company-research
Multi-source company research tool that generates structured due-diligence reports. Use when the user asks to research, look up, or investigate a company — including questions about shareholders, legal representative, registered capital, equity structure, beneficial owner, funding history, investors, valuation, lawsuits, court judgments, enforcement records, blacklist / dishonest debtor status, administrative penalties, operating anomalies, trademarks, patents, government procurement / bidding, recruitment profile, negative news, competitors, or industry position. Also triggers on: "帮我查一下XX公司", "XX公司背景", "XX的股东是谁", "XX有没有诉讼/被执行/失信", "XX融了多少钱", "XX股权结构", "尽调", "公司调研", "公司背景调查", "is X company reliable", "due diligence on X", "background check on X company".
Install via CLI (Recommended)
clawhub install openclaw/skills/skills/fan31415/company-searchCompany Research — Multi-LLM Adaptive Skill
Multi-source company research tool that generates structured reports with the same information granularity as Tianyancha / Qichacha. Supports Kimi / OpenAI GPT / Claude / Gemini / MiniMax / Cursor / generic Agent environments, with automatic tool detection and adaptation.
基于多源搜索的公司调研工具，生成"对标天眼查/企查查信息颗粒度"的结构化报告。 支持 Kimi / OpenAI GPT / Claude / Gemini / MiniMax / Cursor / 通用 Agent 环境，自动适配可用工具。
🧰 Tool Auto-Detection（执行前必须做）
在开始任何调研前，先检测当前环境可用工具，映射到两个抽象操作：
SEARCH 操作 — 按优先级选第一个可用的：
| 优先级 | 工具名 | 适用环境 |
|---|---|---|
| 1 | kimi_search | Kimi (Moonshot) |
| 2 | web_search_preview | OpenAI Responses API |
| 3 | web_search | Claude.ai / MiniMax / 通用 |
| 4 | brave_web_search | Claude+MCP / Cursor MCP |
| 5 | google_search | Gemini API (grounding) |
| 6 | tavily_search | LangChain / AutoGPT / 通用 Agent |
| 7 | search | 其他通用命名 |
| 8 | bash / run_python / shell 调用本地脚本 | 有 shell 工具的环境（见下方） |
| 9 | 无专用搜索工具 | 见 Fallback 策略 |
FETCH 操作 — 按优先级选第一个可用的：
| 优先级 | 工具名 | 适用环境 |
|---|---|---|
| 1 | kimi_fetch | Kimi (Moonshot) |
| 2 | fetch | Claude+MCP / Cursor MCP |
| 3 | url_context | Gemini 2.0+ |
| 4 | browser_navigate + browser_snapshot | Cursor browser MCP |
| 5 | fetch_url / browse_url | 通用 Agent 框架 |
| 6 | bash / run_python / shell 调用本地脚本 | 有 shell 工具的环境（见下方） |
| 7 | 无 FETCH 工具 | 见 Fallback 策略 |
本地脚本兜底（有 shell/bash 工具时）：
当以上专用工具均不可用，但当前环境有 bash / shell / run_command 类工具时，可调用同目录下的 search_fetch.py：
# Search (via DuckDuckGo, no API key required)
python search_fetch.py search "字节跳动 注册资本 法定代表人" --num 10
# Fetch — default strategy is 'direct' (traffic stays local, no third-party proxies)
python search_fetch.py fetch "https://example.com/announcement.html" --max-chars 12000
# If direct fails and you accept third-party routing (jina/archive), use auto:
python search_fetch.py fetch "https://example.com/page.html" --strategy auto
Data flow:
direct(default) — requests go from your machine straight to the target site.auto/jina/archive— the target URL and page content may pass throughr.jina.aiorarchive.org. Only use these for public URLs; never for internal or sensitive endpoints. 中文站点结果质量取决于网络环境（代理/直连）。
Fallback 策略（工具完全不可用时）：
- 无 SEARCH：尝试用 FETCH 直接抓已知权威站点，或基于已有知识推断（须标注"基于内部知识，未实时验证"）。
- 无 FETCH：仅依赖 SEARCH 返回的摘要/snippet；关键字段标注"仅摘要，未全文核验"。
- 两者均无：告知用户当前环境缺乏实时检索工具，报告仅基于模型训练截止日期的知识，建议用户手动核验。
在报告头部声明当前环境使用了哪些工具（例：
SEARCH=web_search, FETCH=fetch）。
🎯 Output Standard
- 以"天眼查/企查查常见模块"为纲，输出结构化报告
- 每条关键结论尽量做到"至少两处来源交叉验证"
- 对所有关键信息标注：
- 来源 URL/标题
- 抓取日期
- 一致性（多源一致/单源）
- 可信度等级（A/B/C）
- 明确声明：公开搜索 ≠ 付费数据库全量数据；无法获取的字段标注"未检索到/疑似需付费/需内部渠道"。
🧭 Workflow
Step 0 — 工具检测 + 实体识别与消歧（必须做）
Metadata
Not sure this is the right skill?
Describe what you want to build — we'll match you to the best skill from 16,000+ options.
Find the right skillPaste this into your clawhub.json to enable this plugin.
{
"plugins": {
"official-fan31415-company-search": {
"enabled": true,
"auto_update": true
}
}
}