test: expand eval dataset to 56 new cases — 20 happy path, 12 edge, 12 adversarial, 12 multi-step

- Created agent/evals/test_eval_dataset.py with 56 categorized test cases - Happy path (20): portfolio, property CRUD, strategy, wealth position, family, relocation - Edge cases (12): zero values, paid-off property, 1-year strategy, nonexistent IDs, extreme ages - Adversarial (12): SQL injection, negative values, extreme rates, whitespace inputs - Multi-step (12): chained tool calls, stateful CRUD flows, cross-tool data passing - Total suite: 182 tests, 0 failures Made-with: Cursor
1 month ago · 443818bacd
1 changed files with 941 additions and 0 deletions
--- a/agent/evals/test_eval_dataset.py
+++ b/agent/evals/test_eval_dataset.py
@ -0,0 +1,941 @@
+"""
+Comprehensive eval dataset for the Ghostfolio AI Agent.
+
+50+ test cases organized into 4 rubric-required categories:
+  - Happy Path (20+ tests): normal successful user journeys
+  - Edge Cases  (10+ tests): boundary conditions, missing data, zero values
+  - Adversarial (10+ tests): bad inputs, injection attempts, extreme values
+  - Multi-Step  (10+ tests): chained tool calls, stateful flows
+
+Every test documents:
+  TYPE, INPUT (what the user asked), EXPECTED (what should happen),
+  CRITERIA (how pass/fail is determined).
+
+Network calls: all Teleport API calls are mocked via conftest.py autouse
+fixture (mock_teleport_no_network). Tests are deterministic and fast.
+"""
+
+import asyncio
+import os
+import sys
+
+sys.path.insert(0, os.path.join(os.path.dirname(__file__), ".."))
+sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..", "tools"))
+
+os.environ.setdefault("ENABLE_REAL_ESTATE", "true")
+os.environ.setdefault("PROPERTIES_DB_PATH", ":memory:")
+
+import pytest
+
+
+# ---------------------------------------------------------------------------
+# Helper: run coroutine from sync context
+# ---------------------------------------------------------------------------
+
+def _run(coro):
+    loop = asyncio.new_event_loop()
+    try:
+        return loop.run_until_complete(coro)
+    finally:
+        loop.close()
+
+
+def _clear():
+    from property_tracker import property_store_clear
+    property_store_clear()
+
+
+# ============================================================================
+# HAPPY PATH TESTS (20 tests)
+# ============================================================================
+
+# TYPE: happy_path
+# INPUT: add home with purchase price, current value, and mortgage
+# EXPECTED: property created, equity = current_value - mortgage_balance
+# CRITERIA: equity == 175000, success == True
+def test_hp_add_property_basic():
+    _clear()
+    from property_tracker import add_property
+    result = _run(add_property(
+        address="My Primary Home",
+        purchase_price=420000,
+        current_value=490000,
+        mortgage_balance=315000,
+    ))
+    assert result["success"] is True
+    prop = result["result"]["property"]
+    assert prop["equity"] == pytest.approx(175000)
+
+
+# TYPE: happy_path
+# INPUT: add two properties, check combined equity
+# EXPECTED: total equity = sum of both properties' equity
+# CRITERIA: total_equity matches manual calculation
+def test_hp_add_two_properties_combined_equity():
+    _clear()
+    from property_tracker import add_property, get_real_estate_equity
+    _run(add_property("Home A", 400000, 480000, 300000))   # equity=180000
+    _run(add_property("Home B", 300000, 350000, 200000))   # equity=150000
+    result = _run(get_real_estate_equity())
+    assert result["success"] is True
+    assert result["result"]["total_real_estate_equity"] == pytest.approx(330000)
+    assert result["result"]["property_count"] == 2
+
+
+# TYPE: happy_path
+# INPUT: get total net worth with portfolio and property
+# EXPECTED: total = portfolio + real estate equity
+# CRITERIA: total_net_worth == 94000 + 175000 = 269000
+def test_hp_total_net_worth_combined():
+    _clear()
+    from property_tracker import add_property, get_total_net_worth
+    _run(add_property("Test Home", 420000, 490000, 315000))  # equity=175000
+    result = _run(get_total_net_worth(portfolio_value=94000))
+    assert result["success"] is True
+    assert result["result"]["total_net_worth"] == pytest.approx(269000)
+    assert result["result"]["investment_portfolio"] == 94000
+    assert result["result"]["real_estate_equity"] == pytest.approx(175000)
+
+
+# TYPE: happy_path
+# INPUT: strategy simulation with 10-year horizon
+# EXPECTED: final net worth exceeds starting portfolio
