# How AI Berkshire Integrates the Four-Master Methodology: A Technical Deep Dive

> Discover how AI Berkshire integrates the Four-Master Methodology. This technical deep dive explains its Markdown-driven workflow, challenge prompts, and data validation for investment research.

- Repository: [Xbt Lin/ai-berkshire](https://github.com/xbtlin/ai-berkshire)
- Tags: deep-dive
- Published: 2026-07-10

---

**AI Berkshire formalizes the investment philosophies of Warren Buffett, Charlie Munger, Duan Yong-ping, and Li Lu into an executable Markdown-driven workflow that enforces master-specific challenge prompts and mandatory data validation at every research stage.**

The xbtlin/ai-berkshire repository implements a rigorous "four-master" (四大师) framework that transforms qualitative investment principles into automated research pipelines. Unlike traditional AI investment tools that treat legendary investors as loose inspiration, this codebase embeds Buffett, Munger, Duan, and Li's methodologies directly into the execution path through structured Markdown skills and enforced validation gates.

## What Is the Four-Master Framework?

The four-master methodology combines the distinct analytical lenses of Warren Buffett (moat and margin of safety), Charlie Munger (mental models and inversion), Duan Yong-ping (business quality and simplicity), and Li Lu (conviction and circle of competence). In AI Berkshire, this isn't merely reference material but a **formalized, executable scaffold** that drives every research operation.

## Three-Layer Architecture Implementation

The integration operates through three distinct architectural layers that ensure the four-master perspective permeates every analysis.

### Skill Definitions Layer

Each research skill is defined as a Markdown-based "slash-command" that explicitly maps seven analysis modules to specific master viewpoints. The [`skills/investment-research.md`](https://github.com/xbtlin/ai-berkshire/blob/main/skills/investment-research.md) file serves as the primary definition, listing analysis steps and associating each module with targeted challenge prompts. Other skill files—including [`skills/industry-research.md`](https://github.com/xbtlin/ai-berkshire/blob/main/skills/industry-research.md), [`skills/earnings-team.md`](https://github.com/xbtlin/ai-berkshire/blob/main/skills/earnings-team.md), [`skills/investment-team.md`](https://github.com/xbtlin/ai-berkshire/blob/main/skills/investment-team.md), and [`skills/industry-funnel.md`](https://github.com/xbtlin/ai-berkshire/blob/main/skills/industry-funnel.md)—extend this framework across different research contexts while maintaining the same seven-step template.

### Execution Pipeline Layer

When users invoke a skill such as `/investment-research 腾讯`, the Claude-Code or Codex runtime parses the Markdown and executes steps in sequence. The pipeline inserts **master-specific "追问" (challenge) prompts** at the end of each analysis module, forcing the AI to confront questions like "段永平式追问：这门生意好在哪？" (Duan Yong-ping style challenge: What makes this business good?). Simultaneously, the workflow triggers mandatory data-validation tools including [`tools/financial_rigor.py`](https://github.com/xbtlin/ai-berkshire/blob/main/tools/financial_rigor.py) and [`tools/report_audit.py`](https://github.com/xbtlin/ai-berkshire/blob/main/tools/report_audit.py) before allowing qualitative judgments to proceed.

### Report Generation Layer

Every final report template requires inclusion of a "四大师模拟点评" (four-master simulated review) section, a confidence-rating grid, and explicit AI-bias disclosures. As defined in [`skills/investment-research.md`](https://github.com/xbtlin/ai-berkshire/blob/main/skills/investment-research.md) (lines 97-100 and 195-197), these constraints ensure that outputs in `reports/`—such as `reports/领益智造/领益智造投资研究报告-20260707.md`—reflect all four investment lenses regardless of sector or asset class.

## Master-Specific Challenge Prompts

The "追问" system represents the core mechanism for methodology integration. After each analysis block in [`skills/investment-research.md`](https://github.com/xbtlin/ai-berkshire/blob/main/skills/investment-research.md) (lines 100-112), the system injects targeted questions that align with specific masters' thinking patterns. These prompts are hard-coded in the Markdown skill files, ensuring that Buffett's margin-of-safety calculations, Munger's inversion heuristics, Duan's business-quality focus, and Li Lu's conviction requirements are systematically applied rather than optionally referenced.

## Data Validation and Rigor

The four-master methodology emphasizes disciplined, data-first analysis. The workflow mandates calling [`tools/financial_rigor.py`](https://github.com/xbtlin/ai-berkshire/blob/main/tools/financial_rigor.py) to verify metrics like market capitalization before any qualitative assessment occurs. Subsequently, [`tools/report_audit.py`](https://github.com/xbtlin/ai-berkshire/blob/main/tools/report_audit.py) audits the draft report to confirm the four-master structure is complete and biases are disclosed. This automation enforces the quantitative rigor championed by Buffett and Munger while maintaining the qualitative depth of Duan and Li.

