551 everyinfra
everysearch Skill
使用 EverySearch 进行最新的网页、语义、学术、论坛、页面阅读、爬取和交叉验证检索。当任务需要最新来源、多种检索机制、带引用的事实核查,或需要从特定公开 URL 或网站获取内容时使用。
安装方式:把技能目录放入 ~/.claude/skills/(Claude Code)或在 claude.ai 设置中启用;也可复制右侧安装命令一键添加。
技能指令原文(SKILL.md)
EverySearch
Use the MCP tool everyinfra_search. Its live input schema is the source of truth for available
tool values and parameters.
- Use the narrowest retrieval mechanism that matches the user's question.
- Use
crosscheckfor a concrete claim such as a current version, price, policy, availability, or
API behavior. It is designed to compare mechanism-distinct retrieval paths.
- Use page-reading or crawl tools for known URLs instead of searching for the same page again.
- Keep query and result limits proportional to the task. Empty or failed results must not be
presented as proof that nothing exists.
- Cite the original sources returned by the search result when producing factual conclusions.
- Read the returned
billingstate. Failed or empty calls may be uncharged, but do not infer that
without the response.
Do not silently replace semantic retrieval with keyword retrieval when result meaning matters.
If the requested tool is not in the current MCP schema, state that it is unavailable.