What Is the Mirror Test in AI Berkshire's Investment Checklist?

The Mirror Test is a mandatory five-sentence articulation gate that forces analysts to concisely justify business essence, moat, management, valuation, and downside risk before any capital allocation decision is approved.

The mirror test serves as the final decision checkpoint in the xbtlin/ai-berkshire repository's systematic investment workflow. After a company clears six preliminary gates—including ability-circle, moat analysis, and safety-margin calculations—the system requires a compressed, mirror-image thesis that exposes any logical gaps or unexamined risks. This mechanism operationalizes Warren Buffett’s capital allocation discipline by ensuring no investment proceeds without a transparent, defensible rationale.

The Five-Sentence Structure

The mirror test mandates exactly five concise statements that collectively form the investment thesis. As defined in skills/investment-checklist.md (lines 75-86), the structure follows a strict template:

1. Business essence – A one-sentence description of what the company manufactures or provides, coupled with the analyst's confirmation that they genuinely understand the operation.

2. Moat – Identification of the core competitive advantage and a judgment on whether that moat is widening or narrowing.

3. Management – An assessment of leadership credibility, trustworthiness, and capital allocation track record.

4. Valuation safety-margin – The purchase price expressed as a fraction of intrinsic value (e.g., "80% of fair value"), demonstrating downside protection.

5. Downside risk – A clear statement confirming that potential losses are controllable and specifying the protective factors.

These five bullet points must be complete and coherent; otherwise, the checklist automatically fails the company. The repository enforces this rule without exception: "5句话说不完整 = 不买。没有例外。" (Five sentences incomplete = do not buy. No exceptions.)

How the Mirror Test Enforces Decision Discipline

The mirror test functions as a hardened gate that prevents "analysis paralysis" and cognitive bias from obscuring flawed theses.

Forces intellectual honesty – Analysts cannot hide behind vague spreadsheets or technical indicators. The requirement to state the thesis in five sentences exposes any fuzzy thinking about competitive advantages or valuation assumptions.

Serves as a real-time sanity check – Before advancing to deeper research stages, the system evaluates the five-sentence block. If any clause is missing or ambiguous, the output immediately displays ❌ 未通过镜子测试 (Failed Mirror Test), halting the workflow regardless of how attractive other metrics appear.

Aligns with Buffett's first rule – By mandating a specific statement on downside controllability, the test enforces the "never lose money" principle before capital commitment.

Evaluation Process and Pass/Fail Criteria

The evaluation logic implemented in scripts/sync-codex-skills.py automates the mirror test validation through a strict parsing sequence:

1. Generation – The system auto-populates placeholders for price, moat description, and risk factors based on data from tools/financial_rigor.py.

2. Human review – The analyst validates the generated five-sentence block; any missing element triggers an automatic fail state.

3. Binary output – The system prints a clear marker: ✅ 通过镜子测试 (Passed) or ❌ 未通过镜子测试 (Failed).

4. Workflow gating – If the test fails, the final checklist marks the company as 未通过 (Not Passed), preventing progression to portfolio consideration.

Practical Usage Examples

Invoke the mirror test through the /investment-checklist command in single-ticker or batch mode.

Analyze a single company:

/investment-checklist 腾讯

Expected mirror test output:

#### 镜子测试

"我以 380 港元 买入 腾讯，因为：
1. 这门生意的本质是 **社交网络 + 数字内容平台**，我理解它；
2. 护城河是 **12 亿用户的社交关系链**，且在变宽；
3. 管理层 **马化腾低调务实、资本配置优秀**，值得信赖；
4. 当前价格相当于内在价值的 **80%**，有足够安全边际；
5. 即使我错了，下行风险可控，因为 **账上净现金>2000 亿、游戏现金流强劲**。"
✅ 通过镜子测试

Batch analysis with automatic comparison:

/investment-checklist 腾讯, 茅台, 拼多多

The system generates individual mirror test blocks for each ticker and aggregates results into a comparison table. Any company missing a complete five-sentence justification displays:

❌ 未通过镜子测试
→ 该公司将在最终表格中标记为 “未通过 Checklist”

Implementation in the Codebase

The mirror test logic spans four critical files in the repository:

• skills/investment-checklist.md – Contains the step-by-step mirror test format and the enforcement rule requiring five complete sentences.

• README.md – Documents the high-level philosophy of the mirror test and its role as the final decision gate.

• tools/financial_rigor.py – Calculates the precise valuation metrics that populate the "price vs intrinsic value" clause in the mirror test.

• scripts/sync-codex-skills.py – Generates the executable Codex skill that parses the /investment-checklist command and renders the pass/fail determination.

Summary

• The mirror test requires exactly five sentences covering business essence, moat, management, valuation, and downside risk.
• Incomplete articulations automatically trigger a ❌ 未通过 status, blocking further analysis.
• The mechanism enforces Buffett-style discipline by forcing transparent, testable theses before capital allocation.
• Implementation resides in skills/investment-checklist.md and scripts/sync-codex-skills.py, with valuation data supplied by tools/financial_rigor.py.
• Batch and single-ticker modes both apply the same rigorous five-sentence standard.

Frequently Asked Questions

What happens if I cannot complete all five sentences in the mirror test?

If any of the five required statements are missing or ambiguous, the system automatically marks the investment as ❌ 未通过镜子测试 (Failed Mirror Test). According to the source code in skills/investment-checklist.md, this failure state prevents the company from advancing to deeper research stages, regardless of other favorable metrics.

How does the mirror test differ from the other six gates in the checklist?

While the six preliminary gates (ability-circle, good-business, moat, management, safety-margin, and discipline) evaluate specific analytical components, the mirror test serves as the final integration gate. It forces a coherent synthesis of all prior analysis into five concise sentences, acting as a "sanity check" that catches inconsistencies the granular gates might miss.

Can the mirror test be bypassed or overridden in the AI Berkshire system?

No. The repository explicitly states "没有例外" (no exceptions) regarding the five-sentence rule. The evaluation logic in scripts/sync-codex-skills.py treats the mirror test as a binary pass/fail gate; there is no configuration option to disable or override this requirement for specific tickers.

Where does the valuation data for the fourth sentence come from?

The fourth sentence requiring a price-to-intrinsic-value fraction pulls data from tools/financial_rigor.py. This module performs discounted cash flow calculations and margin-of-safety mathematics, then injects the resulting percentage into the mirror test template before the human review stage.

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