valuation-router

Use this skill to classify a public company by economic characteristics and select primary, cross-check, downside, implied-expectation, and overlay valuation models. Trigger for stock analysis, intrin

SKILL.md

GitHub

Valuation Router Skill

Purpose

Select primary, cross-check, downside, implied-expectation, and overlay-specific valuation models by classifying the company through economic characteristics, not by company name alone.

Trigger Examples

  • "Which valuation model should I use for UNH / NVDA / AAPL?"
  • "Classify this company before running valuation."
  • "Compare AAPL, GOOGL, and AMZN as quality compounders."
  • "Is this a bank, compounder, managed care company, platform, REIT, or cyclical?"

Inputs

  • Target company / ticker
  • Market
  • Industry
  • Business segments
  • Revenue / earnings / cash-flow profile
  • ROE / ROIC and capital intensity
  • Dividend, buyback, share cancellation, and debt reduction history
  • Segment disclosure for technology-related businesses
  • Cyclicality exposures
  • Available data and freshness status
  • Structured assumptions and evidence for material forecast, normalization, multiple, discount, probability, or recovery inputs
  • User requested depth
  • Prior assumptions, if any

Required Inputs

Minimum usable routing requires target name or ticker plus at least one of: industry, business description, segment profile, revenue / earnings / cash-flow profile, or explicit user-provided classification clues.

If the request is current and valuation/action guidance will follow, current price, latest filing period, and freshness status should be supplied or marked missing.

Outputs

  • Base Business Type
  • Overlay Tags
  • Structured valuation model stack
  • Missing data / blocked models
  • Assumption confidence and blocked / low-confidence assumption items
  • Confidence level
  • Next required action

Blocking Behavior

Do not invent company type from ticker alone when evidence is missing. If classification evidence is insufficient, return:

Valuation Routing Blocked:
- Missing company / industry evidence:
- Models that cannot be selected:
- Minimum data needed:

If only partial classification is possible, return a low-confidence route and defer non-material overlays.

Script Mapping

  • Routing implementation: scripts/routing/select_valuation_models.py
  • Unified execution handoff: scripts/valuation/valuation_executor.py
  • Structured assumption policy: references/valuation_rules/structured_assumption_policy.md
  • Specialist routes: references/valuation_rules/specialized_company_routes_v17.md

Company Classification Flow

Step 1: Base Type

Classify as Mature Quality Compounder if the company has mature operations, stable earnings, durable FCF conversion, above-industry ROE / ROIC, and a defensible competitive position.

Default models:

  • Primary: Owner Earnings DCF / FCFE DCF
  • Cross-check: EPV / Quality-adjusted P/E / Normalized P/E
  • Downside: No-growth EPV
  • Implied: Reverse DCF

Step 2: Dividend / Shareholder Return Overlay

Add this overlay when the company has meaningful dividends, buybacks, share cancellation, or debt reduction that are supported by owner earnings / FCFE and balance-sheet strength.

Additional models / checks:

  • Two-stage DDM
  • Gordon Growth terminal check
  • Dividend Yield Band
  • Dividend Yield + Net Buyback Yield + Debt Reduction Yield
  • Dividend Safety / payout stress test
  • Implied Dividend Growth

If dividend coverage is weak, do not use DDM as valuation support; treat it as a risk flag.

Step 3: Technology Optionality Overlay

Add this overlay when technology-related businesses are financially visible, commercially monetized, and potentially material.

Technology treatment:

  • Efficiency-only technology: reflect in margin, ROIC, FCF conversion, and moat quality.
  • Separable monetized segment: use SOTP or scenario-weighted optionality.
  • Narrative-only technology: no separate valuation premium.

Step 4: Light Cyclical Manufacturing Overlay

Add this overlay when the company is exposed to real estate, consumer cycle, raw material prices, exports, FX, channel inventory, or manufacturing capex cycles.

Additional models / checks:

  • Mid-cycle earnings
  • Normalized margin
  • Stress-case FCF
  • No-growth EPV

Scoring Rubric

The router may use internal 0 / 1 / 2 scoring.