+# CRITERIA: total_net_worth > initial_portfolio_value
+def test_hp_strategy_10_year_growth():
+    from realestate_strategy import simulate_real_estate_strategy
+    result = simulate_real_estate_strategy(94000, 120000, 400000, total_years=10)
+    assert result is not None
+    assert "final_picture" in result
+    assert result["final_picture"]["total_net_worth"] > 94000
+
+
+# TYPE: happy_path
+# INPUT: strategy simulation with user-provided 3% appreciation
+# EXPECTED: strategy uses 3% not default 4%
+# CRITERIA: assumptions show 3.0%
+def test_hp_strategy_user_appreciation():
+    from realestate_strategy import simulate_real_estate_strategy
+    result = simulate_real_estate_strategy(
+        94000, 120000, 400000,
+        annual_appreciation=0.03,
+    )
+    assert result["strategy"]["assumptions"]["annual_appreciation"] == "3.0%"
+
+
+# TYPE: happy_path
+# INPUT: analyze wealth position for 34-year-old with $94k portfolio
+# EXPECTED: Fed Reserve comparison with correct median for under_35 bracket
+# CRITERIA: median_for_age == 39000, percentile_estimate present
+def test_hp_wealth_position_age_34():
+    from wealth_visualizer import analyze_wealth_position
+    result = analyze_wealth_position(94000, 34, 120000)
+    assert "current_position" in result
+    assert result["current_position"]["median_for_age"] == 39000
+    assert "percentile_estimate" in result["current_position"]
+    assert "retirement_projection" in result
+
+
+# TYPE: happy_path
+# INPUT: analyze wealth position for 42-year-old
+# EXPECTED: uses 35_to_44 bracket median of $135,000
+# CRITERIA: median_for_age == 135000
+def test_hp_wealth_position_age_42():
+    from wealth_visualizer import analyze_wealth_position
+    result = analyze_wealth_position(200000, 42, 150000)
+    assert result["current_position"]["median_for_age"] == 135000
+
+
+# TYPE: happy_path
+# INPUT: equity options for property with substantial equity
+# EXPECTED: 3 distinct options returned (keep, refi, rental)
+# CRITERIA: len(options) >= 3, each option has projection
+def test_hp_equity_options_three_scenarios():
+    _clear()
+    from property_tracker import add_property, analyze_equity_options
+    prop = _run(add_property("Equity Home", 400000, 520000, 370000))
+    pid = prop["result"]["property"]["id"]
+    result = analyze_equity_options(pid)
+    assert "options" in result
+    assert len(result["options"]) >= 3
+
+
+# TYPE: happy_path
+# INPUT: family planning for Austin with 1 child
+# EXPECTED: childcare costs, monthly surplus, income_impact
+# CRITERIA: income_impact present, monthly_surplus_after is a number
+def test_hp_family_plan_one_child_austin():
+    from family_planner import plan_family_finances
+    result = plan_family_finances("Austin", 120000, num_planned_children=1)
+    assert "income_impact" in result
+    assert "monthly_surplus_after" in result["income_impact"]
+    assert isinstance(result["income_impact"]["monthly_surplus_after"], (int, float))
+
+
+# TYPE: happy_path
+# INPUT: relocation runway calculation Austin → Seattle
+# EXPECTED: verdict returned, months to emergency fund calculated
+# CRITERIA: verdict present, milestones_if_you_move has emergency_fund milestone
+def test_hp_relocation_runway_seattle():
+    from relocation_runway import calculate_relocation_runway
+    result = calculate_relocation_runway(
+        current_salary=120000,
+        offer_salary=180000,
+        current_city="Austin",
+        destination_city="Seattle",
+        portfolio_value=94000,
+    )
+    assert "verdict" in result
+    assert "milestones_if_you_move" in result
+    assert "months_to_6mo_emergency_fund" in result["milestones_if_you_move"]
+
+
+# TYPE: happy_path
+# INPUT: job offer affordability check Austin → SF
+# EXPECTED: COL-adjusted comparison, is_real_raise boolean
+# CRITERIA: is_real_raise present, verdict non-empty
+@pytest.mark.asyncio
+async def test_hp_job_offer_affordability():
+    from wealth_bridge import calculate_job_offer_affordability
+    result = await calculate_job_offer_affordability(
+        offer_salary=180000,
+        offer_city="Seattle",
+        current_salary=120000,
+        current_city="Austin",
+    )
+    assert "is_real_raise" in result
+    assert isinstance(result["is_real_raise"], bool)
+    assert "verdict" in result
+    assert len(result["verdict"]) > 10
+
+
+# TYPE: happy_path
+# INPUT: down payment power for $94k portfolio
+# EXPECTED: at least one Austin-area market affordable at full 20% down
+# CRITERIA: can_afford_full True for at least one market
+def test_hp_down_payment_94k_portfolio():
