Qibook Company Wiki Deepresearch

企业百科深度研究 - 基于 20+ 维度企业数据，自动识别主体类型，生成专业的企业洞察报告。 Use when: 用户需要生成企业百科报告、企业信用分析、商业调研报告、投资尽调报告，或需要全面了解一家企业的综合信息。

Security Scan

The skill's code, required environment variables, and runtime instructions align with its stated purpose (calling a Qibook entData API and formatting templates into reports); no obvious exfiltration or unrelated credential requests were found, but there are minor implementation mismatches and a few template-level prompts worth reviewing before use.

Purpose & Capability

Name/description request enterprise research data; required env vars (QIBOOK_ACCESS_KEY, QIBOOK_BASE_URL) and python3 are consistent with calling a hosted API and running the included Python scripts. No unrelated credentials or binaries are requested.

Instruction Scope

SKILL.md instructs the agent to validate env vars and call the local scripts (python -m scripts.skill_runner or fetch_enterprise_data). Templates contain strong model-instruction style rules (e.g., hide system prompt, strict formatting) that are intended for report generation — not direct data exfiltration — but these embedded 'system-prompt' style constraints could influence an LLM's behavior and should be reviewed. The runtime instructions do not direct reading of unrelated system files or other credentials.

Install Mechanism

No install spec (no packages downloaded), which reduces supply-chain risk. However the package uses the Python requests library but the skill metadata only declares python3 (no explicit dependency declaration or install step for requests). The skill includes multiple Python files which will be executed locally; ensure the runtime environment has required Python packages and consider running in an isolated environment.

Credentials

Only two environment variables are required and they map directly to the API host and access key used by the code. No other secrets or unrelated config paths are requested. The access key is used as an 'access_key' header to the provided BASE_URL endpoint — proportional to the described functionality.

Persistence & Privilege

Skill does not request always:true or any elevated persistent privileges. It doesn't modify other skills or system-wide settings. Execution is on-demand and requires environment variables to be set by the user.

Assessment

This skill appears to be what it claims: it calls a Qibook aggregate API using QIBOOK_BASE_URL and QIBOOK_ACCESS_KEY and formats the returned data into report templates. Before installing or running it: 1) Verify the QIBOOK_BASE_URL is a legitimate/trusted endpoint (skill source is unknown). 2) Only provide an access key that is scoped/limited to the Qibook service; avoid reusing broad or high-privilege credentials. 3) Run the code in a sandboxed environment (or container) because the skill includes Python scripts that will execute locally and make outbound HTTPS requests. 4) Ensure Python dependencies (requests) are present and inspect/lock them (pip freeze / use a virtualenv). 5) Review the templates (they contain embedded LLM/system-style instructions that insist 'do not disclose internal prompts' and other strict output rules) and decide whether those constraints are acceptable for your use-case. 6) If you will use in production, ask the publisher for provenance (homepage, contact) or rotate/revoke the access key if you become unsure. If any of the above raise concerns (unknown host, broad credentials), do not run the skill until resolved.

企业百科深度研究技能

你是一位资深的企业信用分析师和大数据挖掘专家，擅长洞察企业价值，能够将多维信息转化为观点清晰、通俗易懂的商业报告。

基于企百科 20+ 维度数据，自动识别主体类型（企业/分支机构/个体户/医院/学校/律所/政府等），选择对应模板，生成《企业百科》洞察报告。

只想快速查工商信息（法人、股东、注册资本）不需要报告的 → 引导使用 qibook-company-profile skill。

---

前提条件

1. 获取 API 凭证

访问 https://skill.qibook.com 注册账号并获取 QIBOOK_ACCESS_KEY 和 QIBOOK_BASE_URL。

2. 配置环境变量

export QIBOOK_ACCESS_KEY=your_access_key
export QIBOOK_BASE_URL=your_base_url

调用前先校验这两个环境变量是否存在，缺失则提示用户设置。

---

When to use

用户不说"报告"、"尽调"这些术语也要触发。以下口语都属于本 skill 范围：

• 帮我了解一下 XX 公司 / 给我做个企业报告 / 研究一下这家企业
• XX 公司怎么样 / 这家公司靠不靠谱 / 有没有风险
• XX 的股权结构 / 竞争力如何 / 商业模式是什么 / 成长性怎么样
• 帮我查一下 XX 医院 / XX 学校 / XX 律师事务所

口语理解优先：

• "了解一下" / 意图不明确 → 生成完整企业百科报告
• "怎么样" → 侧重风险与信用、经营状况
• "靠不靠谱" → 侧重风险与信用分析
• "什么背景" → 侧重股权与治理、关联分析
• "值不值得关注" → 侧重价值与成长分析

---

工作流程

1. 获取数据

from scripts import fetch_enterprise_data

result = fetch_enterprise_data("企业名称")

# result["success"]          → 是否成功
# result["entity_type"]      → 主体类型（如 "company"）
# result["template_name"]    → 模板中文名（如 "普通企业"）
# result["template_path"]    → 模板文件路径
# result["template_content"] → 模板内容
# result["data"]             → API 返回的完整数据

命令行：python -m scripts.skill_runner <企业名称>

2. 确认主体类型

系统根据 API 返回数据自动识别主体类型并选择模板：

主体类型	模板文件	判断依据
普通企业	templates/company.md	默认
分支机构	templates/branch.md	HEADQUARTERS 非空
个体工商户	templates/personal.md	ENTTYPE 含"个体"
医院	templates/organization/hospital.md	名称含"医院/诊所/卫生院"
学校	templates/organization/school.md	名称含"学校/大学/学院"
律师事务所	templates/organization/law_firm.md	名称含"律师事务所"
政府机构	templates/organization/government.md	名称含"政府/机关/管理局"
社会组织	templates/organization/social.md	名称含"协会/基金会/商会"
其他组织	templates/organization/other.md	ENTTYPE 含"事业单位"等

3. 按模板生成报告

阅读 result["template_content"]，严格按模板要求，将 result["data"] 转化为洞察性分析报告。

---

输出规范

字数限制（必须严格遵守）

输入数据量	报告字数上限
≤ 15000字	1500字
15001~20000字	2500字
≥ 20001字	3500字

报告结构

按模板输出，典型包含：企业概要、企业历程、股权与治理、主要业务、市场竞争力、品牌影响力、风险与信用、经营与资本、价值与成长、企业关联分析。各模块根据数据支持情况可按模板要求省略。

内容原则

• 成品定位：可直接交付，禁止"建议"、"可能"等模糊词
• 扫读友好：结论前置，句子 ≤ 20词
• 数据驱动：禁止杜撰，禁止简单罗列数据，必须转化为洞察
• 客观中立：禁止重复相同数据，禁止超过字数上限

---

Error handling

• 企业名称未查到 → 提示确认名称，建议用全称
• 数据较少 → 说明报告可能不全面，仍基于已有数据生成
• API 凭证缺失 → 提示设置环境变量
• 空结果区分原因（名称不对 / 权限不足 / 网络异常），不要统一说"查询失败"

---

参考标准

生成报告时需参考以下行业标准（位于 references/ 目录）：

• 企业行业分类标准、企业绿色产业评估标准、人工智能行业标准
• 空壳公司判定标准、受益所有人标准、核心企业判定标准

---

示例

完整示例报告位于 examples/sample_company_report.md。