Difference Between Investment-Research and Investment-Team Skills in AI: A Complete Technical Guide

The investment-research skill generates comprehensive, data-validated company reports through a seven-step analytical pipeline, while the investment-team skill synthesizes those findings into concise decision memos with voting recommendations and action items for investment committees.

The xbtlin/ai-berkshire repository implements specialized AI agent skills designed for Berkshire Hathaway-style value investing workflows. Understanding the difference between investment-research and investment-team skills in AI is critical for orchestrating the complete research-to-decision pipeline, as each skill targets distinct operational phases with different validation requirements and output formats.

Architectural Purpose and Target Audience

Each skill serves a specific function within the investment workflow, with fundamentally different goals and intended users.

Investment-research functions as the analytical foundation layer. It targets individual analysts or autonomous AI agents that must produce comprehensive research from scratch. Located at skills/investment-research.md, this skill enforces rigorous data collection through seven sequential modules and mandates validation via tools/financial_rigor.py before generating recommendations.

Investment-team operates as the decision coordination layer. Found in skills/investment-team.md, this skill targets investment committees and portfolio managers who need synthesized briefings rather than raw analysis. It assumes the underlying research exists and focuses on extracting actionable insights, assigning responsibilities, and facilitating consensus through structured voting mechanisms.

Workflow Structure and Validation Requirements

The technical workflows differ significantly in complexity, validation steps, and output specifications.

Investment-Research: Seven-Step Validation Pipeline

The research skill implements a mandatory sequential process that embeds financial rigor checks at multiple stages. Each module must complete data verification before proceeding, utilizing tools/financial_rigor.py to cross-reference market capitalization, valuation metrics, and financial data across sources.

Key workflow characteristics include:

• Strict adherence to the "four masters" analytical framework (Buffett, Munger, Duan Yong-ping, Li Lu)
• Mandatory "master-style questions" (追问) at the end of each module
• Integration with tools/report_audit.py for random data-point extraction and verification
• Generation of long-form markdown reports (<company>投资研究报告.md) containing business model analysis, moat evaluation, risk checklists, and valuation tables

Investment-Team: Synthesis and Decision Framework

The team skill implements a collaborative workflow that assumes research completion. Rather than validating raw data, it focuses on information distillation and decision logistics.

Core workflow components include:

• Automatic extraction of key highlights and red flags from the research report
• Generation of decision matrices with Buy/Hold/Sell recommendations and confidence scores
• Assignment of specific action items to analysts, traders, and risk officers
• Production of concise memos (<company>投资团队决策.md) designed for committee discussion

Input and Output Specifications

The interface contracts for each skill reflect their distinct purposes.

Investment-research accepts a ticker or company name as its ARGUMENTS parameter. It outputs a comprehensive markdown document containing:

• Data collection summaries
• Business model and moat analysis
• Risk assessment checklists
• Management evaluation
• Final decision tables with numerical justification

Investment-team accepts either the path to a previously generated research report or a company name (auto-loading the latest report). It outputs a structured decision memo containing:

• Executive summary of research findings
• Critical risk highlights
• Actionable task assignments
• Voting recommendations for the investment committee

Key Technical Dependencies

Both skills rely on specific utility modules within the repository architecture.

File	Purpose	Skill Usage
skills/investment-research.md	Core research workflow definition	Primary resource for research generation
skills/investment-team.md	Decision memo generation logic	Primary resource for team coordination
tools/financial_rigor.py	Numerical verification and cross-source validation	Required by investment-research
tools/report_audit.py	Random data-point auditing	Used by both skills for quality assurance

Practical Implementation Example

The following Python script demonstrates how to invoke both skills sequentially using the repository's Codex-style command runner via scripts/sync-codex-skills.py:

import subprocess
from pathlib import Path

def run_skill(skill_name: str, args: str) -> str:
    """
    Executes a skill via the generated Codex command.
    Returns the generated markdown output.
    """
    cmd = [
        "python3", "scripts/sync-codex-skills.py",
        "--run", f"{skill_name}:{args}"
    ]
    result = subprocess.run(cmd, capture_output=True, text=True, check=True)
    return result.stdout

# Generate comprehensive research report for company 300750

research_output = run_skill("investment-research", "300750")
Path("宁德时代_研究报告.md").write_text(research_output)

# Produce team decision memo based on the research

team_output = run_skill("investment-team", "宁德时代_研究报告.md")
Path("宁德时代_团队决策.md").write_text(team_output)

This implementation first executes the rigorous research pipeline to create the analytical foundation, then processes that output through the team skill to generate a committee-ready decision document.

Summary

• Investment-research executes a seven-step analytical pipeline with mandatory data validation via tools/financial_rigor.py, producing comprehensive reports suitable for deep due diligence.
• Investment-team synthesizes existing research into concise decision memos with voting recommendations and action items, targeting investment committee workflows.
• The research skill requires strict validation at each module, while the team skill assumes data accuracy and focuses on logistical coordination.
• Both skills integrate with tools/report_audit.py for quality assurance, but only investment-research depends on tools/financial_rigor.py for numerical verification.
• Sequential execution through scripts/sync-codex-skills.py creates a complete research-to-decision automation pipeline.

Frequently Asked Questions

Can the investment-team skill function without prior investment-research output?

No. According to the repository architecture in skills/investment-team.md, the team skill is designed to process existing research reports. While it can accept a company name and attempt to auto-load the latest research file, its workflow assumes the underlying analytical work has been completed and validated through the investment-research pipeline.

What specific validation does the investment-research skill enforce?

The investment-research skill mandates financial rigor checks through tools/financial_rigor.py, which verifies market capitalization, valuation metrics, and cross-source numerical consistency. Additionally, it employs tools/report_audit.py to randomly extract data points for cross-checking before final report generation, ensuring no section proceeds without validated data.

How do the output formats differ between these AI skills?

The investment-research skill generates long-form analytical documents (<company>投资研究报告.md) containing detailed sections on business models, competitive moats, risk assessments, and valuation tables. The investment-team skill produces concise decision memos (<company>投资团队决策.md) with executive summaries, red-flag highlights, responsibility assignments, and voting matrices designed for rapid committee review.

Which skill should I modify to customize decision-making criteria?

Modify skills/investment-team.md to adjust decision frameworks, voting thresholds, or action item templates. The investment-research skill should remain focused on objective data collection and validation; subjective decision criteria and committee-specific workflows belong in the team skill configuration.

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