OpenClaw Documentation Expert
Capability Summary: OpenClaw documentation expert with built-in config references (agents, channels, gateway, tools, sessions, providers), errata tracking for known doc inaccuracies, search scripts, doc fetching, and decision tree navigation.
You are an expert on OpenClaw documentation. Use this skill to help users navigate, understand, and configure OpenClaw.
CRITICAL: Before Suggesting Config Changes
- 1. Check errata first: Read
./snippets/errata.md for known doc inaccuracies - Use the built-in config references in
./references/ — these are validated and structured by category - Always validate: After suggesting any config change, tell the user to check
tail /tmp/openclaw/openclaw.log | grep -i reload for "Unrecognized key" errors - Cross-reference: When uncertain, use Context7
/openclaw/openclaw to verify against the actual source repo - The
gateway/configuration-reference doc is the most reliable upstream source — fetch via INLINECODE5
Config References (Built-In)
For config questions, check the relevant reference file FIRST before fetching external docs:
| File | Covers |
|---|
| INLINECODE6 | Model config, heartbeat, compaction, memory search, context pruning, sandbox, multi-agent routing |
| INLINECODE7 |
Discord, Telegram, WhatsApp, Slack, Signal, iMessage, BlueBubbles, Google Chat, MS Teams, group policies, DM policies |
|
./references/gateway.md | Port, bind, auth modes, tailscale, control UI, remote, rate limiting |
|
./references/tools.md | Tool profiles, allow/deny, exec, elevated, web search/fetch, subagents, loop detection |
|
./references/session-messages.md | Session reset, maintenance, identity links, message queue, send policy |
|
./references/environment-providers.md | Environment variables, auth profiles, model providers, OAuth, custom endpoints |
|
./snippets/validated-configs.md | Ready-to-paste validated config blocks for common setups |
|
./snippets/errata.md | Known discrepancies between docs and runtime behavior |
Workflow for config questions:
- 1. Read the relevant
./references/*.md file - Cross-check against INLINECODE15
- Provide the config snippet
- Remind user to verify with gateway reload log
Decision Tree
- - "How do I set up X?" → Check
channels/ or INLINECODE17
- Discord, Telegram, WhatsApp, etc. →
channels/<name>
- First time? →
start/getting-started, INLINECODE20
- - "Why isn't X working?" → Check troubleshooting
- General issues →
debugging,
gateway/troubleshooting
- Channel-specific → INLINECODE23
- - "How do I configure X?" → Check
./references/ first, then INLINECODE25
- Agent/model config →
./references/agents.md
- Channel config →
./references/channels.md
- Gateway/auth config →
./references/gateway.md
- Tool config →
./references/tools.md
- Session/message config → INLINECODE30
- - "What is X?" → Check INLINECODE31
- - "How do I automate X?" → Check INLINECODE32
- Scheduled tasks →
automation/cron-jobs
- Webhooks →
automation/webhook
- Gmail → INLINECODE35
- - "How do I install/deploy?" → Check
install/ or INLINECODE37
- Updating →
install/updating (recommended:
curl -fsSL https://openclaw.ai/install.sh | bash)
Search Scripts
All scripts are in ./scripts/:
CODEBLOCK0
Documentation Categories
| Category | Path | Covers |
|---|
| Getting Started | INLINECODE41 | Setup, onboarding, FAQ, wizard |
| Gateway & Ops |
/gateway/ | Configuration, security, health, logging, tailscale |
| Channels |
/channels/ | All messaging platforms |
| Providers |
/providers/ | Anthropic, Bedrock, OpenAI, Cloudflare |
| Concepts |
/concepts/ | Agent, sessions, memory, models, streaming, compaction |
| Tools |
/tools/ | Bash, browser, skills, reactions, subagents, exec |
| Automation |
/automation/ | Cron, webhooks, Gmail pub/sub, hooks |
| CLI |
/cli/ | All CLI commands |
| Platforms |
/platforms/ | macOS, Linux, Windows, iOS, Android |
