Property-Report-Skill-Judges

Comparison Advisor (subagent)

Part of the Property Investigator skill. Runs independently (“which house should I buy”) or is invoked by the orchestrator when a buyer is weighing multiple properties. Produces one clear best-buy pick with an explicit switch condition.

Trigger phrases (independent invocation)

Shared state contract (multi-property — no single journey.json)

Inputs & outputs (quick reference)

| Input | Where to add it | Type | Importance | |——-|—————–|——|————| | Two or more property folders | each under properties/<Address>/ | existing report/README per property | Critical — each needs at least a snapshot to score | | Your priorities / assumptions | tell the agent | — | Important — re-weights the scoring |

Output What it tells you
properties/<A>-vs-<B>-Best-Buy-YYYY-MM-DD.html (parent properties/ folder) Side-by-side scorecard, one best-buy pick, and the explicit condition to switch

Workflow

  1. Inputs: two or more property folders under properties/. If a property lacks a report, first build at least its README/snapshot. Honor any user assumptions (e.g., “assume Property B at $610–630K”) and state them.
  2. Normalize: pull each property’s key facts from its own folder artifacts (never re-derive from another property’s data). Build the side-by-side table.

  3. Score — the “Best-Buy Top-10 Scorecard” (standard method). Rate each property 0–10 on the 10 weighted factors below, multiply by weight, sum, and normalize to 100. Default weights are resale-tilted; re-weight to the user’s stated emphasis and say so. Missing data is flagged, never guessed.

    # Factor Weight What it measures
    1 Price vs market / instant equity 5 Real discount vs CAD value and validated closed comps — not a market-rejected estimate
    2 Resale demand (size + school ratings) 5 Livable size, bed count, ISD/campus ratings, pocket price ceiling
    3 $/sqft value 4 Price normalized to size vs street/subdivision
    4 Condition, age & inspection exposure 4 Roof/HVAC/foundation/systems age, disclosed defects, likely repair $
    5 Location & lot quality 4 Lot size, corner/cul-de-sac/greenbelt vs busy road/backing, flood zone, community prestige
    6 Marketability track record 3 DOM, relists, price cuts, prior failed sales (liquidity)
    7 Entry facing / Vastu + community desirability 3 Facing demand in the local buyer pool (real liquidity driver in some markets)
    8 Builder & build-quality reputation 3 The 🟢/🟡/🔴 builder rating from cited reviews (confirmed builder only; neutral if unconfirmed)
    9 Carrying cost & financials 2 Taxes, HOA, insurance/roof insurability, special assessments
    10 Legal / title / HOA / flood risk 2 Easements, encroachments, deed restrictions, HOA litigation, flood exposure

    Max weighted points = 350 (Σ weights = 35 × 10). Normalized score = weighted total ÷ 350 × 100. Tie-breakers, in order: (1) real comp-validated equity, (2) inspection/structural/insurability safety, (3) resale liquidity. A low price never cures a safety/structural/insurability defect. Always show the rating-scale note for factor 8 and keep Vastu framed as a belief-based but real demand driver.

  4. Decide: always output one clear pick + the explicit condition under which to switch to the runner-up (e.g., “pivot if the foundation report fails”). Avoid wishy-washy conclusions. Near-ties must still name a pick and say it’s close.
  5. Output location: save the comparison in the parent properties/ folder (outside every individual property folder), named with both property names: properties/<A>-vs-<B>-Best-Buy-YYYY-MM-DD.html, rendered as self-contained HTML (side-by-side + Top-10 scorecard in <table>s). Never write the comparison inside a single property’s folder.
  6. Honesty rules: distinguish a real below-market discount (comp-validated) from an unreliable estimate (e.g., a home that repeatedly failed to sell at that price). Flag marketability red flags. Keep Vastu framed as a belief system that is nonetheless a real demand driver in some markets.

Guardrails