| name | eu-mortgage-rules |
| description | European mortgage feasibility rules used by MortgageCheck — DTI thresholds (35/40), LTV thresholds (80/90, no PMI), employment contract type handling (permanent/fixed_term/self_employed), and +3% rate stress test. Use when designing, implementing, testing, or reviewing the mortgage feasibility logic, or when writing boundary tests against any of these criteria. |
| metadata | {"version":"1.0"} |
European mortgage feasibility rules
This is the canonical rule spec for MortgageCheck. The implementation in lib/calculations.ts follows this document; if the two ever disagree, this skill is the source of truth and the code should be fixed.
The app models European (Italian-flavored) mortgage practice. Two things that are NOT modeled — and should not be invented by an agent consuming this skill — are FICO-style credit scores and US-style Private Mortgage Insurance (PMI). High-LTV risk is reflected in the LTV thresholds themselves; borrower creditworthiness is reflected in the employment contract criterion.
DTI (Debt-to-Income)
Formula: (monthlyDebts + monthlyPayment) / monthlyIncome, where monthlyIncome = annualIncome / 12 and monthlyPayment is the standard amortizing payment for the loan amount, term, and interest rate.
| Ratio | Status |
|---|
| ≤ 35% | pass |
| > 35%, ≤ 40% | warning |
| > 40% | fail |
40% is the European maximum most banks apply. Above 40% the bank is expected to deny.
LTV (Loan-to-Value)
Formula: loanAmount / propertyValue, where loanAmount = propertyValue - downPayment.
| Ratio | Status |
|---|
| ≤ 80% | pass |
| > 80%, ≤ 90% | warning |
| > 90% | fail |
PMI does not apply. Do not generate PMI estimates, do not add PMI line items to results, and do not suggest "down payment to eliminate PMI" — that is US-specific and out of scope. A high-LTV warning means the bank is likely to charge a higher rate or impose stricter conditions, not to require insurance.
Employment contract type
Single enum field on MortgageInput: employmentType: "permanent" | "fixed_term" | "self_employed".
| Value | Status | Rationale |
|---|
permanent | pass | Standard underwriting applies. |
fixed_term | warning | Banks may require a guarantor or additional collateral. |
self_employed | warning | Banks typically require 2–3 years of tax returns and recent financials. |
Any other value is invalid input and the API must reject it with a 400.
Rate stress test (+3 percentage points)
Recompute the monthly payment using interestRate + 3 (same loan amount, same term, same formula). Then compute the stressed DTI: (monthlyDebts + stressedMonthlyPayment) / monthlyIncome.
| Stressed DTI | Status |
|---|
| ≤ 40% | pass |
| > 40%, ≤ 45% | warning |
| > 45% | fail |
The stress delta is fixed at 3 percentage points (constant STRESS_TEST_RATE_DELTA in lib/calculations.ts). This mirrors the kind of affordability buffer European regulators (and the ECB-aligned national supervisors) typically expect.
Boundary values for tests
When writing unit tests, exercise both sides of every threshold. The exact pivot values:
- DTI: 35.0% and 40.0% (do both
<= and > cases)
- LTV: 80.0% and 90.0%
- Stressed DTI: 40.0% and 45.0%
employmentType: one test for each of "permanent", "fixed_term", "self_employed", and one for an invalid value (e.g. "contractor") which the API must reject with a 400.
When asserting amortization output, use a known-good fixture: a 25-year fixed at 4.0% on a €280,000 loan amount has a deterministic monthly payment that you can compute independently and compare against.
Out of scope (do not invent)
These are real European concerns but not modeled by MortgageCheck v1. An agent consuming this skill should not synthesize them, and should not write tests that depend on them:
- Notary and registration fees
- Age-vs-term cap (e.g., loan must end before borrower turns 75)
- Variable vs fixed rate product type
- Energy class / EPC bonuses
- Regional first-home tax incentives
If a feature request would require any of these, it's a scope expansion — flag it in the implementation plan rather than guessing the rule.