Mature Quality Compounder Score

  • Stable revenue and earnings
  • Strong ROE / ROIC
  • Durable FCF conversion
  • Competitive advantage
  • Mature but positive growth

If score >= 7, classify as Mature Quality Compounder.

Dividend / Shareholder Return Score

  • Stable dividend history
  • Reasonable payout ratio
  • Dividend covered by FCF / FCFE
  • Buyback, share cancellation, or debt reduction
  • Balance sheet supports payout

If score >= 6, add Dividend / Shareholder Return Overlay.

Technology Optionality Score

  • Technology-related segment exists
  • Segment revenue is disclosed
  • Segment growth is above core business
  • Segment has profit or credible margin path
  • Technology can improve ROIC, margin, or TAM

If score >= 6, add Technology Optionality Overlay.

If score <= 3, treat technology as narrative-only and do not assign separate valuation premium.

Cyclicality Score

  • Real estate / consumption cycle exposure
  • Raw material exposure
  • FX / export exposure
  • Inventory or manufacturing cycle exposure

If score >= 4, add Light Cyclical Manufacturing Overlay.

Reverse DCF Execution Gate

When Reverse DCF is selected as the implied-expectation model, the downstream workflow must call skills/reverse-dcf/SKILL.md.

The final report must include either:

  1. A Reverse DCF result, or
  2. A Reverse DCF blocked section.

It is not valid to only mention Reverse DCF as a suggested next step.

User-facing Output Rules

Default output should show:

  • Company classification result
  • Overlay tags
  • Selected valuation model stack
  • Missing data / blocked models
  • Confidence level
  • Investment implication

Default output must not show:

  • Internal score-by-score details
  • Full formulas
  • Spreadsheet-style valuation calculations
  • Discounting schedules
  • Derived metric tables

Calculation detail is allowed only when the user explicitly requests audit, debug, appendix, formulas, or workbook-style detail.

General Rules

  • State assumptions explicitly.
  • Convert material assumptions into structured, evidence-constrained inputs before relying on valuation output.
  • Flag missing or low-confidence data.
  • Prefer conservative assumptions.
  • Keep output auditable.
  • Never use a single valuation multiple as the full valuation conclusion.

v17 Token-Efficient Specialist Route Expansion

The valuation router supports specialist company routes but must lazy-load them.

Default procedure:

  1. Run compact classification.
  2. Select exactly one base route.
  3. Add at most two overlays for L0/L1.
  4. Load only the selected route section from references/valuation_rules/specialized_company_routes_v17.md when needed.
  5. Defer non-material routes and list them briefly.
  6. Do not expose internal scorecards or step-by-step valuation calculations by default.

Specialist routes include AI / semiconductor, digital platform, cloud platform, managed care, insurance-float holding company, financial institutions, SaaS, pharma, commodity cyclicals, REIT/infrastructure, and auto/EV platforms.

v17.1 Modular Workflow Quality Gate

The router now selects workflows as well as valuation models. Default behavior:

  1. Select one primary workflow.
  2. Add at most two auxiliary workflows for L0/L1.
  3. Load references/core/investment_philosophy_layer.md and references/core/investment_quality_gate.md before final output.
  4. Keep domain-specific workflow details out of the main prompt unless selected.
  5. Do not expose internal scorecards, workflow internals, quality-gate internals, or valuation calculation traces unless explicitly requested.

Final output must include business quality, valuation attractiveness, margin of safety, data confidence, and action.

Fixed Output Contract

The router must return routing and model-selection fields to the renderer. It must not write a custom final report.

Default output must include these user-facing valuation/action fields through the fixed renderer:

  • selected valuation model stack
  • Bear / Base / Bull intrinsic value range, or blocked status
  • current price, or blocked status
  • margin of safety, or blocked status
  • price zones
  • position-aware suggestions for empty, half, full, and overweight investors
  • tranche plan
  • add / hold / trim / exit conditions

If valuation range is unavailable, report the valuation and price/action framework as blocked with missing data. Do not drop these sections.