+    from wealth_bridge import calculate_down_payment_power
+    result = calculate_down_payment_power(94000)
+    assert "markets" in result
+    affordable = [m for m in result["markets"] if m["can_afford_full"]]
+    assert len(affordable) > 0
+
+
+# TYPE: happy_path
+# INPUT: list properties when one exists
+# EXPECTED: property appears in list with correct fields
+# CRITERIA: len(properties) == 1, equity field present
+def test_hp_list_properties_one():
+    _clear()
+    from property_tracker import add_property, get_properties
+    _run(add_property("My Home", 400000, 480000, 320000))
+    result = _run(get_properties())
+    assert result["success"] is True
+    props = result["result"]["properties"]
+    assert len(props) == 1
+    assert "equity" in props[0]
+    assert props[0]["equity"] == pytest.approx(160000)
+
+
+# TYPE: happy_path
+# INPUT: remove property by ID
+# EXPECTED: property no longer in list
+# CRITERIA: property count drops from 1 to 0
+def test_hp_remove_property_success():
+    _clear()
+    from property_tracker import add_property, remove_property, get_properties
+    prop = _run(add_property("Remove Me", 300000, 350000, 200000))
+    pid = prop["result"]["property"]["id"]
+    removed = _run(remove_property(pid))
+    assert removed["success"] is True
+    listed = _run(get_properties())
+    ids = [p["id"] for p in listed["result"]["properties"]]
+    assert pid not in ids
+
+
+# TYPE: happy_path
+# INPUT: update property current value
+# EXPECTED: equity recalculates correctly
+# CRITERIA: equity increases after value update
+def test_hp_update_property_value():
+    _clear()
+    from property_tracker import add_property, update_property
+    prop = _run(add_property("Update Test", 400000, 450000, 320000))
+    pid = prop["result"]["property"]["id"]
+    updated = _run(update_property(pid, current_value=470000))
+    assert updated["success"] is True
+    new_equity = updated["result"]["property"]["equity"]
+    assert new_equity == pytest.approx(150000)  # 470000 - 320000
+
+
+# TYPE: happy_path
+# INPUT: strategy simulation single property scenario
+# EXPECTED: result contains at least 1 property
+# CRITERIA: properties_owned >= 1
+def test_hp_strategy_single_property():
+    from realestate_strategy import simulate_real_estate_strategy
+    result = simulate_real_estate_strategy(
+        94000, 120000, 400000,
+        buy_interval_years=10,  # only 1 purchase in 10 years
+        total_years=10,
+    )
+    assert result["final_picture"]["num_properties_owned"] >= 1
+
+
+# TYPE: happy_path
+# INPUT: family plan with partner income
+# EXPECTED: household income reflects both incomes
+# CRITERIA: total_household_income > single income
+def test_hp_family_plan_with_partner():
+    from family_planner import plan_family_finances
+    result = plan_family_finances("Austin", 120000, partner_income=80000, num_planned_children=1)
+    assert "income_impact" in result
+    data = result["income_impact"]
+    assert "total_household_income" in data or "monthly_surplus_after" in data
+
+
+# TYPE: happy_path
+# INPUT: wealth position with real estate equity added
+# EXPECTED: total net worth includes real estate
+# CRITERIA: total_net_worth > portfolio_value alone
+def test_hp_wealth_position_with_real_estate():
+    from wealth_visualizer import analyze_wealth_position
+    result = analyze_wealth_position(
+        portfolio_value=94000,
+        age=34,
+        annual_income=120000,
+        real_estate_equity=175000,
+    )
+    assert "current_position" in result
+    pos = result["current_position"]
+    assert pos["total_net_worth"] == pytest.approx(269000)
+
+
+# TYPE: happy_path
+# INPUT: relocation to cheaper city (SF → Austin)
+# EXPECTED: destination surplus higher than current, verdict positive
+# CRITERIA: destination monthly_surplus > 0, verdict present
+def test_hp_relocation_to_cheaper_city():
+    from relocation_runway import calculate_relocation_runway
+    result = calculate_relocation_runway(
+        current_salary=150000,
+        offer_salary=140000,
+        current_city="San Francisco",
+        destination_city="Austin",
+        portfolio_value=100000,
+    )
+    assert "verdict" in result
+    dest_surplus = result["destination_monthly"]["monthly_surplus"]
+    curr_surplus = result["current_monthly"]["monthly_surplus"]
+    assert dest_surplus > curr_surplus
+
+
+# TYPE: happy_path
+# INPUT: strategy with conservative appreciation preset
+# EXPECTED: conservative < optimistic final net worth
