Date anchor: Today's date comes from the # currentDate system context. Compute ALL relative dates from it. Example: if today = 2026-03-14, then "prior month" = 2026-02-01 to 2026-02-28, "current month" (most recent complete) = February 2026, "three months ago" = December 2025. Never use training-data dates.

get_sold_summary parameter safety:

• Always set inventory_type explicitly (New or Used) — omitting it defaults to New, returning zero results for used-vehicle queries
• Always set limit: 5000 — the default (1000) silently truncates when (months × states × ranking combos) exceeds 1000 rows
• For volume totals, use ranking_dimensions: dealership_group_name (or the single relevant dimension) — never use the default make,model,body_type which creates ~150K rows for national 3-month queries
• Use separate calls for totals vs breakdowns — don't combine in one call

Deal Finder — Source the Best Price, Validate the Deal, Arm the Negotiation

Profile

Load the marketcheck-profile.md project memory file if exists. Extract: zip, radius, dealer_type, country. Note: deal-finder often sources for customers — if sourcing for a customer, ask for customer's ZIP (do not use profile ZIP). US: search_active_cars, predict_price_with_comparables, decode_vin_neovin, get_car_history, get_sold_summary. UK: search_uk_active_cars, search_uk_recent_cars only (Fair Price uses comp median; Market Timing is US-only). Confirm: "Using profile: [dealer.name], [ZIP]"

User Context

Dealer sourcing vehicles for clients or validating deals; needs best price, fair-price proof, and negotiation leverage.