| name | url-insight |
| description | Auto-analyze URLs shared in conversation. Triggers when user sends a message containing one or more HTTP/HTTPS URLs without other instructions. Fetches content, generates structured summary, and provides personalized next-step recommendations based on user context (USER.md background). Use for: link analysis, article summarization, tool/project evaluation, tech discovery review. |
| metadata | {"openclaw":{"emoji":"🔗"}} |
URL Insight
When the user sends a URL (with no other instruction), automatically analyze it.
Workflow
- Detect: Message contains URL(s) and no explicit question/command
- Fetch: Use
web_fetch to extract page content (markdown mode, maxChars=8000)
- Read context: Load
USER.md for user background (skip if unavailable)
- Analyze & respond with this structure:
Output Format
📌 **{页面标题}**
> {一句话摘要,30字以内}
**📝 内容概要**
{3-5个要点,bullet list}
**🔍 关键信息**
- 类型:{文章/工具/项目/产品/论文/新闻}
- 技术栈:{如适用}
- 作者/来源:{如有}
**💡 对老板的建议**
{结合 USER.md 中的背景信息,给出 1-3 条具体的下一步行动建议}
🔗 {原始URL}
Rules
- If URL is unreachable, say so and suggest alternatives (archive.org, Google cache)
- For GitHub repos: focus on stars, language, last update, README highlights
- For news/articles: focus on core argument and implications
- For tools/products: focus on features, pricing, alternatives
- Multiple URLs: analyze each one separately
- Language: match user's language (default Chinese)
- Keep total response under 300 words per URL
- The "对老板的建议" section is the key differentiator — make it specific and actionable based on user's projects and interests