+# CRITERIA: conservative_net_worth < optimistic_net_worth
+def test_hp_strategy_conservative_vs_optimistic():
+    from realestate_strategy import simulate_real_estate_strategy
+    conservative = simulate_real_estate_strategy(
+        94000, 120000, 400000, annual_appreciation=0.02,
+    )
+    optimistic = simulate_real_estate_strategy(
+        94000, 120000, 400000, annual_appreciation=0.06,
+    )
+    assert (conservative["final_picture"]["total_net_worth"] <
+            optimistic["final_picture"]["total_net_worth"])
+
+
+# ============================================================================
+# EDGE CASE TESTS (12 tests)
+# ============================================================================
+
+# TYPE: edge_case
+# INPUT: portfolio value of zero
+# EXPECTED: graceful response, not a crash
+# CRITERIA: returns dict, no exception
+def test_ec_zero_portfolio_value():
+    from wealth_visualizer import analyze_wealth_position
+    result = analyze_wealth_position(0, 30, 50000)
+    assert result is not None
+    assert isinstance(result, dict)
+    assert "current_position" in result
+
+
+# TYPE: edge_case
+# INPUT: property with no mortgage (fully paid off)
+# EXPECTED: equity equals current value
+# CRITERIA: equity == current_value, equity_pct == 100.0
+def test_ec_paid_off_property():
+    _clear()
+    from property_tracker import add_property
+    result = _run(add_property(
+        address="Paid Off Home",
+        purchase_price=300000,
+        current_value=380000,
+        mortgage_balance=0,
+    ))
+    prop = result["result"]["property"]
+    assert prop["equity"] == pytest.approx(380000)
+    assert prop["equity_pct"] == pytest.approx(100.0)
+
+
+# TYPE: edge_case
+# INPUT: strategy simulation for just 1 year
+# EXPECTED: returns valid result, no crash
+# CRITERIA: result has final_picture and timeline, no exception
+def test_ec_strategy_single_year():
+    from realestate_strategy import simulate_real_estate_strategy
+    result = simulate_real_estate_strategy(50000, 80000, 300000, total_years=1)
+    assert result is not None
+    assert "final_picture" in result
+    assert "timeline" in result
+    assert isinstance(result["final_picture"]["num_properties_owned"], int)
+
+
+# TYPE: edge_case
+# INPUT: equity options on nonexistent property ID
+# EXPECTED: error dict returned, not exception
+# CRITERIA: "error" key present in result
+def test_ec_equity_nonexistent_property():
+    from property_tracker import analyze_equity_options
+    result = analyze_equity_options("does-not-exist-999")
+    assert result is not None
+    assert "error" in result
+
+
+# TYPE: edge_case
+# INPUT: family planner with zero income
+# EXPECTED: graceful response, no ZeroDivisionError
+# CRITERIA: returns dict, no exception raised
+def test_ec_family_planner_zero_income():
+    from family_planner import plan_family_finances
+    result = plan_family_finances("Austin", 0, num_planned_children=1)
+    assert result is not None
+    assert isinstance(result, dict)
+
+
+# TYPE: edge_case
+# INPUT: wealth position for someone at exactly the median
+# EXPECTED: correctly identified as median tier
+# CRITERIA: percentile_estimate contains "50th" or similar
+def test_ec_wealth_exactly_at_median():
+    from wealth_visualizer import analyze_wealth_position
+    # median for under_35 is 39000
+    result = analyze_wealth_position(39000, 30, 60000)
+    pos = result["current_position"]
+    assert "50th" in pos["percentile_estimate"] or "median" in pos["percentile_estimate"].lower()
+
+
+# TYPE: edge_case
+# INPUT: empty property list — get net worth with no properties
+# EXPECTED: net worth equals portfolio value, real estate equity is 0
+# CRITERIA: total_net_worth == portfolio_value, real_estate_equity == 0
+def test_ec_net_worth_no_properties():
+    _clear()
+    from property_tracker import get_total_net_worth
+    result = _run(get_total_net_worth(portfolio_value=50000))
+    assert result["success"] is True
+    assert result["result"]["total_net_worth"] == pytest.approx(50000)
+    assert result["result"]["real_estate_equity"] == 0
+
+
+# TYPE: edge_case
+# INPUT: relocation with identical source and destination city
+# EXPECTED: no crash, verdict makes sense
+# CRITERIA: returns dict with verdict field
+def test_ec_same_city_relocation():
+    from relocation_runway import calculate_relocation_runway
+    result = calculate_relocation_runway(
+        current_salary=120000,
+        offer_salary=125000,
+        current_city="Austin",
+        destination_city="Austin",
+        portfolio_value=94000,
