tests (vitest)
pnpm test -- extensions/my-plugin/

# With coverage
pnpm test:coverage

# Scoped
pnpm test -- extensions/my-plugin/src/channel.test.ts -t "resolves account"

# Low-memory mode
OPENCLAW_TEST_PROFILE=low OPENCLAW_TEST_SERIAL_GATEWAY=1 pnpm test

Full testing patterns → see references/testing.md


Security rules (MUST follow)

  • Dependencies installed with npm install --ignore-scripts — no lifecycle scripts
  • Keep deps "pure JS/TS" — avoid packages with postinstall builds
  • Plugin entries must resolve inside the plugin directory (symlink escapes are rejected)
  • Non-bundled plugins without plugins.allow entry emit a startup warning
  • plugins.deny always wins over plugins.allow
  • Use optional: true for tools with side effects or extra binary requirements
  • Treat all plugin code as trusted — review before installing from unknown sources

Pre-submission checklist

  • [ ] package.json has "type": "module" and correct openclaw.extensions
  • [ ] openclaw.plugin.json present with valid id and configSchema (even if empty)
  • [ ] Entry uses definePluginEntry (tools/providers) or defineChannelPluginEntry (channels)
  • [ ] All imports use openclaw/plugin-sdk/<subpath> — never the root barrel
  • [ ] Internal imports go through ./api.ts or ./runtime-api.ts, not SDK self-imports
  • [ ] No scripts in dependency packages (security)
  • [ ] Optional tools marked { optional: true }
  • [ ] uiHints.sensitive: true on any API key / secret config fields
  • [ ] Gateway restarted: openclaw gateway restart
  • [ ] Tests pass: pnpm test -- extensions/my-plugin/

Publishing

# Publish to npm (ClawHub auto-indexes npm)
npm publish --access public

# Users install with:
openclaw plugins install @myorg/my-plugin
# OpenClaw tries ClawHub first, npm fallback automatically

Reference files (read when needed)

  • references/manifest.md — complete openclaw.plugin.json field reference
  • references/sdk-subpaths.md — all 100+ SDK import subpaths grouped by purpose
  • references/channel-plugin.md — full channel plugin walkthrough (ChannelPlugin shape, pairing, actions, setup wizard)
  • references/provider-plugin.md — full provider plugin (all 22 hooks, auth, catalog, dynamic models, multi-capability)
  • references/runtime.md — all api.runtime namespaces with examples
  • references/testing.md — unit testing patterns, mocking, contract tests

平台分布

OpenClaw

90.61%
按下载量换算377

安全审计

VirusTotal

通过

ClawScan

可疑

Static analysis

通过

权限和风险

接触敏感数据

该 Skill 可能接触密钥、Token、环境变量或敏感配置,应进入高风险复核队列,默认不自动发布。

安装前确认

本站仅展示第三方公开信息,不托管安装包,不提供自动安装或运行环境。安装前应自行审查源码、依赖和命令行为。来源安全扫描存在提醒或未通过项,安装前请自行复核。目前只有一个公开来源,建议同时核对源仓库。

继续浏览同类 Skills