360 Secure SkillHub
360 安全技能中心

OpenClaw Docs

该 Skill 通过 shell 脚本缓存 OpenClaw 文档站索引,支持按关键词或正则检索页面标题与网址、抓取原始 Markdown、构建本地全文缓存并记录索引快照变化,用于回答 OpenClaw 安装、配置、CLI、频道、自动化等问题并给出有依据的文档出处。

jszesze
jszesze
热度 32.37
最近更新 2026-05-11 17:47:10
分类:搜索依赖:代码执行 / 第三方 API / 读写文件应用场景:研究

安装技能

复制下面的话术发给你的 Agent,它会通过 secure-skills 检索并安装该技能。

执行命令安装此 skill: npx secure-skills-cli add jszesze-openclaw-docs
Metadata
name
jszesze-openclaw-docs
description
OpenClaw documentation helper that caches the live docs index, searches page titles and URLs, fetches raw markdown docs, builds a local docs cache for full-text grep, and tracks index snapshots over time. Use when answering OpenClaw setup, configuration, CLI, channel, automation, architecture, or troubleshooting questions and you want grounded docs links or current page text.

OpenClaw docs

Use this skill for OpenClaw product docs, not for general repo spelunking.

What this skill is good at

  • finding the right OpenClaw doc page quickly
  • fetching raw markdown from docs.openclaw.ai
  • caching the live docs index from https://docs.openclaw.ai/llms.txt
  • downloading docs locally for full-text search
  • tracking page additions/removals across snapshots of the docs index

What it does not do

  • it does not inject the whole docs site into context automatically
  • it does not guarantee config snippets are current unless you verify them against fetched docs
  • it does not replace local repo inspection when the answer depends on the user’s actual workspace state

Workflow

  1. Start with the live docs index.
  2. Search by keyword to find the most likely page.
  3. Fetch the raw markdown page.
  4. Quote or summarize the relevant section.
  5. Link the exact docs URL in the answer.
  6. For broad questions, build a local docs cache and run a full-text search.

Commands

Run scripts with bash.

Index and cache

bash ./scripts/cache.sh status
bash ./scripts/cache.sh refresh

Search and discovery

bash ./scripts/search.sh telegram
bash ./scripts/search.sh "group.*policy"
bash ./scripts/sitemap.sh
bash ./scripts/fetch-doc.sh gateway/configuration
bash ./scripts/fetch-doc.sh https://docs.openclaw.ai/channels/telegram

Local full-text cache

bash ./scripts/build-index.sh fetch
bash ./scripts/build-index.sh build
bash ./scripts/build-index.sh search "requireMention"
# optional smoke-test limit:
OPENCLAW_DOCS_FETCH_LIMIT=25 bash ./scripts/build-index.sh fetch

Change tracking

bash ./scripts/track-changes.sh snapshot
bash ./scripts/track-changes.sh list
bash ./scripts/track-changes.sh since 2026-05-01
bash ./scripts/recent.sh 7

Heuristics

  • setup or install: check start/, install/, channels/, gateway/
  • config fields or JSON5 shape: check gateway/configuration, gateway/configuration-reference, gateway/configuration-examples
  • CLI questions: check cli/
  • channel behavior: check channels/<provider> and channels/troubleshooting
  • architecture or concepts: check concepts/
  • automation: check automation/
  • control UI, dashboard, web surfaces: check web/
  • node/device questions: check nodes/

Good defaults

  • Prefer fetched raw markdown over guessing from memory.
  • Prefer the smallest relevant page instead of dumping a broad overview page.
  • If a snippet matters operationally, verify it against the fetched page before giving it to the user.
  • For ambiguous terms, search first, then fetch.

Verified snippet notes

See ./snippets/common-configs.md for a few small, verified patterns copied from current docs pages. Treat them as helpers, not as a schema reference.