+    )
+    assert result is not None
+    assert "verdict" in result
+
+
+# TYPE: edge_case
+# INPUT: strategy with very large portfolio (1M)
+# EXPECTED: simulation completes without overflow
+# CRITERIA: total_net_worth is a positive finite number
+def test_ec_strategy_large_portfolio():
+    from realestate_strategy import simulate_real_estate_strategy
+    result = simulate_real_estate_strategy(
+        initial_portfolio_value=1_000_000,
+        annual_income=300000,
+        first_home_price=1_500_000,
+        total_years=10,
+    )
+    assert result is not None
+    nw = result["final_picture"]["total_net_worth"]
+    assert nw > 0
+    assert nw < 1e15  # sanity check: not overflowing
+
+
+# TYPE: edge_case
+# INPUT: analyze property with equity exceeding current value (impossible)
+# EXPECTED: graceful handling, equity capped or error message
+# CRITERIA: no crash, returns dict
+def test_ec_mortgage_exceeds_value():
+    _clear()
+    from property_tracker import add_property
+    result = _run(add_property(
+        address="Underwater Property",
+        purchase_price=400000,
+        current_value=300000,
+        mortgage_balance=380000,
+    ))
+    # Should succeed but equity will be negative (underwater)
+    assert result["success"] is True
+    prop = result["result"]["property"]
+    equity = prop["equity"]
+    # underwater: equity = 300000 - 380000 = -80000 (valid, just negative)
+    assert isinstance(equity, (int, float))
+
+
+# TYPE: edge_case
+# INPUT: family plan for 5 children
+# EXPECTED: costs scaled proportionally, no crash
+# CRITERIA: childcare costs for 5 > childcare for 1
+def test_ec_family_many_children():
+    from family_planner import plan_family_finances
+    result1 = plan_family_finances("Austin", 200000, num_planned_children=1)
+    result5 = plan_family_finances("Austin", 200000, num_planned_children=5)
+    assert result1 is not None and result5 is not None
+    # Costs for 5 children should be higher
+    s1 = result1["income_impact"]["monthly_surplus_after"]
+    s5 = result5["income_impact"]["monthly_surplus_after"]
+    assert s5 < s1  # more children = lower surplus
+
+
+# TYPE: edge_case
+# INPUT: wealth position for very old age (80+)
+# EXPECTED: uses highest bracket (65+), no crash
+# CRITERIA: median_for_age == 409000 (65_to_74 bracket)
+def test_ec_wealth_very_old_age():
+    from wealth_visualizer import analyze_wealth_position
+    result = analyze_wealth_position(500000, 82, 60000)
+    assert result is not None
+    assert result["current_position"]["median_for_age"] == 409000
+
+
+# ============================================================================
+# ADVERSARIAL TESTS (12 tests)
+# ============================================================================
+
+# TYPE: adversarial
+# INPUT: property address contains SQL injection
+# EXPECTED: stored safely as a string, DB still works after
+# CRITERIA: no exception, subsequent queries work normally
+def test_adv_sql_injection_address():
+    _clear()
+    from property_tracker import add_property, get_properties
+    malicious = "'; DROP TABLE properties; --"
+    result = _run(add_property(
+        address=malicious,
+        purchase_price=300000,
+        current_value=350000,
+        mortgage_balance=200000,
+    ))
+    assert result["success"] is True
+    # DB still works after the injection attempt
+    props = _run(get_properties())
+    assert props["success"] is True
+
+
+# TYPE: adversarial
+# INPUT: negative property values
+# EXPECTED: validation catches it OR handles gracefully
+# CRITERIA: no uncaught exception
+def test_adv_negative_property_value():
+    _clear()
+    from property_tracker import add_property
+    try:
+        result = _run(add_property(
+            address="Bad Property",
+            purchase_price=-100000,
+            current_value=-50000,
+            mortgage_balance=0,
+        ))
+        # If it accepts, result should be a dict
+        assert result is not None
+    except (ValueError, AssertionError):
+        pass  # Rejecting invalid input is correct behavior
+
+
+# TYPE: adversarial
+# INPUT: impossible appreciation rate (1000% = 10.0)
+# EXPECTED: simulation runs without crash
+# CRITERIA: returns result dict with final_picture
+def test_adv_extreme_appreciation_rate():
+    from realestate_strategy import simulate_real_estate_strategy
+    result = simulate_real_estate_strategy(
+        94000, 120000, 400000,
+        annual_appreciation=10.0,  # 1000% — extreme input
+    )
+    assert result is not None
