Industry Screening Skills in AI Berkshire: A Complete Guide to Sector Analysis

AI Berkshire provides two dedicated industry-screening skills—Industry Research and Industry Funnel—that systematically evaluate sectors and identify the most investable opportunities.

AI Berkshire (xbtlin/ai-berkshire) is an open-source investment research framework that includes specialized industry screening skills for filtering global sectors and conducting deep-dive analysis. These skills automate the construction of industry value-chain maps and apply quantitative criteria to surface high-conviction investment themes.

Industry Research Skill: Deep-Dive Sector Analysis

The Industry Research skill, defined in skills/industry-research.md, performs comprehensive industry chain scans using the "four-master" analysis framework. This skill builds and validates investment logic chains, draws complete industry value-chain maps, and scans global listed companies across A-share, Hong Kong, and US markets.

Core Analytical Components

The skill applies the four-master framework to headline companies, analyzing:

• Business model sustainability
• Moat and competitive advantages
• Risk factors and regulatory concerns
• Management quality and governance

Before generating final reports, the skill utilizes tools/report_audit.py to perform data-audit checks and ensure analytical integrity.

Output Format

Running this skill generates markdown reports (e.g., ~/新能源产业链投资研究报告.md) containing:

• Complete industry value-chain maps
• Company comparison tables
• Four-master analysis sections
• Sector-level portfolio allocation recommendations

Industry Funnel Skill: High-Level Sector Screening

The Industry Funnel skill, located in skills/industry-funnel.md, acts as a preliminary filter that rapidly screens dozens of sectors to surface the most promising investment themes. Unlike the deep-research skill, this tool focuses on speed and breadth of coverage.

Screening Criteria and Scoring

The skill defines specific screening criteria including:

• Trend strength and momentum
• Market size and growth potential
• Competitive dynamics
• Regulatory environment stability

It ranks industries by a weighted composite score and outputs a short list of "funnel-passed" sectors ready for deeper investigation.

Integration with Quality Screening

According to the xbtlin/ai-berkshire source code, the Quality-Screen skill (skills/quality-screen.md) works hand-in-hand with the Industry Funnel. After generating the initial sector shortlist, users can apply quantitative quality metrics such as ROE thresholds and cash-flow stability filters to validate the preliminary results.

Codex Prompt Integration

Both industry screening skills are exposed as generated Codex slash-prompts for chat-based invocation:

• codex-prompts/industry-research.md wraps the deep-research functionality
• codex-prompts/industry-funnel.md provides the screening interface

These prompts enable analysts to invoke complex screening workflows directly from command-line chat interfaces without manually constructing CLI commands.

Command Line Usage Examples

The repository includes scripts/run-skill.py to execute screening skills from the terminal.

Running Industry Research

To execute a comprehensive analysis of a specific sector:

# Execute the industry-research skill for the "新能源" sector

python3 scripts/run-skill.py industry-research "新能源"

This command generates a detailed markdown report in your home directory containing the full chain analysis and four-master evaluations.

Running Industry Funnel

To screen all available sectors with default parameters:

# Execute the industry-funnel skill with default screening parameters

python3 scripts/run-skill.py industry-funnel

The output displays a ranked table of sectors (e.g., "新能源", "半导体", "数字医疗") with funnel scores and highlights the top candidates.

Automated Funnel-to-Research Workflow

Combine both skills to automate end-to-end sector screening:

# 1️⃣ Get top-3 sectors from the funnel

SECTORS=$(python3 scripts/run-skill.py industry-funnel --top 3 --plain)

# 2️⃣ Loop through each sector and launch full research reports

for S in $SECTORS; do
    python3 scripts/run-skill.py industry-research "$S"
done

This pipeline first filters for the three highest-scoring sectors, then automatically generates comprehensive research reports for each shortlisted candidate.

Summary

• Industry Research (skills/industry-research.md) provides comprehensive value-chain analysis using the four-master framework and generates detailed investment reports audited by tools/report_audit.py
• Industry Funnel (skills/industry-funnel.md) offers rapid quantitative screening across multiple sectors with weighted scoring algorithms
• Both skills can be invoked via scripts/run-skill.py or through generated Codex prompts stored in codex-prompts/
• Quality-Screen (skills/quality-screen.md) complements the Industry Funnel by applying quantitative financial metrics to validated sector shortlists

Frequently Asked Questions

What is the difference between Industry Research and Industry Funnel?

Industry Funnel acts as a high-level filter that quickly screens dozens of sectors to surface promising themes based on trend strength and market dynamics, while Industry Research conducts comprehensive deep-dives including value-chain mapping and four-master analysis on specific sectors identified as investable.

How do I run industry screening from the command line in AI Berkshire?

Use python3 scripts/run-skill.py industry-research "sector_name" for deep research or python3 scripts/run-skill.py industry-funnel for initial screening, with optional flags like --top 3 to limit results and --plain to enable shell scripting integration.

What output files does the Industry Research skill generate?

The skill generates markdown reports (e.g., ~/新能源产业链投资研究报告.md) containing complete industry chain maps, global company tables, four-master analyses, and sector-level portfolio allocation tables, all validated by the tools/report_audit.py module before publication.

Where are the Codex prompts for industry screening stored?

The generated Codex prompts are stored in codex-prompts/industry-research.md and codex-prompts/industry-funnel.md, enabling chat-based invocation of these screening skills without requiring direct CLI interaction.

{
  "mcpServers": {
    "instagit": {
      "command": "npx",
      "args": ["-y", "instagit@latest"]
    }
  }
}