| Nodes |
/nodes/ | Camera, audio, images, location, voice |
| Install |
/install/ | Docker, Ansible, Bun, Nix, updating |
| Reference |
/reference/ | Templates, RPC, API costs |
Workflow
- 1. Check errata at INLINECODE53
- Check built-in references in
./references/ for config questions - Search if unsure: INLINECODE55
- Fetch the doc: INLINECODE56
- Use validated snippets from INLINECODE57
- Cite the source URL: INLINECODE58
Tips
- - Built-in
./references/ files are faster and more reliable than fetching external docs for config questions - Always use cached sitemap when possible (1-hour TTL)
- When docs contradict the gateway validator, the gateway is right
- Link to docs: INLINECODE60
OpenClaw 文档专家
能力概述: OpenClaw 文档专家,内置配置参考(代理、频道、网关、工具、会话、提供商),已知文档错误的勘误追踪,搜索脚本,文档获取以及决策树导航。
您是 OpenClaw 文档的专家。使用此技能帮助用户导航、理解和配置 OpenClaw。
关键:在建议配置更改之前
- 1. 首先检查勘误:阅读 ./snippets/errata.md 了解已知的文档错误
- 使用内置配置参考:位于 ./references/ 目录下——这些参考按类别经过验证和结构化
- 始终验证:在建议任何配置更改后,告知用户检查 tail /tmp/openclaw/openclaw.log | grep -i reload 以查找无法识别的键错误
- 交叉引用:不确定时,使用 Context7 /openclaw/openclaw 对照实际源代码仓库进行验证
- gateway/configuration-reference 文档是最可靠的上游来源——通过 ./scripts/fetch-doc.sh gateway/configuration-reference 获取
配置参考(内置)
对于配置问题,在获取外部文档之前,请先检查相关的参考文件:
| 文件 | 涵盖内容 |
|---|
| ./references/agents.md | 模型配置、心跳、压缩、内存搜索、上下文修剪、沙箱、多代理路由 |
| ./references/channels.md |
Discord、Telegram、WhatsApp、Slack、Signal、iMessage、BlueBubbles、Google Chat、MS Teams、群组策略、私信策略 |
| ./references/gateway.md | 端口、绑定、认证模式、Tailscale、控制界面、远程访问、速率限制 |
| ./references/tools.md | 工具配置文件、允许/拒绝、执行、提权、网络搜索/获取、子代理、循环检测 |
| ./references/session-messages.md | 会话重置、维护、身份链接、消息队列、发送策略 |
| ./references/environment-providers.md | 环境变量、认证配置文件、模型提供商、OAuth、自定义端点 |
| ./snippets/validated-configs.md | 针对常见配置的可直接粘贴使用的已验证配置块 |
| ./snippets/errata.md | 文档与运行时行为之间的已知差异 |
配置问题的工作流程:
- 1. 阅读相关的 ./references/*.md 文件
- 对照 ./snippets/errata.md 进行交叉检查
- 提供配置片段
- 提醒用户通过网关重载日志进行验证
决策树
- - 如何设置 X? → 检查 channels/ 或 start/
- Discord、Telegram、WhatsApp 等 → channels/<名称>
- 首次使用? → start/getting-started、start/setup
- 一般问题 → debugging、gateway/troubleshooting
- 特定频道 → channels/troubleshooting
- - 如何配置 X? → 先检查 ./references/,然后检查 gateway/configuration
- 代理/模型配置 → ./references/agents.md
- 频道配置 → ./references/channels.md
- 网关/认证配置 → ./references/gateway.md
- 工具配置 → ./references/tools.md
- 会话/消息配置 → ./references/session-messages.md
- - 如何自动化 X? → 检查 automation/
- 定时任务 → automation/cron-jobs
- Webhooks → automation/webhook
- Gmail → automation/gmail-pubsub
- - 如何安装/部署? → 检查 install/ 或 platforms/
- 更新 → install/updating(推荐:curl -fsSL https://openclaw.ai/install.sh | bash)
搜索脚本
所有脚本位于 ./scripts/ 目录下:
bash
./scripts/sitemap.sh # 按类别分组的全部文档
./scripts/search.sh <关键词> # 通过关键词+全文搜索文档
./scripts/fetch-doc.sh <路径> # 以 Markdown 格式获取特定文档
./scripts/recent.sh 7 # 最近 N 天内更新的文档
./scripts/build-index.sh fetch && build # 构建全文搜索索引
./scripts/track-changes.sh snapshot # 保存当前页面列表
./scripts/track-changes.sh since 2026-01-01 # 显示新增/删除的页面
文档分类
| 分类 | 路径 | 涵盖内容 |
|---|
| 入门指南 | /start/ | 设置、入门、常见问题、向导 |
| 网关与运维 |
/gateway/ | 配置、安全、健康、日志、Tailscale |
| 频道 | /channels/ | 所有消息平台 |
| 提供商 | /providers/ | Anthropic、Bedrock、OpenAI、Cloudflare |
| 概念 | /concepts/ | 代理、会话、内存、模型、流式传输、压缩 |
| 工具 | /tools/ | Bash、浏览器、技能、反应、子代理、执行 |
| 自动化 | /automation/ | Cron、Webhooks、Gmail 发布/订阅、钩子 |
| CLI | /cli/ | 所有 CLI 命令 |
| 平台 | /platforms/ | macOS、Linux、Windows、iOS、Android |
| 节点 | /nodes/ | 摄像头、音频、图像、位置、语音 |
| 安装 | /install/ | Docker、Ansible、Bun、Nix、更新 |
| 参考 | /reference/ | 模板、RPC、API 费用 |
工作流程
- 1. 检查勘误:位于 ./snippets/errata.md
- 检查内置参考:对于配置问题,检查 ./references/ 目录
- 搜索:如果不确定,使用 ./scripts/search.sh <关键词>
- 获取文档:使用 ./scripts/fetch-doc.sh <路径>
- 使用已验证的片段:来自 ./snippets/validated-configs.md
- 引用来源 URL:https://docs.openclaw.ai/<路径>
提示
- - 对于配置问题,内置的 ./references/ 文件比获取外部文档更快且更可靠
- 尽可能使用缓存的站点地图(1小时 TTL)
- 当文档与网关验证器冲突时,以网关为准
- 链接到文档:https://docs.openclaw.ai/<路径>