# How to Use the Quality-Screen Skill with Hard Indicators for Stock Filtering

> Master stock filtering with the quality-screen skill and hard indicators. Learn to automatically exclude low-quality stocks using key financial thresholds from xbtlin/ai-berkshire.

- Repository: [Xbt Lin/ai-berkshire](https://github.com/xbtlin/ai-berkshire)
- Tags: how-to-guide
- Published: 2026-07-11

---

**The quality-screen skill in the xbtlin/ai-berkshire repository applies seven non-negotiable financial thresholds—ROE, free cash flow, interest coverage, margins, cash conversion, and equity dilution—to automatically exclude low-quality stocks from investment consideration.**

The **quality-screen** skill is a built-in AI Berkshire workflow that automates rigorous Buffett-style quality criteria through hard-coded financial filters. Located in the [`skills/quality-screen.md`](https://github.com/xbtlin/ai-berkshire/blob/main/skills/quality-screen.md) specification, this tool evaluates companies against first-class quality standards to eliminate poor investments before deeper analysis begins.

## The Seven Hard Indicators

The skill defines seven "hard" financial metrics in [`skills/quality-screen.md`](https://github.com/xbtlin/ai-berkshire/blob/main/skills/quality-screen.md) that serve as binary pass/fail gates. Any company failing a single metric is immediately excluded unless an exemption rule applies.

- **10-year average ROE**: Excludes companies below **8%**, measuring capital efficiency and the ability to generate returns above the cost of equity.
- **5-year cumulative free cash flow**: Excludes **negative** values, distinguishing real cash generation from accounting earnings.
- **Interest-coverage ratio**: Excludes ratios below **2×** (EBIT ÷ Interest), assessing debt service capacity.
- **Long-term gross margin**: Excludes margins below **15%**, indicating pricing power and product differentiation.
- **Operating-cash-flow ÷ Net profit (5-year average)**: Excludes ratios below **0.7**, verifying earnings quality and cash conversion.
- **Long-term net profit margin**: Excludes margins below **5%**, testing resilience against revenue fluctuations.
- **5-year total equity dilution**: Excludes dilution above **20%** (excluding M&A), protecting shareholder interests from excessive share issuance.

## How the Quality-Screen Skill Works

According to the implementation in [`codex-skills/quality-screen/SKILL.md`](https://github.com/xbtlin/ai-berkshire/blob/main/codex-skills/quality-screen/SKILL.md), the workflow executes through three distinct stages:

### Input Parsing

The skill determines the screening scope by parsing whether you provided a single ticker, an industry sector, an index constituent list, or a thematic universe.

### Parallel Data Collection

A background agent retrieves required financial figures from company filings, broker research, and financial data platforms simultaneously.

### Rule Evaluation

The system applies the **"7+2" rule set**—seven hard filters plus two exemption rules. Any failure on a hard metric triggers immediate exclusion unless an exemption applies (e.g., a strategic investment phase for a young, high-margin company). This design guarantees **no false positives**, preferring to miss potential winners rather than retain clear losers.

## Invoking the Quality-Screen Command

The skill exposes a slash-command `/quality-screen` documented in [`README_EN.md`](https://github.com/xbtlin/ai-berkshire/blob/main/README_EN.md). Provide comma-separated tickers for single-stock mode or natural-language phrases for batch screening:

```text
/quality-screen 腾讯, 美团, 英伟达
/quality-screen 恒生指数成分股
/quality-screen 全球AI算力链
/quality-screen 中国啤酒行业

```

When processing indices or themes, the skill first searches for constituent lists (typically 10-30 top companies) then applies the hard-indicator filter to each.

## Understanding the Output

The skill generates a markdown table showing for each ticker:
- **Pass (✅)** or **fail (❌)** status per indicator
- Numeric values used for the decision
- A brief "Reason" column explaining exclusion triggers
- Summary statistics including pass-rate and industry ranking

Example output for single-stock mode:

```markdown
| 股票 | ROE (10 yr) | FCF 5yr | 利息覆盖 | 毛利率 | 运营/净利 | 净利率 | 股本膨胀 | 结论 |
|------|-------------|---------|----------|--------|-----------|--------|----------|------|
| 腾讯 | 12% ✅ | 45B ✅ | 3.8× ✅ | 43% ✅ | 1.2 ✅ | 25% ✅ | 7% ✅ | ✅ Pass |
| 美团 | 6% ❌ | 2B ✅ | 5.1× ✅ | 31% ✅ | 1.1 ✅ | 8% ✅ | 12% ✅ | ❌ Fail (ROE) |

```

The prompt wrapper in [`codex-prompts/quality-screen.md`](https://github.com/xbtlin/ai-berkshire/blob/main/codex-prompts/quality-screen.md) forwards requests to the skill engine, which returns the formatted results.

## Summary

- The **quality-screen** skill applies seven hard financial thresholds defined in [`skills/quality-screen.md`](https://github.com/xbtlin/ai-berkshire/blob/main/skills/quality-screen.md) to filter stocks.
- Metrics include ROE (>8%), cumulative FCF (positive), interest coverage (>2×), gross margin (>15%), cash conversion (>0.7), net margin (>5%), and equity dilution (<20%).
- The **"7+2" rule set** ensures non-negotiable exclusion unless exemption criteria apply, eliminating false positives.
- Invoke via `/quality-screen` with tickers, indices, or thematic phrases.
- Source files include [`skills/quality-screen.md`](https://github.com/xbtlin/ai-berkshire/blob/main/skills/quality-screen.md), [`codex-skills/quality-screen/SKILL.md`](https://github.com/xbtlin/ai-berkshire/blob/main/codex-skills/quality-screen/SKILL.md), and [`codex-prompts/quality-screen.md`](https://github.com/xbtlin/ai-berkshire/blob/main/codex-prompts/quality-screen.md).

## Frequently Asked Questions

### What happens if a company fails only one hard indicator?

The company is immediately excluded from the pass list unless it qualifies for one of the two exemption rules. The quality-screen skill treats the seven hard indicators as non-negotiable gates according to the specification in [`skills/quality-screen.md`](https://github.com/xbtlin/ai-berkshire/blob/main/skills/quality-screen.md).

### Can I adjust the threshold values for the hard indicators?

No. The thresholds are hard-coded in the skill specification to maintain consistency with first-class quality standards. The design intentionally prevents customization to avoid diluting the screening criteria.

### Where does the financial data come from?

The skill uses a parallel data collection process that aggregates figures from company filings, broker research, and financial data platforms. This happens automatically when you invoke the `/quality-screen` command.

### How is the quality-screen skill different from other screening tools?

Unlike flexible screeners that allow partial matches, the AI Berkshire quality-screen skill implements a **"7+2" rule set** that prioritizes specificity over sensitivity. It guarantees no false positives by excluding any company that fails hard metrics, whereas traditional tools might rank companies on a sliding scale.