Company Due Diligence Skill

Overview

This skill enables comprehensive company background research using publicly available information sources. When the user provides a company name, gather and analyze data from multiple sources to produce a structured due diligence report.

Workflow

Step 1: Gather Basic Information

Use web search to collect initial company information:

Search query: "{公司名称}" 企业信息 天眼查
Search query: "{公司名称}" 工商信息 注册资本 参保人数
Search query: "{公司名称}" 公司规模 成立时间

Step 2: Search Legal/ Judicial Risks

Search for legal issues and litigation records:

Search query: "{公司名称}" 劳动仲裁 劳动纠纷
Search query: "{公司名称}" 裁判文书网 诉讼
Search query: "{公司名称}" 被执行人 失信

Step 3: Search Layoff & Sentiment Risks

Search for recent layoff news and employee sentiment:

Search query: "{公司名称}" 裁员
Search query: "{公司名称}" 降薪 拖欠工资
Search query: "{公司名称}" 离职 负面

Step 4: Search on Social Platforms

Use available skills to search on social media:

# 小红书舆情搜索
Use xiaohongshu skill to search: "{公司名称} 避雷"
Use xiaohongshu skill to search: "{公司名称} 裁员"

# 脉脉职场舆情搜索（调用maimai-sentiment skill）
When searching for workplace sentiment, use maimai-sentiment skill with queries like:
- "{公司名称}" 脉脉
- "{公司名称}" 裁员 脉脉
- "{公司名称}" 怎么样 脉脉

Step 5: Compile Due Diligence Report

Structure the final report following the template in references/report_template.md.

Output Format

The final output MUST follow the structured report format defined in references/report_template.md.

Include risk ratings (🟢 Low / 🟡 Medium / 🟠 High / 🔴 Critical) for each dimension.

Data Sources Priority

1. Official sources (preferred when accessible): National Enterprise Credit Information Publicity System
2. Search engines: Consolidate information from multiple search results
3. Social platforms: Xiaohongshu, Weibo, Zhihu for sentiment
4. Third-party platforms: Tianyancha, Qichacha for supplementary data

Important Notes

• Always distinguish between facts (official data) and sentiment (user reviews)
• For litigation records, note the type and recency of disputes
• For sentiment data, note the volume and recency of complaints
• Include a disclaimer that this is based on publicly available information and may not be complete
• If certain data cannot be obtained, note it in the report as "Unable to verify"

⚠️ Template Cleanup Rule (PDF Generation)

CRITICAL: When generating a PDF report, you MUST clean up template residuals from generate_pdf.py:

Before running the script, verify and update ALL of the following for the NEW company:

| Field | Common Template Residuals to Check |

|-------|-----------------------------------|

| company_name | Old company name from previous report |

| basic_info table | Old company data (registration date, capital, etc.) |

| extra_info | Old business info, old subsidiaries count |

| risk_info | Old lawsuit details (e.g., "安居客" content) |

| biz_risk | Old risk items (e.g., IPO failure, financial decline) |

| sentiment_data | Old layoff data (e.g., "20%-30% layoffs") |

| sentiment_details | Old sentiment quotes |

| score_data | Star ratings with old company descriptions |

| conclusion_elements | MOST LIKELY to have residuals - check verdict AND all reasons |

| suggestions | MOST LIKELY to have residuals - old advice like "如果不是走投无路..." |

| sources | Old source list |

Common residual patterns:

• Conclusion: "不建议入职" / "如果不是走投无路，强烈建议避开"
• Sources: References to previous companies (e.g., "安居客", "58同城")
• Sentiment: Previous company's layoff percentages or quotes
• Score descriptions: Old company-specific text

Always verify: Run a search for "安居客", "58同城", "不建议", "走投无路" in the script before generating PDF.

Resources

• Report template: references/report_template.md
• PDF generator: scripts/generate_pdf.py

PDF Report Generation

When the user requests a PDF report (e.g., "生成PDF"、"导出PDF"、"保存到桌面"):

1. Use the scripts/generate_pdf.py script to generate a professional PDF report
2. The script accepts the following data via hardcoded variables:
• company_name: Full company name
• query_date: Report generation date
• Various data tables (basic_info, risk_info, sentiment_data, score_data, etc.)
3. Output file: C:\Users\lee\Desktop\{company_short_name}_公司背调报告.pdf

PDF Generation Steps:

1. Run: python scripts/generate_pdf.py
2. The script will generate a PDF with:
• Professional blue color scheme
• Risk ratings with color coding (Red/Yellow/Green)
• Structured sections for each data category
• Conclusion box with overall recommendation
3. Rename the output file if needed for better readability

Note on Fonts:

• The script uses SimHei font for Chinese text
• Emoji characters are replaced with text markers like [高风险], [中风险], [正常]
• If fonts are not available, the script falls back to Helvetica