Company Investment Research Skill

A research skill for analyzing a company's long-term competitiveness and investment value, compatible with both Codex and Claude Code.

It helps users analyze a company's underlying business, position in the value chain, competitive moat, financial quality, potential competitive attack vectors, valuation-implied assumptions, thesis-breaking conditions, and the key metrics to track over the next three years.

The runtime workflow is based on the English Company Long-Term Competitiveness and Investment Value Research Prompt | Final Industrial Version, with additional safeguards for source traceability, restricted research environments, valuation discipline, and cross-platform use.

Key Features

  • Business boundaries and profit pool analysis
  • Moat assessment validated against financial outcomes
  • Greenfield challenges and asymmetric competitive attacks
  • Bear case and pre-mortem analysis
  • Comparisons of financial quality, peers, and alternatives
  • DCF, relative valuation, cycle normalization, and reverse valuation
  • Three operating modes: online research, user-provided materials, and offline framework
  • Traceable evidence labels, source periods, and calculation methodologies
  • No direct buy, sell, position-sizing, or technical trading recommendations

Repository Structure

company-investment-research-skill/
├── SKILL.md
├── agents/
│   └── openai.yaml
├── assets/
│   ├── icon.svg
│   ├── input-brief-template.md
│   └── tracking-dashboard-template.md
├── evals/
│   ├── cases.yaml
│   └── evaluation-rubric.md
├── examples/
├── docs/
│   └── original-industrial-prompt.md
├── references/
│   ├── research-modes.md
│   ├── evidence-policy.md
│   ├── task-router.md
│   ├── module-library.md
│   ├── valuation-framework.md
│   ├── output-templates.md
│   ├── data-source-checklist.md
│   └── quality-checklist.md
└── scripts/
    └── validate_skill.py

SKILL.md and references/ form the shared core used by both Codex and Claude Code. agents/openai.yaml only provides Codex UI metadata and is ignored by Claude Code.

docs/original-industrial-prompt.md preserves the English source prompt and is not loaded as runtime instructions.

Installation

Codex

Project-level installation:

mkdir -p .agents/skills
git clone [email protected]:bingofreedom/company-investment-research-skill.git \
  .agents/skills/company-investment-research

User-level installation:

mkdir -p ~/.agents/skills
git clone [email protected]:bingofreedom/company-investment-research-skill.git \
  ~/.agents/skills/company-investment-research

Explicit invocation:

Use $company-investment-research to analyze Apple's long-term moat.

Claude Code

Project-level installation:

mkdir -p .claude/skills
git clone [email protected]:bingofreedom/company-investment-research-skill.git \
  .claude/skills/company-investment-research

User-level installation:

mkdir -p ~/.claude/skills
git clone [email protected]:bingofreedom/company-investment-research-skill.git \
  ~/.claude/skills/company-investment-research

Example invocation:

Use the company-investment-research skill to analyze NVIDIA's long-term competitiveness in the AI accelerator market.

Both hosts can automatically select the skill based on the description in SKILL.md, or you can invoke it explicitly.

The skill follows the user's requested output language while keeping all runtime instructions and repository content in English.

Usage Examples

Give me a quick assessment of Apple's long-term moat, focusing on the most important metrics to track over the next three years.

If I had a $500 million budget to challenge Coinbase's exchange business from scratch, what would be the most realistic asymmetric path?

Analyze the quality of a company and use a reverse DCF to explain what growth and margin assumptions are implied by its current price.

Use only the annual report I provide. Do not access the internet or use information outside the supplied materials.

Research Modes

  • Online research mode: Uses the search and reading tools available in the current environment to retrieve the latest primary-source information.
  • User-provided materials mode: Uses only files and data supplied by the user.
  • Offline framework mode: When reliable information is unavailable, provides only inferences, assumptions, validation paths, and a list of required data.

The skill does not assume that Codex or Claude Code has browser access, network permissions, or MCP services configured.

Validation

Run the dependency-free static validation:

python3 scripts/validate_skill.py .

Behavioral evaluation cases and scoring criteria are located in evals/. After modifying any core rules, run the same test set in both Codex and Claude Code.

Limitations

  • Outputs are research aids and do not constitute investment advice.
  • Valuation results depend on assumptions and do not represent a definitive intrinsic value.
  • The skill will not claim that an analysis reflects the latest facts unless it has actually reviewed current primary sources.
  • Different models or tool configurations may produce different results.

License

This project is licensed under the MIT License.