How the Quality-Screen Skill Filters Non-First-Tier Companies Using 7 Hard Indicators

The quality-screen skill applies seven deterministic financial thresholds—ROE, cumulative free cash flow, interest coverage, gross margin, operating cash flow quality, net margin, and share-base inflation—to mechanically exclude inferior companies, while three strategic exemption clauses rescue high-quality firms in temporary investment or high-turnover phases.

The ai-berkshire repository provides an institutional-grade equity analysis framework that emulates Warren Buffett’s capital allocation discipline. At its core, the quality-screen skill documented in skills/quality-screen.md implements a deterministic filter designed to rapidly eliminate non-first-tier companies before deeper fundamental analysis begins.

The Seven Hard Financial Indicators

The filter evaluates every candidate against seven absolute exclusion rules defined in lines 28–36 of skills/quality-screen.md. A single failure marks the company for exclusion unless rescued by an exemption clause.

• 10-Year Average ROE – Exclude if below 8 %. This measures long-term capital efficiency and whether shareholders’ capital earns above the opportunity cost. Source: L28‑L29.

• 5-Year Cumulative Free Cash Flow – Exclude if negative. This distinguishes accounting profits from actual cash generation. Source: L30‑L31.

• Interest-Coverage Ratio (EBIT / Interest) – Exclude if below 2×. This gauges debt-service safety and the ability to meet interest obligations during downturns. Source: L31‑L32.

• Long-Term Gross Margin – Exclude if below 15 %. This indicates pricing power and the ability to command premiums over production costs. Source: L32‑L33.

• Operating Cash Flow ÷ Net Profit (5-Year Average) – Exclude if below 0.7. This proxies profit quality by measuring the proportion of earnings actually collected as cash. Source: L33‑L34.

• Long-Term Net Profit Margin – Exclude if below 5 %. This assesses risk resilience and the ability of earnings to survive revenue volatility. Source: L34‑L35.

• 5-Year Total Share-Base Inflation – Exclude if above 20 % (excluding M&A-driven dilution). This protects shareholder interests by flagging excessive equity dilution beyond normal growth. Source: L35‑L36.

The Three Strategic Exemption Clauses

The skill applies three rule-sets after hard-indicator failures to prevent over-filtering of high-quality firms in transitional phases. These exemptions convert a failure into an “exempt-pass” status.

• Exemption A (Strategic-Investment Phase) – Rescues young companies (listed < 10 years) with gross margin > 30 % and positive operating cash flow in the last two years, allowing low current ROE. Example: Meituan.

• Exemption B (Deliberate Low-Profit Strategy) – Saves firms intentionally sacrificing short-term margins for market share, provided gross margin > 30 % and the net-profit margin is recovering toward ≥ 5 %. Example: Amazon.

• Exemption C (High-Turnover, Thin-Margin Model) – Ignores gross-margin and net-margin failures for membership, platform, or high-turnover models when ROE > 20 % and OCF/NI > 1.0. Example: Costco.

Execution Pipeline

The quality-screen skill processes requests through six deterministic stages:

1. Input Normalisation – Parses $ARGUMENTS to classify the request as single-stock, industry/market batch, or custom list mode.

2. Scope Determination – Single-stock mode routes the ticker directly to data collection; batch mode triggers a web search or index pull to assemble 10–30 representative tickers.

3. Parallel Data Collection – Autonomous agents gather the seven raw metrics from prioritized sources: company annual reports → brokerage research → financial data platforms.

4. Hard-Indicator Evaluation – Each metric is compared against its exclusion threshold. Failures trigger initial exclusion flags.

5. Exemption Processing – The system evaluates Exemptions A, B, and C against any flagged failures, overturning exclusions where criteria are satisfied.

6. Result Synthesis – Outputs a classification of Pass, Exempt-Pass, Exclude, or Borderline (for insufficient or threshold-proximate data), including pass-rate statistics and sector-quality judgments in batch mode.

Usage Examples

Invoke the skill via the Instagit interface using the patterns defined in codex-prompts/quality-screen.md:

# Single-stock mode – filter three Chinese tech giants

instagit quality-screen "腾讯, 美团, 英伟达"

# Industry mode – filter the top 15 listed Chinese breweries

instagit quality-screen "中国啤酒行业"

# Index mode – filter all constituents of the Hang Seng Index

instagit quality-screen "恒生指数成分股"

The skill returns structured JSON output classifying each entity:

{
  "date": "2026-07-26",
  "total_companies": 18,
  "passed": ["腾讯", "英伟达"],
  "exempt_pass": ["美团"],
  "excluded": ["某低ROE公司"],
  "summary": {
    "pass_rate": "66.7%",
    "sector_quality": "High"
  }
}

Summary

• The quality-screen skill in skills/quality-screen.md applies seven hard financial thresholds to eliminate non-first-tier companies deterministically.
• Exemption clauses A, B, and C prevent false exclusions of high-quality firms in strategic investment, low-profit, or high-turnover phases.
• The filter supports single-stock, batch, and index modes, using parallel data collection from annual reports and research.
• Results are bucketed into Pass, Exempt-Pass, Exclude, or Borderline, with quantitative pass-rate statistics for sector-level assessment.
• Source files: skills/quality-screen.md (primary logic) and codex-prompts/quality-screen.md (CLI wrapper).

Frequently Asked Questions

What happens if a company fails only one hard indicator but passes the other six?

The company is initially marked for exclusion. However, if it satisfies any of the three exemption clauses (A, B, or C), the status converts to Exempt-Pass, allowing it to proceed for further analysis. Only companies failing multiple indicators with no qualifying exemption receive a final Exclude classification.

How does the skill handle data scarcity or incomplete financial histories?

Companies with insufficient data to calculate the seven indicators—or with metrics near the threshold—are classified as Borderline. The skill prioritizes company annual reports and brokerage research to minimize data gaps, but recommends manual review when automated collection yields incomplete 5- or 10-year histories.

Can the threshold values be customized for specific industries or regions?

As implemented in ai-berkshire, the thresholds are fixed constants (e.g., ROE < 8 %, gross margin < 15 %). The deterministic design philosophy favors uniform standards over sector-specific adjustments, though the exemption clauses effectively accommodate structural differences in business models such as platform or membership economics.

What is the difference between "Pass" and "Exempt-Pass" classifications?

Pass indicates the company satisfied all seven hard indicators without requiring exemptions. Exempt-Pass indicates the company failed at least one hard indicator but qualified for an exemption clause (A, B, or C), signaling high underlying quality despite temporary financial metrics that would normally trigger exclusion.

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