Token导航 LogoToken导航

Agent Skill · 研究检索

chatclaw(chatclaw 搜索)

chatclaw 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。

chatclaw 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 12,829。

总安装 9,937周安装 402GitHub Stars 1下载 3,120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
0.4.0

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:chatclaw(chatclaw 搜索)
来源仓库:https://github.com/sumeralabs/chatclaw
安装命令:
openclaw skills install chatclaw
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 chatclaw

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。

将您的 OpenClaw 机器人连接到 ChatClaw 云仪表板,以进行实时远程聊天、令牌跟踪和任务管理

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 chatclaw;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

chatclaw 怎么安装?

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

chatclaw 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。

chatclaw 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。

chatclaw 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。

chatclaw 用的人多吗?

公开数据是累计安装 9,937 次,近一周安装 402 次,下载 3,120 次,GitHub 1 星。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。

chatclaw 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
chatclaw

chatclaw 搜索 · 本页

9,9371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
chatclaw
version
0.4.0
description
Connect your OpenClaw bot to the ChatClaw cloud dashboard — remote chat, token tracking, task management, agent workspace file browsing (read-only), and skills management (list, enable/disable, reinstall)
author
ChatClaw Team
homepage
https://chatclaw.sumeralabs.com
metadata
{ "openclaw": { "os": ["darwin", "linux"], "requires": { "bins": ["python3"], "env": ["CHATCLAW_API_KEY"], "config": ["gateway.http.endpoints.chatCompletions", "gateway.auth.token"] }, "primaryEnv": "CHATCLAW_API_KEY", "install": { "uv": "aiohttp>=3.9 websockets>=12.0 cryptography>=41.0" } } }

ChatClaw

ChatClaw is a persistent background bridge that connects your local OpenClaw agent to the ChatClaw cloud dashboard. Once installed and enabled, it runs automatically with OpenClaw and allows you to chat with your agent, monitor token usage, and manage tasks from any browser or mobile device — without exposing any ports or configuring a firewall.

What it does

Transparency note: The cloud relay endpoint (api.sumeralabs.com) is operated by SumeraLabs, the company behind ChatClaw. chatclaw.sumeralabs.com is the product domain; sumeralabs.com is the infrastructure domain. Both are owned and operated by the same team.

ChatClaw opens two connections when enabled:

  1. Cloud relay — an outbound WebSocket to wss://api.sumeralabs.com/ws/agent/{api_key}. This is how your dashboard communicates with the skill. It is a purely outbound connection and requires no inbound port forwarding.
  2. Local gateway — a WebSocket connection to ws://localhost:18789 for Ed25519 authentication only, plus HTTP SSE calls to http://localhost:18789/v1/chat/completions for all chat traffic. The HTTP endpoint is auto-enabled in openclaw.json on first start.

Messages flow bidirectionally in real time. User messages sent from the dashboard are forwarded to the OpenClaw agent via HTTP SSE streaming. Each token of the agent's response is relayed back to the dashboard as it is generated, producing a live typewriter effect.

Permissions & Capabilities

This section fully discloses every action the skill can perform in response to a cloud relay message. All operations require a valid ChatClaw API key; the relay rejects unauthenticated connections before any message reaches the skill.

CapabilityTrigger messageScopeNotes
Chat relayany message with text/content fieldSends text to OpenClaw agent via HTTP SSE, streams response backCore feature
File listingfiles.listRead-only; returns file metadata for the agent workspace via gateway RPC (agents.files.list)No file content returned at this step
File readfiles.getRead-only; returns content of a named workspace file via gateway RPC (agents.files.get); scoped to agentId: "main" workspaceEach request is logged with the filename
Skills listskills.listRuns openclaw skills list --json locally; returns skill metadataRead-only
Enable/disable skillskills.set_enabledRuns openclaw config set skills.entries.<name>.enabled true/false; affects local OpenClaw configEach invocation is logged with skill name and new state
Reinstall skillskills.reinstallRuns clawhub install <name> --force; replaces skill files from ClawHub registryEach invocation is logged with skill name

Why operator-level credentials are required: The gateway uses Ed25519 device authentication with operator scopes (operator.admin, operator.approvals, operator.pairing). This is required by the OpenClaw gateway handshake protocol to establish an authenticated connection — there is no lower-privilege connection mode for skill bridges. The operator token is read from device-auth.json and used only for the initial WebSocket handshake; it is never transmitted to the cloud relay.

Security mitigations:

  • All capabilities require a valid API key authenticated at the cloud relay layer before any message reaches the skill
  • File access is read-only and scoped to the agent workspace via the gateway RPC (agentId: "main")
  • Skills enable/disable and reinstall are logged locally on every invocation
  • No shell commands with user-controlled arguments are executed — skill names and file names are passed as positional arguments to controlled subprocess calls, not interpolated into shell strings
  • Disabling this skill reverts any openclaw.json changes it made

Architecture

ChatClaw Dashboard (browser / mobile)
        ↕  wss://api.sumeralabs.com
  ChatClaw Cloud Relay (FastAPI)
        ↕  wss://api.sumeralabs.com/ws/agent/{api_key}
  ChatClaw Skill  ←── this package
        ↕  ws://localhost:18789 (auth handshake, Ed25519)
        ↕  http://localhost:18789/v1/chat/completions (SSE streaming)
  OpenClaw Gateway
        ↕
  OpenClaw Agent (LLM)

Requirements

  • OpenClaw: v2026.3.28 or later
  • Python: 3.8+
  • Dependencies: websockets, cryptography, aiohttp
Important: ChatClaw v0.3.0+ requires OpenClaw v2026.3.28 or later due to authentication changes. If you're using an older OpenClaw version, please upgrade OpenClaw or use ChatClaw v0.2.9 (legacy).

