用途
OpenClaw Extend 是一个用于扩展 OpenClaw 系统能力的技能,定位为「OpenClaw 扩展操作者」。它的作用是向系统中添加新的组件,以扩展系统能力。该技能明确覆盖「添加新事物」这一范围,而非配置调优、诊断或运维维护。
主要能力
该技能覆盖以下可扩展对象:
- 插件(Plugins):扩展网关功能。
- 钩子(Hooks):智能体生命周期钩子。
- 智能体(Agents):隔离的智能体实例。
- 频道(Channels):消息集成,包括 WhatsApp、Telegram、Discord、Matrix。
- 节点(Nodes):远程执行节点。
- 模型(Models):AI 模型提供方。
- 设备(Devices):已配对设备管理。
- Webhook:外部事件集成。
- 技能(Skills):查看和管理可用技能。
使用方式
插件
列出与查看:
openclaw plugins list # all plugins
openclaw plugins info <name> # plugin details
openclaw plugins doctor # diagnose load issues
安装与管理:
openclaw plugins install <spec> # install (path, archive, or npm spec)
openclaw plugins enable <name> # enable a plugin
openclaw plugins disable <name> # disable a plugin
openclaw plugins update # update npm-installed plugins
安装插件后:
openclaw plugins list # verify it appears
openclaw daemon restart # restart to load
openclaw plugins doctor # check for load issues
插件编写指南见 references/plugin-hook-authoring.md。
钩子(Hooks)
列出与查看:
openclaw hooks list # all hooks
openclaw hooks info <name> # hook details
openclaw hooks check # eligibility status
安装与管理:
openclaw hooks install <spec> # install (path, archive, or npm spec)
openclaw hooks enable <name> # enable a hook
openclaw hooks disable <name> # disable a hook
openclaw hooks update # update npm-installed hooks
安装钩子后:
openclaw hooks list # verify it appears
openclaw hooks check # verify eligibility
openclaw daemon restart # restart to activate
钩子编写指南见 references/plugin-hook-authoring.md。
智能体(Agents)
列出与管理:
openclaw agents list # all configured agents
openclaw agents add # add a new agent (interactive)
openclaw agents set-identity <name> # update identity (name/theme/emoji/avatar)
openclaw agents delete <name> # delete agent + prune workspace/state
每个智能体拥有:隔离的工作区、隔离的认证/凭据、路由规则(哪些频道/目标映射到该智能体)。
频道(Channels)
列出与状态:
openclaw channels list # configured channels + auth profiles
openclaw channels status # gateway channel status
openclaw channels status --probe # deep probe all channels
openclaw channels capabilities # per-provider features
添加与配置:
openclaw channels add # add or update a channel account
openclaw channels login # link a channel (if supported)
openclaw channels logout # disconnect a session
openclaw channels remove # disable or delete
各频道类型的设置要求不同,详见 references/channel-setup.md,涵盖:WhatsApp(通过二维码进行 Web 链接)、Telegram(Bot token)、Discord(Bot token + guild 设置)、Matrix(homeserver + access token)。
添加频道后:
openclaw channels status --probe # verify connection
openclaw channels capabilities # check features
openclaw daemon restart # if needed
节点(Nodes)
列出与状态:
openclaw nodes list # all nodes (paired + pending)
openclaw nodes status # connection status + capabilities
openclaw nodes pending # pending pairing requests
openclaw nodes describe <name> # detailed capabilities
配对:
openclaw nodes approve <id> # approve a pending request
openclaw nodes reject <id> # reject a pending request
openclaw nodes rename <id> <name> # set display name
节点操作:
openclaw nodes invoke <node> <cmd> # invoke command on node
openclaw nodes run <node> <command> # run shell command (mac only)
openclaw nodes camera <node> # capture camera
openclaw nodes screen <node> # capture screen
openclaw nodes canvas <node> # capture/render canvas
openclaw nodes location <node> # get location
openclaw nodes notify <node> <msg> # send notification (mac only)
运行节点主机:
openclaw node run # run headless node host (foreground)
openclaw node install # install as system service
openclaw node status # check node host status
openclaw node restart # restart node host
openclaw node stop # stop node host
openclaw node uninstall # remove service
节点架构与安全见 references/node-pairing.md。
模型(Models)
状态:
openclaw models status --probe # probe all providers
配置:模型配置通过 openclaw config set 完成(变更规范使用 openclaw-admin):
openclaw config get models # current model config
openclaw config set agents.defaults.model "anthropic/claude-sonnet-4-5"
openclaw config set agents.defaults.subagents.model "anthropic/claude-haiku-3-5"
最佳实践:为子智能体使用更便宜的模型以降低成本。
设备(Devices)
列出与管理:
openclaw devices list # paired devices
openclaw devices approve <id> # approve pairing request
openclaw devices reject <id> # reject pairing request
openclaw devices revoke <id> # revoke device token
openclaw devices rotate <id> # rotate device token
Webhook
Gmail 集成:
openclaw webhooks gmail # Gmail Pub/Sub hooks (via gogcli)
技能(Skills)
查看:
openclaw skills list # all skills
openclaw skills check # readiness status
openclaw skills info <name> # skill details
扩展工作流
添加新能力时:
- 安装/添加 — 使用相应命令
- 验证 — 检查其是否出现在列表中
- 重启 — 若网关需要加载,执行 openclaw daemon restart
- 测试 — 探测或调用以确认可用
- 诊断 — 运行 /openclaw-doctor 验证系统健康
流程概括为:安装 → 验证 → 重启 → 测试 → 诊断。
适用场景
适用于需要向 OpenClaw 系统添加新集成或新能力的场景,例如安装插件、编写并启用钩子、创建隔离智能体、接入消息频道、配对远程节点、配置模型提供方、管理配对设备、设置 Webhook 等。
限制与边界
- 该技能仅覆盖「添加新事物」的范围。配置调优请使用 openclaw-admin;诊断请使用 openclaw-doctor;运维维护请使用 openclaw-maintain。
- 部分节点操作有平台限制:nodes run 与 nodes notify 标注为 mac only。
- 各频道类型的设置要求不同,需参考 references/channel-setup.md。
- 关于收费、价格、平台支持范围、效果等具体信息,资料中暂未提供。
参考文档
- references/plugin-hook-authoring.md — 如何编写插件和钩子
- references/channel-setup.md — 各频道设置指南
- references/node-pairing.md — 节点架构、配对流程、安全
相关技能
- openclaw-doctor — 添加新能力后验证系统健康
- openclaw-admin — 以「验证-应用-验证」规范应用配置变更
- openclaw-maintain — 守护进程运维、更新、cron、日志轮转