跳到正文
TakoAPI
返回技能列表

call-web-search-agent-strategy

由 alvinecarn 提供

Web & Frontend Development
464下载量由 clawskills.sh 提供
0浏览量
0 个赞登录后即可为这个技能点赞
—暂无评分
你的评分

简介

AI agent for call web search agent strategy tasks.

描述

该技能的功能

An AI research agent that gathers web-sourced information through a structured three-phase workflow: breadth search, focused iteration, and report synthesis. It enforces inline citations for every data point and caps research logs at 5,000 words. A deadlock mechanism stops repeated failed attempts after two tries and moves on to the next sub-goal.

The deadlock limit and hard word-count cap prevent the agent from looping on failed searches or producing bloated, uncited output that generic research agents commonly produce.

何时使用

  • Researching a newly announced product or version not in training data
  • Compiling sourced competitive intelligence on a niche topic
  • Finding recent financial data or social media trend reports
  • Building a cited research log to support a written report
  • Verifying unfamiliar technical terms or product names online

示例工作流

输入

User asks: Research the latest developments in Claude 4 Sonnet

智能体
  1. 1

    Creates a wiki research log document and records the file path

  2. 2

    Executes 3-5 parallel searches across core angles such as 'Claude 4 Sonnet review' and 'Claude 4 Sonnet release timeline'

  3. 3

    Selects up to 4 authoritative URLs and scrapes them, recording each finding with an inline citation

  4. 4

    Iterates focused one-question-at-a-time searches, appending new inline-cited findings until the word limit approaches

  5. 5

    Generates a deduplicated numbered bibliography and submits the completed log via submit_result

输出

A wiki document with inline-cited research findings and a numbered bibliography of all sources used

安装

clawhub install alvinecarn/call-web-search-agent-strategy

运行 CLI 命令,或把这段提示词粘贴到 Claude Code、Cursor、Codex 或你在用的任意编程智能体里。

链接