Installation

Via OpenClaw Control UI (recommended)

  1. Open the Control UI at http://localhost:18789
  2. Go to Skills → Marketplace
  3. Search for ChatClaw
  4. Click Install, enter your API key, click Enable
Note: On first enable, the skill automatically sets gateway.http.endpoints.chatCompletions.enabled = true in your openclaw.json. If the skill made that change, disabling it will revert it automatically. If chatCompletions was already enabled before install, the skill leaves your config unchanged on disable.

The skill starts immediately and auto-starts with OpenClaw on every subsequent boot.

Via OpenClaw CLI

npx clawhub install chatclaw
openclaw skills enable chatclaw
openclaw skills logs chatclaw

Before enabling, add your API key to openclaw.json under skills.entries.chatclaw.config:

"skills": {
  "entries": {
    "chatclaw": {
      "enabled": true,
      "config": {
        "apiKey": "ck_your_key_here",
        "cloud_url": "wss://api.sumeralabs.com"
      }
    }
  }
}

Agent-assisted setup

Your OpenClaw agent can install and configure ChatClaw for you without any manual steps.

Get the one-click setup prompt from app.chatclaw.sumeralabs.com/setup — it includes your API key pre-filled and full instructions to paste directly into chat.

Configuration

KeyRequiredDefaultDescription
apiKeyYes—ChatClaw API key from app.chatclaw.sumeralabs.com
cloud_urlNowss://api.sumeralabs.comWebSocket relay URL (leave default unless self-hosting)

Environment variables

VariableDescription
OPENCLAW_DATA_DIROverride the OpenClaw data directory. Defaults to /data/.openclaw on Docker/VPS or ~/.openclaw on standard installs.
CHATCLAW_CLOUD_URLOverride the cloud relay URL. Defaults to wss://api.sumeralabs.com. Leave unset unless self-hosting the relay.

Lifecycle hooks

HookBehaviour
on_enablePatches openclaw.json to enable the /v1/chat/completions endpoint, then starts the cloud ↔ gateway relay loop
on_disableCloses both WebSocket connections, stops the relay loop, and reverts the chatCompletions patch in openclaw.json if the skill was the one that enabled it

Reconnection behaviour

Both the cloud relay and the local gateway implement automatic reconnection with exponential backoff (5 s → 10 s → 20 s … up to 60 s). The skill never exits on a connection drop.

Verify it is working

openclaw skills logs chatclaw

Expected output:

Connected to cloud relay ✓
Gateway authenticated ✓
Both connections established — relaying messages ✓

Troubleshooting

SymptomCauseFix
Cloud connection failedWrong API key or relay unreachableVerify key at app.chatclaw.sumeralabs.com; check network connectivity
OpenClaw identity files not foundOpenClaw not initialised or non-default install pathRun openclaw wizard or set OPENCLAW_DATA_DIR
Gateway HTTP 403chatCompletions endpoint not enabledRestart skill — on_enable auto-patches openclaw.json
Gateway auth rejectedOperator token expiredRe-pair device via openclaw devices approve
Streaming works but token count shows 0sessions.json not yet writtenSend one message first; counts appear after the first completion

File access

This skill reads and writes the following local files:

PathAccessPurpose
$OPENCLAW_DATA_DIR/openclaw.jsonRead + WriteReads gateway auth token; enables chatCompletions HTTP endpoint on first start
$OPENCLAW_DATA_DIR/identity/device.jsonReadEd25519 device identity for gateway authentication
$OPENCLAW_DATA_DIR/identity/device-auth.jsonReadOperator token for gateway authentication
$OPENCLAW_DATA_DIR/agents/main/sessions/sessions.jsonReadToken usage counts (input/output/context) per session

$OPENCLAW_DATA_DIR defaults to /data/.openclaw on Docker/VPS installs and ~/.openclaw on standard installs. It can be overridden via the OPENCLAW_DATA_DIR environment variable. No files outside this directory are accessed.

External connections

This skill makes the following outbound network connections:

  • wss://api.sumeralabs.com — ChatClaw cloud relay (authentication and message relay)
  • ws://localhost:18789 — OpenClaw gateway WebSocket (Ed25519 auth handshake only)
  • http://localhost:18789/v1/chat/completions — OpenClaw gateway HTTP (SSE streaming chat)

No inbound ports are opened. No user data is stored by the skill itself — messages are persisted by the ChatClaw backend (Supabase) for chat history.

Changelog

v0.3.0 (2026-03-30)

  • Breaking: Now requires OpenClaw v2026.3.28+
  • Fixed authentication for OpenClaw v2026.3.28 (added required scope headers)
  • Fixed model field validation (changed from "auto" to "openclaw")
  • Added cross-platform path resolution (Docker/VPS/macOS)
  • Improved error messages for troubleshooting
  • Added support for approval workflows (operator.approvals scope)

v0.2.9 (2026-03-12)

  • Initial ClawHub release

License

MIT-0 (No Attribution Required)

平台分布

OpenClaw

90.94%
按下载量换算2,837

安全审计

VirusTotal

可疑

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