跳至正文
OpenAgentSkill
浏览已开放 · 目录记录日期 2026-11-21 · 自动更新和账户服务暂时关闭。
金融

quantitative-screening

Quantitative stock screening, forward-looking valuation outlier detection, backward-looking financial statement validation, PEG ratio analysis, earnings growth profile assessment, turnaround vs value trap discrimination, data mining bias prevention

agentii-ai · 已有 AI 审核记录

安装与使用

运行命令前先阅读来源与依赖说明。复制命令不会自动安装或执行。

npx skills add agentii-ai/agentii-investment-intelligence --skill quantitative-screening

技能说明


name: quantitative-screening description: Quantitative stock screening, forward-looking valuation outlier detection, backward-looking financial statement validation, PEG ratio analysis, earnings growth profile assessment, turnaround vs value trap discrimination, data mining bias prevention multi_ticker_semantics: single_target temporal_scope: default_quarters: 8 max_quarters: 20 description: "Multi-year financial data required for trend analysis; 8 quarters default." allowed_tools:

  • search_investment_strategies
  • get_investment_strategy
  • search_investment_cases retrieval_scope: structured_only layer_tags: ["L2"] min_tool_diversity: 2 parameter_free: false

Methodology fused from professional trading and investment frameworks; all text is an original paraphrase.

Defaults

ParameterDefault ValueRationale
screening_universeS&P 500 + Russell 1000 liquidBroad enough for diversity, liquid enough for execution
historical_years5Minimum years of financial data for trend analysis
peg_threshold1.0PEG < 1.0 suggests undervaluation relative to growth
fcf_conversion_min70%FCF/Net Income below 70% flags earnings quality issues
earnings_beat_threshold70%Beat frequency above 70% suggests conservative guidance

Preflight

Run canonical pre-flight per contracts/preflight.md. Propagate X-Agentii-Trace per contracts/x-agentii-trace-header.md.

Data Source Priority

  1. Quantitative methodology — references/quant-methodology.md (bundled screening framework)
  2. Financial data — SEC XBRL facts via agentii MCP for historical financials
  3. Market data — ~~market_data placeholder for real-time valuation multiples
  4. Strategy frameworks — search_investment_strategies(domain=fundamental, kind=screening)

Methodology

Retrieval Scope

structured_only

Retrieval Strategy

Ownership & insider signals: search_institutional_holdings (top-10 holders + whale portfolios, direction=accumulating|reducing|new|exited) and search_insider_trades (Form-4 transactions with SEC URLs) are available as signal inputs.

Branch (a) Structured Data Query from contracts/retrieval.md: primary retrieval via XBRL facts for financial statement data. Supplement with search_investment_strategies for screening methodology validation. Detailed methodology in references/quant-methodology.md.

Temporal Scope

See frontmatter temporal_scope block.

Tool Allowlist

See frontmatter allowed_tools.

Protocol

This skill implements a two-directional screening process: forward-looking valuation discovery and backward-looking financial statement validation. Core principle: the market is mostly efficient. An outlier exists because either the market is wrong (your edge) or you are missing something. Non-participation is always an option.

Detailed methodology: peer selection protocol, turnaround financial scorecard, 7-step sector cleaning, and data mining bias catalog are in references/quant-methodology.md.

Foundational principle: P/E measures what the market is willing to pay for forward earnings — it is a market psychology metric, not intrinsic value. "Cheap" and "expensive" are not analytical conclusions. The question is: why has the market assigned this multiple? PEG < 1.0 is not a universal buy signal — calibrate sector-relatively, growth-rate-adjust, and cross-check with EV/EBITDA-to-Growth. This skill uses PEG as a screening filter only; for a standalone PEG-based valuation, defer to the peg-valuation skill.

Steps

  1. Universe and Macro Filter: Apply portfolio bias from orchestrator. Long → $3B-$10B mid-caps. Short → $20B+ large caps. Neutral → both, emphasize pairs. Weight sectors by macro regime preferences.

  2. Forward-Looking Valuation Scan: Screen using four-pillar framework (PE1, PE2; EG1, EG2; PEG1, PEG2; revenue multiples). Rank by deviation from sector median. Top/bottom decile advance. Calibrate PEG sector-relatively. Use EV/EBITDA-to-Growth as cross-check; prefer EBIT over EBITDA for capital-intensive sectors.

  3. Backward-Looking Financial Validation (execute in this order):

    • Revenue: growth trajectory, organic vs. acquisition quality, concentration risk
    • Earnings quality: GAAP vs. non-GAAP (> 20% gap = investigate), SBC > 10% revenue = red flag, "non-recurring" in 3+ of 4 quarters = recurring
    • Margin: gross margin trend, incremental margins (> 50% strong, < 20% weak)
    • Cash flow: FCF/Net Income conversion. > 80% excellent, 70-80% acceptable, 50-70% explain, < 50% hard stop for longs. DSO + inventory both rising = channel stuffing risk.
  4. Peer Selection (dual-path): Sector path (GICS → 10-K competition → sell-side → merger docs) + Fundamentals path (cluster by growth, margins, ROIC). Must converge on 4-6 names. Divergence = classification error. Use median. For a formal benchmarked peer set, hand off to peer-bench; for full multiple spreading and calendarization, hand off to comps — do not rebuild either here.

  5. Growth Profile and Trap Detection: EPS CAGR 3-5yr (consistency > magnitude). Estimate trajectory: rising + rising = aligned; falling + rising = danger. Beat/raise = strongest signal. Decompose growth source (revenue vs. cost-cutting vs. buybacks). Turnaround scorecard (0-10): 7-10 investigate long, 0-3 avoid/short. Exclude revenue-growth stories from turnaround classification. Scan for data mining biases.

  6. Sector Cleaning (when data errors suspected): Apply 7-step protocol from reference. Only clean < 20 candidates that pass initial screen.

  7. Output: Score each candidate (valuation × validation × growth). Flag GREEN/AMBER/RED. Handoff: ranked list, peer data, turnaround scores, data quality flags.

Output File

{ticker}/{YYYY-MM-DD_HHMM}_quantitative-screening_{affix}.md

Output Structure

  1. Executive Summary — Universe scanned, outliers found, top 5 candidates ranked
阅读完整来源 ↗