+    assert "final_picture" in result
+
+
+# TYPE: adversarial
+# INPUT: strategy with zero annual income
+# EXPECTED: graceful, not ZeroDivision
+# CRITERIA: returns dict, no crash
+def test_adv_strategy_zero_income():
+    from realestate_strategy import simulate_real_estate_strategy
+    try:
+        result = simulate_real_estate_strategy(
+            initial_portfolio_value=100000,
+            annual_income=0,
+            first_home_price=300000,
+            total_years=5,
+        )
+        assert result is not None
+        assert isinstance(result, dict)
+    except ZeroDivisionError:
+        pytest.fail("ZeroDivisionError raised with zero income")
+
+
+# TYPE: adversarial
+# INPUT: mortgage rate of 100% (extreme)
+# EXPECTED: simulation completes, not crash
+# CRITERIA: result is a dict with final_picture
+def test_adv_extreme_mortgage_rate():
+    from realestate_strategy import simulate_real_estate_strategy
+    result = simulate_real_estate_strategy(
+        94000, 120000, 400000,
+        mortgage_rate=1.00,  # 100% rate
+        total_years=5,
+    )
+    assert result is not None
+    assert "final_picture" in result
+
+
+# TYPE: adversarial
+# INPUT: address that is only whitespace
+# EXPECTED: validation returns structured error
+# CRITERIA: success=False, error dict present
+def test_adv_whitespace_address():
+    _clear()
+    from property_tracker import add_property
+    result = _run(add_property(address="   \t\n  ", purchase_price=300000))
+    assert result["success"] is False
+    assert "error" in result
+
+
+# TYPE: adversarial
+# INPUT: purchase price is zero
+# EXPECTED: validation error returned
+# CRITERIA: success=False
+def test_adv_zero_purchase_price():
+    _clear()
+    from property_tracker import add_property
+    result = _run(add_property(address="Test", purchase_price=0))
+    assert result["success"] is False
+
+
+# TYPE: adversarial
+# INPUT: wealth analysis with negative portfolio
+# EXPECTED: graceful handling, no crash
+# CRITERIA: returns dict
+def test_adv_negative_portfolio_wealth():
+    from wealth_visualizer import analyze_wealth_position
+    try:
+        result = analyze_wealth_position(-5000, 30, 60000)
+        assert result is not None
+        assert isinstance(result, dict)
+    except Exception:
+        pass  # Rejecting negative portfolio is acceptable
+
+
+# TYPE: adversarial
+# INPUT: family plan for a city not in the lookup
+# EXPECTED: uses default costs, no crash
+# CRITERIA: returns dict with income_impact
+def test_adv_unknown_city_family_plan():
+    from family_planner import plan_family_finances
+    result = plan_family_finances("Xanadu City", 80000, num_planned_children=1)
+    assert result is not None
+    assert "income_impact" in result
+
+
+# TYPE: adversarial
+# INPUT: remove a nonexistent property ID
+# EXPECTED: structured error dict returned
+# CRITERIA: success=False, error code = NOT_FOUND
+def test_adv_remove_nonexistent_property():
+    _clear()
+    from property_tracker import remove_property
+    result = _run(remove_property("nonexistent-id-999"))
+    assert result["success"] is False
+    assert isinstance(result["error"], dict)
+    assert result["error"]["code"] == "PROPERTY_TRACKER_NOT_FOUND"
+
+
+# TYPE: adversarial
+# INPUT: strategy with negative down payment percentage
+# EXPECTED: uses fallback/default or handles gracefully
+# CRITERIA: no crash
+def test_adv_negative_down_payment_pct():
+    from realestate_strategy import simulate_real_estate_strategy
+    try:
+        result = simulate_real_estate_strategy(
+            94000, 120000, 400000,
+            down_payment_pct=-0.5,
+        )
+        assert result is not None
+    except (ValueError, AssertionError):
+        pass  # Validation rejection is fine
+
+
+# TYPE: adversarial
+# INPUT: empty string city for relocation
+# EXPECTED: no crash, returns dict
+# CRITERIA: result is a dict
+def test_adv_empty_city_relocation():
+    from relocation_runway import calculate_relocation_runway
+    try:
+        result = calculate_relocation_runway(
+            current_salary=100000,
+            offer_salary=120000,
+            current_city="",
+            destination_city="",
+            portfolio_value=50000,
+        )
+        assert isinstance(result, dict)
+    except Exception:
+        pass  # Failing gracefully is acceptable
+
+
+# ============================================================================
+# MULTI-STEP TESTS (12 tests)
+# ============================================================================
