openclaw-help（OpenClaw help 搜索）

openclaw-help Skill

Language: Respond in the same language as the user (中文提问则中文回答).

Answer OpenClaw usage questions using the bundled reference (橙皮书 extracts) and official docs (docs.openclaw.ai). When the two conflict, prefer official docs. When neither suffices, trigger WebSearch.

Reference is derived from OpenClaw橙皮书 v1.2.0; for latest details, prefer docs.openclaw.ai.

Knowledge Sources

1. Embedded Reference (bundled)

• Path: reference.md (sibling to SKILL.md in the openclaw-help skill directory)
• Content: FAQ, Command Cheat Sheet, Resources, Architecture, Deployment, Channels, Skills, Model providers, Security
• Retrieval: Grep reference.md → Read matching regions

2. Official Docs (Remote)

• Base: https://docs.openclaw.ai
• Getting Started: https://docs.openclaw.ai/start/getting-started
• Doc Index: https://docs.openclaw.ai/llms.txt (discover channels, cli, install pages)
• Retrieval: WebFetch to fetch relevant pages

3. Alignment Rule

• On conflict (e.g., Node version, command syntax): prefer official docs
• Install / quick start / commands: MUST reference official docs in the answer

Execution Flow

Step 0 · Trigger Check

Activate when user asks about OpenClaw usage: deployment, channel integration, configuration, errors, Skills, 养虾, etc.

Step 1 · Retrieve from Reference

1. Extract 2–5 keywords from the user question (e.g., 部署, Telegram, Skills, openclaw doctor, FAQ)
2. Grep reference.md in the skill directory for those keywords
3. Read matching regions with context (e.g., ±50 lines)
4. If no hits or content insufficient → treat as "reference no cover"

Query → Section mapping (优先 Read 对应段落以快速获取完整回答):

Step 2 · Retrieve from Official Docs

Use WebFetch (or mcp_web_fetch) to fetch relevant docs pages:

If WebFetch fails → continue with reference or fallback; note network issue.

Step 3 · Synthesize Answer

Step 4 · WebSearch Fallback

When neither reference nor official docs has sufficient content:

1. Invoke WebSearch with the user question (and OpenClaw-related refinements)
2. Organize answer from search results
3. Note: "Reference 与官方文档未覆盖，以下来自网络搜索"

Reference Attribution

• From reference: Cite section (e.g., "据 reference 渠道章节…")
• From official docs: Include docs.openclaw.ai link (e.g., https://docs.openclaw.ai/start/getting-started)
• From web search: Note "Reference 与官方文档未覆盖，以下来自网络搜索"

Examples

Example A — Telegram 接入 (Chinese)

Input: "OpenClaw 如何接入 Telegram？"

Actions:

1. Grep reference.md for "Telegram"
2. WebFetch https://docs.openclaw.ai/channels/telegram.md
3. Synthesize from both; prefer official docs on setup steps
4. Cite reference Channels section and link to docs.openclaw.ai/channels/telegram

Example B — openclaw doctor (Chinese)

Input: "openclaw doctor 是干什么的？"

Actions:

1. Grep reference.md for "doctor" or "诊断"
2. WebFetch https://docs.openclaw.ai/cli/doctor.md
3. Synthesize; prefer official docs on command flags

Example C — Quick start (官方文档为主)

Input: "OpenClaw 快速开始怎么装？"

Actions:

1. WebFetch https://docs.openclaw.ai/start/getting-started (primary)
2. Optional: Grep reference.md for "安装"
3. Answer primarily from official docs; add reference background if relevant

Example D — ClawHub Skill 安全 (Reference FAQ)

Input: "ClawHub Skill 安全吗？"

Actions:

1. Grep reference.md for "ClawHub" or "安全"
2. WebFetch docs if relevant; reference FAQ Q9 likely sufficient
3. Cite reference FAQ section

Example E — Fallback

Input: Question not covered by reference or official docs

Actions:

1. WebSearch
2. Organize answer; note "Reference 与官方文档未覆盖，以下来自网络搜索"

Error Handling