What Are the 19 Specialized Skills in the AI Berkshire Framework?

The AI Berkshire repository does not contain a fixed set of "19" or "I-9" specialized skills; instead, it offers a modular architecture of stand-alone skills defined as Markdown files in the skills/ directory, each exposing specific investment analysis workflows as slash-commands.

The AI Berkshire framework (xbtlin/ai-berkshire) provides a specialized AI environment for institutional investment research. While searches often refer to "19 specialized skills" or an "I-9" skill group, the framework actually organizes capabilities as individual, composable skill files that Claude Code and OpenAI Codex execute as native slash-commands.

Skill Architecture and Organization

The framework’s capabilities are organized as stand-alone "skills" expressed as human-written Markdown files. Each file describes a specific workflow, its parameters, and expected output format.

Component	Role	Location
Skill definitions	Markdown files describing workflows, parameters, and output formats.	skills/*.md
Codex generator	Converts Markdown skills into Codex-compatible artifacts.	scripts/sync-codex-skills.py
Claude installer	Registers Markdown skills as Claude slash-commands.	scripts/install-claude-commands.sh
Helper tools	Python utilities for financial calculations and data retrieval.	tools/*.py

When invoked (e.g., /investment-research), the framework parses the associated Markdown file, executes required helper scripts, and returns a structured report.

Core Specialized Skills Available

The repository includes several production-ready skills. Below are the primary examples documented in the source code.

Investment Research

The skills/investment-research.md file defines an end-to-end workflow for full company analysis. It covers financial statement review, valuation modeling, and risk assessment. When invoked, this skill coordinates with tools/financial_rigor.py to perform exact arithmetic and cross-source validation.

Investment Team Analysis

Defined in skills/investment-team.md, this skill provides a structured methodology for researching an investment team’s background, track record, and decision-making history. It focuses on qualitative due diligence rather than quantitative modeling.

Earnings Review

The skills/earnings-review.md skill implements a structured approach to quarterly earnings call analysis. It extracts key metrics, management commentary, and forward guidance from transcripts and financial releases.

Industry Research

Located at skills/industry-research.md, this skill offers a methodology for sector-wide analysis, including competitive landscape mapping, market sizing, and trend identification across industry verticals.

Technical Implementation

Skill Synchronization

The scripts/sync-codex-skills.py utility converts each human-written Markdown file in skills/ into a Codex-compatible format stored in codex-skills/*/SKILL.md. This ensures parity between Claude and OpenAI Codex implementations.

Command Registration

The scripts/install-claude-commands.sh bash script registers the Markdown skills as native Claude slash-commands, making them available in the Claude Code interface.

Helper Libraries

Skills requiring financial calculations invoke modules from the tools/ directory. The tools/financial_rigor.py library is particularly critical, providing exact arithmetic operations and validation logic that ensure computational accuracy across all investment-related skills.

Discovering and Invoking Skills

You can list available skills by inspecting the filesystem directly. The following Python snippet demonstrates how to enumerate skills and simulate invocation:

import subprocess
from pathlib import Path

# List all skill definition files

skill_dir = Path("skills")
skill_files = sorted(skill_dir.glob("*.md"))

print("Available AI Berkshire skills:")
for f in skill_files:
    print(f" • {f.stem}")

# Example: Invoke the investment-research skill

skill_name = "investment-research"
cmd = [
    "python3",
    "scripts/install-claude-commands.sh",  # Registers the command

    f"/{skill_name}",
    "--input", "Apple Inc. 2024 outlook"
]

result = subprocess.run(cmd, capture_output=True, text=True)
print("\nSkill output:")
print(result.stdout or result.stderr)

In practice, Claude users invoke skills directly:

/investment-research
Company: Apple Inc.
Focus: 2024 outlook

The backend reads skills/investment-research.md, executes the necessary Python helpers, and returns a formatted markdown report.

Summary

• Modular design: Skills are individual Markdown files in skills/, not a fixed "19" or "I-9" group.
• Four documented examples: Investment research, team analysis, earnings review, and industry research.
• Sync workflow: Use scripts/sync-codex-skills.py to generate Codex versions and scripts/install-claude-commands.sh for Claude integration.
• Extensible: Add new skills by creating Markdown files and re-running the sync script.
• Backed by tools: Financial calculations rely on tools/financial_rigor.py for accuracy.

Frequently Asked Questions

Does the AI Berkshire framework have exactly 19 specialized skills?

No, the repository does not specify a fixed count of 19 skills. The reference to "19" likely stems from confusion with "I-9" (which also does not exist as a category). The framework uses a modular architecture where skills are added as individual Markdown files in the skills/ directory, allowing the total count to vary as the repository evolves.

How do I add a new specialized skill to the framework?

Create a new Markdown file in the skills/ directory that defines the workflow, required parameters, and output format. Next, run scripts/sync-codex-skills.py to generate the Codex-compatible version, then execute scripts/install-claude-commands.sh to register the new command with Claude Code.

What is the difference between Claude skills and Codex skills in this framework?

The source of truth is the human-written Markdown in skills/*.md. The sync-codex-skills.py script generates Codex-compatible artifacts in codex-skills/*/, while install-claude-commands.sh registers the same logic as Claude slash-commands. Both interfaces utilize the underlying Python tools in tools/*.py for execution.

Which Python modules support the financial analysis skills?

The tools/financial_rigor.py module provides core functionality for exact arithmetic and cross-source validation. Skills like investment-research and earnings-review invoke this library to ensure computational rigor when processing financial data and generating reports.

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