+
+# TYPE: multi_step
+# INPUT: add property → get total net worth → verify property appears
+# EXPECTED: net worth increases by property equity after adding
+# CRITERIA: total_net_worth > portfolio_value_alone
+def test_ms_add_then_net_worth():
+    _clear()
+    from property_tracker import add_property, get_total_net_worth
+    _run(add_property(
+        address="Multi Step Test",
+        purchase_price=350000,
+        current_value=420000,
+        mortgage_balance=280000,
+    ))  # equity = 140000
+    result = _run(get_total_net_worth(portfolio_value=94000))
+    assert result["result"]["total_net_worth"] == pytest.approx(234000)
+    assert result["result"]["real_estate_equity"] == pytest.approx(140000)
+
+
+# TYPE: multi_step
+# INPUT: add property → analyze equity options
+# EXPECTED: equity options reference added property, 3 scenarios returned
+# CRITERIA: options has 3 entries, each has 10-year projection
+def test_ms_add_then_equity_options():
+    _clear()
+    from property_tracker import add_property, analyze_equity_options
+    prop = _run(add_property(
+        address="Equity Chain Test",
+        purchase_price=400000,
+        current_value=520000,
+        mortgage_balance=370000,
+    ))
+    result = analyze_equity_options(prop["result"]["property"]["id"])
+    assert "options" in result
+    assert len(result["options"]) >= 3
+
+
+# TYPE: multi_step
+# INPUT: add two properties → remove one → verify only one remains
+# EXPECTED: after removal, list shows exactly 1 property
+# CRITERIA: len(properties) == 1 after removal
+def test_ms_add_two_remove_one():
+    _clear()
+    from property_tracker import add_property, remove_property, get_properties
+    p1 = _run(add_property("Home 1", 400000, 450000, 300000))
+    p2 = _run(add_property("Home 2", 350000, 400000, 250000))
+    id1 = p1["result"]["property"]["id"]
+    _run(remove_property(id1))
+    listed = _run(get_properties())
+    props = listed["result"]["properties"]
+    assert len(props) == 1
+    assert props[0]["id"] == p2["result"]["property"]["id"]
+
+
+# TYPE: multi_step
+# INPUT: strategy simulation → use final net worth in wealth position check
+# EXPECTED: chaining results without errors
+# CRITERIA: both return valid dicts, wealth check uses strategy output
+def test_ms_strategy_then_wealth_check():
+    from realestate_strategy import simulate_real_estate_strategy
+    from wealth_visualizer import analyze_wealth_position
+    strategy = simulate_real_estate_strategy(94000, 120000, 400000, total_years=10)
+    final_worth = strategy["final_picture"]["total_net_worth"]
+    wealth = analyze_wealth_position(
+        portfolio_value=final_worth,
+        age=44,
+        annual_income=150000,
+    )
+    assert "current_position" in wealth
+    assert "retirement_projection" in wealth
+
+
+# TYPE: multi_step
+# INPUT: family planning → use reduced savings in wealth position
+# EXPECTED: lower savings rate flows into retirement projection
+# CRITERIA: both tools return valid dicts
+def test_ms_family_then_wealth():
+    from family_planner import plan_family_finances
+    from wealth_visualizer import analyze_wealth_position
+    family = plan_family_finances("Austin", 120000, num_planned_children=1)
+    assert "income_impact" in family
+    reduced_savings = max(0, family["income_impact"]["monthly_surplus_after"] * 12)
+    wealth = analyze_wealth_position(
+        portfolio_value=94000,
+        age=32,
+        annual_income=120000,
+        annual_savings=reduced_savings,
+    )
+    assert "retirement_projection" in wealth
+
+
+# TYPE: multi_step
+# INPUT: job offer check → relocation runway → both use same cities
+# EXPECTED: consistent data across both tools for Austin/Seattle
+# CRITERIA: both return their respective fields correctly
+@pytest.mark.asyncio
+async def test_ms_job_offer_then_runway():
+    from wealth_bridge import calculate_job_offer_affordability
+    from relocation_runway import calculate_relocation_runway
+    offer = await calculate_job_offer_affordability(
+        offer_salary=180000,
+        offer_city="Seattle",
+        current_salary=120000,
+        current_city="Austin",
+    )
+    runway = calculate_relocation_runway(
+        current_salary=120000,
+        offer_salary=180000,
+        current_city="Austin",
+        destination_city="Seattle",
+        portfolio_value=94000,
+    )
+    assert "is_real_raise" in offer
+    assert "verdict" in runway
+
+
+# TYPE: multi_step
+# INPUT: add property → update value → check net worth reflects update