## Cross-Skill Consistency

All research-related skills share identical seven-step templates, guaranteeing that whether analyzing a single stock, conducting an industry funnel review, or performing an earnings-call deep-dive, the analysis passes through the same four-master lenses. The [`scripts/sync-codex-skills.py`](https://github.com/xbtlin/ai-berkshire/blob/main/scripts/sync-codex-skills.py) utility copies Markdown definitions into `codex-skills/*/SKILL.md` artifacts, preserving methodology consistency across both Claude Code and Codex execution environments as documented in [`AGENTS.md`](https://github.com/xbtlin/ai-berkshire/blob/main/AGENTS.md).

## How to Trigger the Four-Master Workflow

Users initiate the integrated methodology through simple slash commands. The system then executes the full validation and challenge pipeline automatically.

```bash

# Invoke the investment-research skill for a specific ticker

/ investment-research 腾讯

```

Behind the scenes, the parser reads [`skills/investment-research.md`](https://github.com/xbtlin/ai-berkshire/blob/main/skills/investment-research.md) and extracts the seven modules. Step 1 gathers data and runs `tools/financial_rigor.py verify-market-cap`. Steps 2-6 run the analysis blocks, each terminating with master-specific challenges. Step 7 calls [`tools/report_audit.py`](https://github.com/xbtlin/ai-berkshire/blob/main/tools/report_audit.py) to audit the draft before writing the final report to `~/腾讯投资研究报告.md` with the mandatory four-master review section.

## Summary

- AI Berkshire embeds the four-master methodology (Buffett, Munger, Duan Yong-ping, Li Lu) directly into executable Markdown skill definitions rather than treating them as optional guidelines.
- The architecture operates through three layers: **skill definitions** ([`skills/investment-research.md`](https://github.com/xbtlin/ai-berkshire/blob/main/skills/investment-research.md)), **execution pipelines** (Claude-Code/Codex runtime with challenge prompts), and **report generation** (mandatory four-master sections).
- **Master-specific "追问" prompts** are hard-coded into analysis modules to enforce disciplined thinking patterns at every step.
- Mandatory validation via [`tools/financial_rigor.py`](https://github.com/xbtlin/ai-berkshire/blob/main/tools/financial_rigor.py) and [`tools/report_audit.py`](https://github.com/xbtlin/ai-berkshire/blob/main/tools/report_audit.py) ensures data integrity before qualitative judgments.
- The [`scripts/sync-codex-skills.py`](https://github.com/xbtlin/ai-berkshire/blob/main/scripts/sync-codex-skills.py) utility maintains cross-platform consistency by generating Codex-compatible artifacts that preserve the four-master logic.

## Frequently Asked Questions

### What are the four masters in AI Berkshire's methodology?

The four masters are Warren Buffett, Charlie Munger, Duan Yong-ping, and Li Lu. Each contributes distinct analytical lenses: Buffett contributes moat analysis and margin-of-safety calculations, Munger provides mental models and inversion heuristics, Duan focuses on business quality and simplicity, and Li Lu emphasizes conviction investing and staying within one's circle of competence.

### How does the repository enforce the four-master framework during execution?

The framework is enforced through hard-coded Markdown skill files that define seven analysis modules, each terminating with master-specific "追问" (challenge) prompts. Additionally, the execution pipeline mandates calling [`tools/financial_rigor.py`](https://github.com/xbtlin/ai-berkshire/blob/main/tools/financial_rigor.py) for data validation and [`tools/report_audit.py`](https://github.com/xbtlin/ai-berkshire/blob/main/tools/report_audit.py) for structural auditing before generating any final report, ensuring all four perspectives are included.

### Can users modify the four-master challenge prompts?

Yes, because the methodology is implemented in plain-text Markdown files such as [`skills/investment-research.md`](https://github.com/xbtlin/ai-berkshire/blob/main/skills/investment-research.md), users can edit the challenge prompts (lines 100-112) without modifying source code. After editing, running [`scripts/sync-codex-skills.py`](https://github.com/xbtlin/ai-berkshire/blob/main/scripts/sync-codex-skills.py) propagates changes to `codex-skills/*/SKILL.md` artifacts, making the methodology transparent and customizable while maintaining execution integrity.

### Does the four-master methodology work for different types of research?

Yes, the methodology extends across all research contexts through specialized skill files. [`skills/industry-research.md`](https://github.com/xbtlin/ai-berkshire/blob/main/skills/industry-research.md) applies the framework to sector-wide analysis, [`skills/earnings-team.md`](https://github.com/xbtlin/ai-berkshire/blob/main/skills/earnings-team.md) uses it for earnings-call deep-dives, and [`skills/industry-funnel.md`](https://github.com/xbtlin/ai-berkshire/blob/main/skills/industry-funnel.md) employs the "四大师深度分析" (four-master deep analysis) layer to select headline companies, ensuring consistent application regardless of asset class or research scope.