What Are the Available Skills in AI Berkshire? A Complete Guide to the I-6 Framework

AI Berkshire provides six core "I-6" skills—/investment-research, /industry-research, /financial-data, /earnings-review, /portfolio-review, and /bottleneck-hunter—that form a complete investment research workflow defined in Markdown under skills/ and synchronized to codex-skills/.

The available skills in AI Berkshire power an end-to-end investment research engine housed in the xbtlin/ai-berkshire repository. These skills ship as plain-text Markdown specifications that define exact workflows, data validation rules, and CLI commands for rigorous financial analysis. Each skill is designed to be composable, allowing users to chain together research tasks while enforcing a "two-source, <1% error" data discipline.

The Six I-6 Skills Available in AI Berkshire

AI Berkshire organizes its capabilities into six distinct skill packages, each accessible via a slash command. These skills are located in skills/*.md and mirrored as Codex-compatible packages in codex-skills/*/SKILL.md.

/investment-research

Full-cycle research using the four investment masters (Buffett, Munger, Duan Yong-ping, Li Lu). This skill drives comprehensive equity analysis by applying multi-master mental models to individual securities. The source definition lives in skills/investment-research.md and generates codex-skills/investment-research/SKILL.md.

/industry-research

Industry-level analysis including market-size sizing and "civilization-evolution" framing. Use this skill to assess secular trends and competitive dynamics across sectors. Defined in skills/industry-research.md with its Codex counterpart at codex-skills/industry-research/SKILL.md.

/financial-data

Data extraction and rigorous verification of financial numbers. This skill enforces the dual-source verification rule (US equities: Macrotrends + StockAnalysis; HK equities: AASTocks + Macrotrends; CN equities: Eastmoney + 巨潮). It invokes tools/financial_rigor.py for exact-arithmetic calculations to eliminate LLM "mental math" errors. Source: skills/financial-data.md.

/earnings-review

Deep earnings-report reading with data reconciliation and post-earnings commentary. This skill cross-validates transcript data against filings and generates audit trails. Source file: skills/earnings-review.md.

/portfolio-review

Portfolio-level monitoring, risk aggregation, and rebalancing recommendations. This skill aggregates holdings data to assess concentration risk and correlation exposure. Defined in skills/portfolio-review.md.

/bottleneck-hunter

Global supply-chain bottleneck detection and arbitrage opportunity scouting. Use this skill to identify capacity constraints and supply shortages across global markets. Source: skills/bottleneck-hunter.md.

Skill Architecture and Implementation

Understanding how AI Berkshire structures these skills reveals how the repository maintains consistency between human-readable documentation and executable automation.

Skill Definition Format

Each skill is a plain-text Markdown file under skills/ that describes:

• The exact workflow steps
• Required data sources and validation procedures
• CLI command specifications (e.g., calls to tools/financial_rigor.py)

Codex Generation Pipeline

The repository uses scripts/sync-codex-skills.py to read the canonical Markdown sources and render Codex-compatible packages. This synchronization ensures that Claude Code and Codex users receive identical skill definitions in codex-skills/<skill>/SKILL.md.

Tooling Backbone

Core Python utilities under tools/ enforce the Financial Data Discipline:

• tools/financial_rigor.py: Performs exact-arithmetic verification of market-cap calculations, valuation metrics, and financial ratios
• tools/report_audit.py: Generates embedded audit tables for reproducibility and traceability

Workflow Orchestration

Skills are designed to be composable. For example, invoking /investment-research internally chains /financial-data and /earnings-review to gather and cross-validate data before producing a final report. This modularity ensures that data integrity rules propagate through every research operation.

How to Invoke AI Berkshire Skills

After installing the repository via scripts/install-codex-skills.sh, interact with the I-6 skills through the ai-berkshire CLI. Each command produces a Markdown report containing embedded audit tables generated by tools/report_audit.py.

# Run full investment research on a US ticker

ai-berkshire /investment-research --ticker AAPL

# Analyze the semiconductor industry

ai-berkshire /industry-research --sector semiconductor

# Pull and validate financial data for a Chinese A-share

ai-berkshire /financial-data --ticker 600519

# Review latest earnings for a US stock

ai-berkshire /earnings-review --ticker MSFT

# Generate portfolio health check

ai-berkshire /portfolio-review --holdings portfolio.csv

# Scan for AI hardware supply chain bottlenecks

ai-berkshire /bottleneck-hunter --focus AI

Each execution creates a Markdown report in the working directory with links back to the originating skill definition for full traceability.

Key Files Supporting the I-6 Skills

Category	File Path	Purpose
Skill Sources	skills/investment-research.md	Four-master research workflow
	skills/industry-research.md	Industry analysis framework
	skills/financial-data.md	Data-source specifications
	skills/earnings-review.md	Earnings ingestion logic
	skills/portfolio-review.md	Portfolio monitoring rules
	skills/bottleneck-hunter.md	Supply-chain detection
Generated Packages	codex-skills/*/SKILL.md	Codex compatibility layer
Tools	tools/financial_rigor.py	Exact-arithmetic verification
	tools/report_audit.py	Report auditing
Scripts	scripts/sync-codex-skills.py	Skill synchronization
Metadata	AGENTS.md	Layout and compatibility standards

Summary

• AI Berkshire offers six I-6 skills: /investment-research, /industry-research, /financial-data, /earnings-review, /portfolio-review, and /bottleneck-hunter.
• Each skill exists as Markdown in skills/ and is synchronized to codex-skills/ via scripts/sync-codex-skills.py.
• Data integrity is enforced through tools/financial_rigor.py and dual-source verification rules specified in skills/financial-data.md.
• Skills are composable, allowing complex research workflows that chain multiple capabilities while maintaining audit trails via tools/report_audit.py.

Frequently Asked Questions

How do I access the available skills in AI Berkshire?

Install the repository using scripts/install-codex-skills.sh, then invoke skills via the ai-berkshire CLI using slash commands like /investment-research or /financial-data. Each skill is defined in the skills/ directory as a Markdown file that specifies exact workflows and data requirements.

What is the difference between the skills/ and codex-skills/ directories?

The skills/ directory contains the canonical source definitions written in plain Markdown. The codex-skills/ directory contains generated Codex-compatible packages created by scripts/sync-codex-skills.py. This dual structure ensures that both human readers and automated Codex agents use identical skill definitions.

How does AI Berkshire ensure data accuracy across its skills?

Every skill adheres to the Financial Data Discipline mandated in skills/financial-data.md. This requires dual-source verification (e.g., Macrotrends + StockAnalysis for US equities) and uses tools/financial_rigor.py to perform exact-arithmetic calculations, preventing the approximation errors common in LLM "mental math."

Can I combine multiple AI Berkshire skills in a single workflow?

Yes. The I-6 skills are designed to be composable. For example, running /investment-research automatically chains /financial-data and /earnings-review to gather and validate data before generating the final report. This orchestration ensures that validation rules propagate through multi-step research tasks.

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