+# EXPECTED: net worth uses updated value not original
+# CRITERIA: net worth after update > net worth after initial add
+def test_ms_add_update_then_net_worth():
+    _clear()
+    from property_tracker import add_property, update_property, get_total_net_worth
+    prop = _run(add_property("Growing Home", 400000, 450000, 320000))
+    pid = prop["result"]["property"]["id"]
+    nw_before = _run(get_total_net_worth(portfolio_value=50000))
+    _run(update_property(pid, current_value=500000))
+    nw_after = _run(get_total_net_worth(portfolio_value=50000))
+    assert nw_after["result"]["total_net_worth"] > nw_before["result"]["total_net_worth"]
+
+
+# TYPE: multi_step
+# INPUT: add property → get equity → use in wealth position
+# EXPECTED: wealth position uses real estate equity from property tracker
+# CRITERIA: position with RE equity > position without
+def test_ms_property_equity_in_wealth_position():
+    _clear()
+    from property_tracker import add_property, get_real_estate_equity
+    from wealth_visualizer import analyze_wealth_position
+    _run(add_property("Wealth Test", 400000, 500000, 300000))  # equity=200000
+    equity_result = _run(get_real_estate_equity())
+    equity = equity_result["result"]["total_real_estate_equity"]
+    pos_with_re = analyze_wealth_position(94000, 34, 120000, real_estate_equity=equity)
+    pos_without = analyze_wealth_position(94000, 34, 120000)
+    assert pos_with_re["current_position"]["total_net_worth"] > pos_without["current_position"]["total_net_worth"]
+
+
+# TYPE: multi_step
+# INPUT: full CRUD cycle — create, read, update, delete
+# EXPECTED: each operation succeeds, state consistent throughout
+# CRITERIA: all 4 operations return success=True, final list is empty
+def test_ms_full_crud_cycle():
+    _clear()
+    from property_tracker import add_property, get_properties, update_property, remove_property
+
+    # CREATE
+    prop = _run(add_property("CRUD Home", 400000, 450000, 320000))
+    assert prop["success"] is True
+    pid = prop["result"]["property"]["id"]
+
+    # READ
+    listed = _run(get_properties())
+    ids = [p["id"] for p in listed["result"]["properties"]]
+    assert pid in ids
+
+    # UPDATE
+    updated = _run(update_property(pid, current_value=480000))
+    assert updated["success"] is True
+    assert updated["result"]["property"]["equity"] == pytest.approx(160000)
+
+    # DELETE
+    removed = _run(remove_property(pid))
+    assert removed["success"] is True
+
+    # Verify empty
+    after = _run(get_properties())
+    assert after["result"]["properties"] == []
+
+
+# TYPE: multi_step
+# INPUT: multiple properties → wealth position uses combined equity
+# EXPECTED: total equity from all properties flows into net worth
+# CRITERIA: net worth = portfolio + combined equity
+def test_ms_multiple_properties_net_worth():
+    _clear()
+    from property_tracker import add_property, get_total_net_worth
+    _run(add_property("Home A", 400000, 480000, 300000))  # equity=180000
+    _run(add_property("Home B", 300000, 380000, 260000))  # equity=120000
+    result = _run(get_total_net_worth(portfolio_value=94000))
+    expected = 94000 + 180000 + 120000  # = 394000
+    assert result["result"]["total_net_worth"] == pytest.approx(expected)
+
+
+# TYPE: multi_step
+# INPUT: relocation runway → family plan in destination city
+# EXPECTED: can chain city data across both tools
+# CRITERIA: both return valid dicts for Seattle
+def test_ms_runway_then_family_plan():
+    from relocation_runway import calculate_relocation_runway
+    from family_planner import plan_family_finances
+    runway = calculate_relocation_runway(
+        current_salary=120000,
+        offer_salary=180000,
+        current_city="Austin",
+        destination_city="Seattle",
+        portfolio_value=94000,
+    )
+    assert "verdict" in runway
+    family = plan_family_finances("Seattle", 180000, num_planned_children=1)
+    assert "income_impact" in family
+
+
+# TYPE: multi_step
+# INPUT: strategy simulation over 20 years → verify timeline length
+# EXPECTED: timeline has 20+ entries covering each year
+# CRITERIA: len(timeline) >= 20
+def test_ms_strategy_long_horizon():
+    from realestate_strategy import simulate_real_estate_strategy
+    result = simulate_real_estate_strategy(
+        94000, 120000, 400000,
+        total_years=20,
+        buy_interval_years=4,
+    )
+    assert result is not None
+    timeline = result.get("timeline", [])
+    assert len(timeline) >= 20
+    assert result["final_picture"]["num_properties_owned"] >= 1