Company Investment Research

Company Long-Term Competitiveness and Investment Value Research

Objective

Help long-term investors build a structured, falsifiable view of a company. Do not provide direct buy, sell, hold, or position-sizing recommendations.

Keep the analysis centered on one question:

> Can this company convert its current advantages into sustained revenue growth, margins, free cash flow, and high returns on invested capital, and what assumptions are already embedded in the valuation?

Core Principles

1. Define the company's real business before discussing moat or valuation.
2. Examine strengths, counterexamples, bypass paths, and thesis failure conditions.
3. Connect every moat claim to revenue growth, margins, free cash flow, ROIC, retention, pricing power, or unit economics.
4. Separate company quality, valuation attractiveness, and market-price direction.
5. Never present unverified or stale information as current fact.
6. Do not assign precise probabilities without evidence. Use high, medium, or low and explain the basis.
7. Do not use macro risk, competition, or regulation as substitutes for company-level causal analysis.
8. If evidence is insufficient, write: 【Currently cannot be confirmed; further verification required】.

Workflow

1. Identify the Task

Identify the company or business, market, scope, investment horizon, desired depth, and whether the user wants valuation, peer comparison, competitor simulation, or a pre-mortem.

Read references/task-router.md and choose the smallest sufficient mode. Do not default to a full report.

If inputs are incomplete, proceed with reasonable defaults but label them 【Assumption requiring verification】.

2. Identify the Research Environment

Read references/research-modes.md and select:

• Online research mode;
• User-provided materials mode;
• Offline framework mode.

Use the search, web, file, and calculation capabilities available in the current host. Do not assume that Codex or Claude Code has a specific tool, and never fabricate tool use, file access, or retrieval results.

3. Establish the Evidence Boundary

For company facts, financial data, market share, customer data, valuation inputs, or "latest/current" claims, read and follow references/evidence-policy.md.

Use one of these labels for every material claim:

• 【Confirmed fact | source summary】
• 【Reasonable inference based on facts】
• 【Assumption requiring verification】

Only primary sources actually opened and read may support a confirmed fact. Downgrade any claim whose source, period, or accounting basis cannot be verified.

4. Perform the Analysis

Read only the references needed for the selected mode:

• references/module-library.md: business boundary, value chain, moat, attack simulation, fragility, financial validation, peer comparison, and tracking;
• references/valuation-framework.md: valuation, reverse valuation, cycle normalization, and sensitivity analysis;
• references/data-source-checklist.md: source discovery and data verification;
• references/output-templates.md: structured output templates.

Prefer explicit causal chains:

Resource or capability
→ customer value and competitive behavior
→ pricing, retention, cost, or scale effect
→ revenue, margins, FCF, and ROIC
→ valuation expectations and falsification conditions

5. Produce the Conclusion

Match the selected mode rather than mechanically filling every table. At minimum, state:

1. Research mode and information cutoff date;
2. The company's real business;
3. Its most important advantage and whether financial results validate it;
4. The strongest bear argument or failure mechanism;
5. Critical information still requiring verification;
6. The most important three-year tracking indicators.

If valuation is requested, separate "company quality" from "valuation assessment." Disclose the method, assumptions, source periods, and sensitivity. Do not present a model output as definitive intrinsic value.

6. Run the Quality Check

Before finalizing, read and apply references/quality-checklist.md.

Full Research Report

Use this structure only when the user explicitly requests a full, deep, investment-committee, or industrial analysis:

1. Research mode, source status, and evidence boundary;
2. Business boundary;
3. Value chain and profit pool;
4. Building from zero;
5. Moat and financial validation;
6. Head-on and asymmetric competitor attacks;
7. Defense, counterexamples, bypass paths, and pre-mortem;
8. Financial and business-model quality;
9. Peers and substitutes;
10. Valuation and market-implied expectations, if requested;
11. Long-term tracking and thesis failure conditions;
12. Primary classification or classification deferred.

Primary Classification

When evidence is sufficient, select one primary category:

1. Short-term cyclical opportunity driven by industry momentum;
2. Overvalued market narrative;
3. Growth company with an unproven business model;
4. Mature company with a strong moat;
5. Company turning investment into long-term barriers;
6. Company whose moat is weakening and requires caution.

If critical information is missing or the categories cannot be distinguished reliably, output:

> 【Classification deferred | insufficient evidence】Missing: ___. A reliable primary classification requires verification of ___.

You may add one or two secondary feature tags, but do not use a blended primary classification to avoid making a decision.

Output Rules

• Respond in the user's language. Use English only when the user's preference is unclear.
• Keep tables to seven columns or fewer; split wide tables.
• State the evidence level, causal basis, and verification path for material conclusions.
• Do not fabricate historical failure cases. Label comparisons as 【Analogy case, not direct evidence】.
• Do not provide direct trading recommendations, target position sizes, or short-term price predictions.
• Do not fill a template with unsupported information.

Reference Routing

• Every task: read references/task-router.md and references/research-modes.md.
• Facts or data: read references/evidence-policy.md.
• Analytical modules: read references/module-library.md.
• Valuation: read references/valuation-framework.md.
• Source discovery: read references/data-source-checklist.md.
• Structured output: read references/output-templates.md.
• Before final output: read references/quality-checklist.md.

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: bingofreedom
• Source: bingofreedom/company-investment-research-skill
• License: MIT
• Homepage: https://github.com/bingofreedom/company-investment-research-skill

Install and usage instructions live in the source repository linked above.