← 返回 Skills 市场
windseeker1111

FlowSearch — Deep Web Research via Claude Native Search

作者 windseeker1111 · GitHub ↗ · v1.2.0 · MIT-0
cross-platform ⚠ suspicious
356
总下载
0
收藏
0
当前安装
3
版本数
在 OpenClaw 中安装
/install flow-search
功能描述
Deep web research using Claude's native search tool. Use for comprehensive research, market analysis, competitor intelligence, or when standard search isn't...
使用说明 (SKILL.md)

FlowSearch — Deep Web Research

Free deep research powered by Claude Max subscription.

Stop getting shallow answers. FlowSearch uses Claude's native search to go wide and deep — pulling from multiple sources, synthesizing conflicting information, and giving you the kind of answer a thorough analyst would write, not a one-liner.

How it works

FlowSearch spawns Claude CLI as a subprocess with your CLAUDE_CODE_OAUTH_TOKEN. Claude performs multi-source web search, synthesizes the results, and returns a structured report. No data is sent to third-party APIs — it's just Claude doing what Claude does best.

Security note: This skill spawns a claude process that inherits your shell environment to access CLAUDE_CODE_OAUTH_TOKEN. Review the source (search.ts) before running — it's 150 lines and straightforward.

Prerequisites

  1. Claude Code CLI — install from claude.ai/code
  2. Authenticate: run claude auth login (opens browser)
  3. Claude Max subscription recommended for best results (more search depth)

Installation

npx clawhub@latest install flow-search
cd ~/.openclaw/skills/flow-search
npm install

Usage

Quick search (15–30s)

npx tsx search.ts "Kling AI pricing 2026"

Deep research (60–180s)

npx tsx search.ts --deep "AI video generation competitive landscape"

In your own skills or agents

import { claudeSearch, claudeResearch } from "./search.ts";

// Quick answer with sources
const result = await claudeSearch("What is Kling AI pricing?");
if (result.success) console.log(result.answer);

// Deep research with guiding questions
const report = await claudeResearch("AI video market 2026", [
  "Who are the top 5 players and their pricing?",
  "What's the total addressable market?",
  "Which companies raised funding in the last 6 months?"
]);

Output format

Quick search returns a direct answer with cited sources.

Deep research (--deep) returns a full structured report:

## Summary
2–3 sentence overview of findings.

## Key Findings
- Specific fact with source
- Specific fact with source
- ...

## Details
Full analysis with context.

## Sources
- [Publication Name](https://url)

When to use FlowSearch vs Brave Search

Use FlowSearch Use Brave
Market analysis Quick fact check
Competitor deep-dives Single answer needed
Multi-source synthesis Fast lookup
Recent news + context Basic info
"What's the full picture on X?" "What year was X founded?"

Environment variables

Variable Required Description
CLAUDE_CODE_OAUTH_TOKEN Yes From claude auth login

Options

Flag Description
--deep / -d Deep research mode (slower, more thorough)
--help / -h Show usage

Integration example — use inside another OpenClaw skill

// In your skill's handler
import { claudeResearch } from "~/.openclaw/skills/flow-search/search.ts";

const intel = await claudeResearch(`Competitor analysis: ${companyName}`, [
  "What are their pricing tiers?",
  "Who are their target customers?",
  "What are their main weaknesses?"
]);

if (intel.success) {
  // Pass intel.answer to your LLM or store it
}

Timeouts

  • Quick search: 2 minutes max
  • Deep research: 3 minutes max

About

Built by the Flow team. Free, MIT-0 licensed — use it, fork it, build on it.

