Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Simulate the car-buying gauntlet before you walk in — the four-square worksheet, the payment-question trap, the trade-in shuffle, and the finance office's second sales floor, all run against your actual deal. Use when asked practice negotiating at a dealership, simulate the finance office, what tricks will the dealer use, or prep me before I buy a car. Produces the showroom and finance-office transcripts with the salesperson's playbook notes, the deal outcome vs your targets, and a debrief with
.claude/skills/mohitagw15856-the-car-dealership/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-02 | ✗→✓ | ▲ Improved | -8% | 0% |
| case-11 | ✗→✓ | ▲ Improved | 32% | 0% |
| case-13 | ✗→✓ | ▲ Improved | 179% | 0% |
| case-16 | ✗→✓ | ▲ Improved | 58% | 0% |
| case-20 | ✗→✓ | ▲ Improved | 89% | 0% |
The dealership is two negotiations wearing one showroom: the car price out front, and the finance office in back — where the real margin often lives, sold to a buyer whose guard dropped the moment they heard "congratulations." This skill runs both rooms against your actual numbers: the four-square worksheet that blends price/trade/payment into fog, the "what payment works for you?" trap, the trade-in shuffle, and then the F&I gauntlet of add-ons and rate markup. Every move gets named in the dealer's private notes, and the debrief hands you the exact holds.
Ask for these if not provided:
Room one — the showroom moves: the payment question ("what monthly works for you?" — answering it surrenders the price; the counter is "let's settle the out-the-door price first") · the four-square (price, trade, down, payment on one sheet — fog by design; the counter is one number at a time, out-the-door) · the trade-in shuffle (great car price, terrible trade number, or vice versa — the counter is negotiating them as separate transactions) · the manager walk ("let me talk to my manager" — theater and a pacing device) · the today-only urgency (the deal that dies at midnight rarely does).
Room two — the finance office: the rate markup (the quoted APR may carry dealer reserve — the counter is your pre-approval on the desk: "beat this or use it") · the add-on cascade (VIN etching, paint protection, nitrogen, extended warranty, GAP — each priced at 3–10× cost; some, like GAP on a long loan, are legitimately worth pricing elsewhere) · the payment-packing move (add-ons quoted as "just $9 more a month" — ×72 months, out loud, every time) · the menu close (three columns of bundles, none labeled "none of these" — that column exists; ask for it).
Mechanics to honor: out-the-door price is the only number that can't be shuffled · everything is negotiable until signed, nothing after · the walk-out is a legitimate and often winning move, and one run should show it working · four hours in the building is a tactic, not bad luck.
> Simulation — a plausible adversarial reading, not a prediction.
Transcript. Dealer's note: after each move, naming the play and reading you: "asked their payment — they answered. Working payment now, price is loose."]
Transcript: rate, add-ons, menu — with the ×72 math shown wherever "per month" appears]
Deal vs. targets · the dealer's notes on where you left money · the walk-out branch: what re-engagement looked like]
| Moment you softened | The move | The hold (exact words) | |---|---|---| Plus: the crib sheet — out-the-door only · trade separate · pre-approval on the desk · ×months every add-on · the none-column ask]
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 53,560 | 45,163 | -16% | 1 | 1 | 0% | 8,348 | 7,239 | -13% | 0 | 0 | — |
case-07 | pass→pass | 28,470 | 25,566 | -10% | 1 | 1 | 0% | 2,918 | 5,552 | +90% | 0 | 0 | — |
case-02 | fail→pass | 46,989 | 46,311 | -1% | 1 | 1 | 0% | 8,246 | 7,601 | -8% | 0 | 0 | — |
case-03 | fail→fail | 45,367 | 46,867 | +3% | 1 | 1 | 0% | 6,819 | 7,976 | +17% | 0 | 0 | — |
case-04 | fail→fail | 12,271 | 54,141 | +341% | 1 | 1 | 0% | 1,074 | 8,984 | +736% | 0 | 0 | — |
case-05 | pass→pass | 28,122 | 39,412 | +40% | 1 | 1 | 0% | 3,046 | 7,525 | +147% | 0 | 0 | — |
case-06 | pass→pass | 23,831 | 53,536 | +125% | 1 | 1 | 0% | 2,599 | 7,735 | +198% | 0 | 0 | — |
case-08 | pass→pass | 22,795 | 33,892 | +49% | 1 | 1 | 0% | 2,857 | 4,880 | +71% | 0 | 0 | — |
case-09 | pass→pass | 18,722 | 16,015 | -14% | 1 | 1 | 0% | 1,825 | 2,618 | +43% | 0 | 0 | — |
case-10 | pass→pass | 27,500 | 40,147 | +46% | 1 | 1 | 0% | 2,832 | 7,148 | +152% | 0 | 0 | — |
case-11 | fail→pass | 23,806 | 16,409 | -31% | 1 | 1 | 0% | 2,513 | 3,318 | +32% | 0 | 0 | — |
case-12 | fail→fail | 25,055 | 22,280 | -11% | 1 | 1 | 0% | 2,481 | 3,952 | +59% | 0 | 0 | — |
case-13 | fail→pass | 25,828 | 51,727 | +100% | 1 | 1 | 0% | 2,614 | 7,281 | +179% | 0 | 0 | — |
case-14 | fail→fail | 13,614 | 21,995 | +62% | 1 | 1 | 0% | 1,216 | 3,405 | +180% | 0 | 0 | — |
case-15 | fail→fail | 21,123 | 39,261 | +86% | 1 | 1 | 0% | 2,315 | 6,336 | +174% | 0 | 0 | — |
case-16 | fail→pass | 18,832 | 21,251 | +13% | 1 | 1 | 0% | 2,172 | 3,436 | +58% | 0 | 0 | — |
case-17 | fail→fail | 23,884 | 46,885 | +96% | 1 | 1 | 0% | 2,844 | 8,510 | +199% | 0 | 0 | — |
case-18 | pass→pass | 18,257 | 28,957 | +59% | 1 | 1 | 0% | 2,757 | 4,580 | +66% | 0 | 0 | — |
case-19 | pass→pass | 29,428 | 33,893 | +15% | 1 | 1 | 0% | 3,148 | 4,924 | +56% | 0 | 0 | — |
case-20 | fail→pass | 21,277 | 29,710 | +40% | 1 | 1 | 0% | 2,675 | 5,065 | +89% | 0 | 0 | — |
case-21 | fail→fail | 24,291 | 42,074 | +73% | 1 | 1 | 0% | 2,480 | 7,178 | +189% | 0 | 0 | — |
case-22 | fail→pass | 17,470 | 41,973 | +140% | 1 | 1 | 0% | 1,594 | 6,657 | +318% | 0 | 0 | — |
DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 22 cases were attempted. The headline lift of +27 percentage points is the difference between those two pass rates over the 22 comparable cases.
Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.
Other measured skills in the registry, with their headline benchmark lift.