Other Flow skills: FlowCrawl · FlowConcierge · FlowForge

安全使用建议
This skill appears to do what it claims: run a local 'claude' CLI using your CLAUDE_CODE_OAUTH_TOKEN to perform searches. Before installing or running it, (1) verify the source and integrity of the 'claude' CLI you will run (ensure it's the official binary), (2) inspect search.ts yourself (the file is included) so you understand what it spawns and which env vars are passed, (3) avoid running it in environments that contain other sensitive environment variables you don't want the spawned process to see, and (4) consider using a token with limited scope or a dedicated machine/container if you are concerned about token exposure. Note: the script adds a '--dangerously-skip-permissions' argument when not running as root — this is unusual and may alter the CLI's permission checks; investigate that flag in the official Claude CLI docs before use.
功能分析
Type: OpenClaw Skill Name: flow-search Version: 1.2.0 The skill is a legitimate wrapper for the Claude CLI, designed to perform web searches and research synthesis. It correctly uses `child_process.spawn` with an argument array to avoid shell injection and requires the `CLAUDE_CODE_OAUTH_TOKEN` environment variable as expected for its stated purpose. While it uses the `--dangerously-skip-permissions` flag to facilitate automation, this behavior is documented in `SKILL.md` and `search.ts`, and there is no evidence of data exfiltration or malicious intent.
能力评估
Purpose & Capability
Name/description promise deep web research via Claude's native search and the code + SKILL.md only require the Claude CLI and a CLAUDE_CODE_OAUTH_TOKEN; these requirements align with the stated purpose.
Instruction Scope
SKILL.md and search.ts instruct the agent to spawn the local 'claude' CLI and inherit the shell environment so the token is available — this is expected for a CLI-based integration but means the external 'claude' binary will see your full environment. The SKILL.md explicitly tells users to review search.ts, which is good practice.
Install Mechanism
There is no remote install script or download in the bundle (instruction-only with a local script file). The package.json/search.ts are included but nothing fetches code from untrusted URLs. Lowest-risk install model.
Credentials
The only required secret is CLAUDE_CODE_OAUTH_TOKEN, which is justified for calling the Claude CLI. That token is sensitive — the skill passes the entire process.env through to the spawned process, so any other environment secrets present in the agent will be visible to the Claude CLI process.
Persistence & Privilege
The skill is not force-included (always:false) and requests no system-level persistence or modifications; autonomous invocation is allowed by default but not combined with other concerning privileges.
如何使用
  1. 确保已安装 OpenClaw(本地或 Docker 部署)
  2. 在对话框中输入安装命令:/install flow-search
  3. 安装完成后,直接呼叫该 Skill 的名称或使用 /flow-search 触发
  4. 根据 Skill 的参数说明提供必要输入,即可获得结构化输出
版本历史
v1.2.0
Deep web research using Claude's native search tool. Comprehensive research, market analysis, competitor intelligence. Free from the Flow team.
v1.1.0
Improved docs, explicit credential requirements, security transparency, integration examples, comparison table
v1.0.0
Initial public release — deep web research via Claude CLI
元数据
Slug flow-search
版本 1.2.0
许可证 MIT-0
累计安装 0
当前安装数 0
历史版本数 3
常见问题

FlowSearch — Deep Web Research via Claude Native Search 是什么?

Deep web research using Claude's native search tool. Use for comprehensive research, market analysis, competitor intelligence, or when standard search isn't... 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件,目前累计下载 356 次。

如何安装 FlowSearch — Deep Web Research via Claude Native Search?

在 OpenClaw 或 Claude Code 对话框中运行命令「/install flow-search」即可一键安装,无需额外配置。

FlowSearch — Deep Web Research via Claude Native Search 是免费的吗?

是的,FlowSearch — Deep Web Research via Claude Native Search 完全免费,采用 MIT-0 许可证,可自由下载、安装和使用。

FlowSearch — Deep Web Research via Claude Native Search 支持哪些平台?

FlowSearch — Deep Web Research via Claude Native Search 跨平台运行,可在任意部署了 OpenClaw / Claude Code 的环境中使用(cross-platform)。

谁开发了 FlowSearch — Deep Web Research via Claude Native Search?

由 windseeker1111(@windseeker1111)开发并维护,当前版本 v1.2.0。

💬 留言讨论