Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Decode a used-car listing before you drive an hour to see it — what the seller's phrasing is hiding, the history-check items that matter, a test-drive and inspection checklist ordered by cost-of-miss, the questions that make evasive sellers visible, and the walk-away signs ranked 🔴🟡🟢. Use when someone says 'is this car listing legit', 'what should I check on a used car', 'decode this ad', or is about to buy their first car. Produces a listing decode, the viewing checklist, and the negotiation fr
.claude/skills/mohitagw15856-used-car-decoder/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-07 | ✗→✓ | ▲ Improved | 79% | 0% |
| case-10 | ✗→✓ | ▲ Improved | 97% | 0% |
| case-14 | ✓→✗ | ▼ Worse | 145% | 0% |
| case-12 | ✓→✓ | = Same ✓ | 40% | 0% |
| case-04 | ✓→✓ | = Same ✓ | 83% | 0% |
A used-car listing is a document written by someone who knows the car's problems, for someone who doesn't. The dialect is learnable: "selling for a friend" (distance from liability), "drives well for its age" (adjusted expectations), "minor cosmetic damage" (photographed from the good side), "no test drives without deposit" (walk away now). This skill decodes the ad the way lease-decoder]] reads a lease — severity-ranked, money math attached — then arms the viewing: the checks a non-mechanic can actually do, the questions that surface evasion, and the honest boundary: which findings mean pay a professional for an inspection and which mean leave.
typically means and the question that tests it
(accident/write-off status, finance owing, mileage consistency, recalls) — each flagged as country-specific-verify-locally with what to search, since registries differ everywhere
cold-start, the panel-gap walk, fluids, tires including the spare-match, electronics sweep, the specific listen-fors on the drive
would you fix next?", "can I take it for a pre-purchase inspection?" — the last one is the real test: honest sellers say yes)
worth in negotiation, and the walk-away list where no price is right
Ask for (if not already provided):
(make/model/year/mileage)
might inspect with them
buffer (insurance, immediate maintenance)
(this skill never asserts registry procedures)
distance ("for a friend/relative") · condition adjectives doing heavy lifting ("well for its age", "usual marks") · photo tells (no full-front shot, wet-car photos hiding paint, interior only) · urgency pressure ("first to see will buy") · the deposit-before- viewing 🔴s. Each flag gets its test question, not just suspicion.
consistent? History check run (accident, write-off category, outstanding finance — finance owing can follow the CAR, not the seller, in many places: 🔴 verify locally)? Recalls outstanding? Seller matches the registered keeper? All flagged with what-to-search terms, none asserted as universal procedure.
engines are a tell) → head-gasket-adjacent signs (mayonnaise under the oil cap, white smoke — described plainly for a novice) → panel gaps and paint-tone walk (accident repair) → tires including date-codes and the spare (uneven wear = alignment or worse) → electronics sweep (every window, warning-light theater: which lights come on at ignition and go off) → the drive: straight-line braking, full-lock turns, a listen with the radio OFF.
the sorting hat: "can my mechanic look at it?" — yes means proceed, any version of no means the listing decoded itself. "What would you fix next?" beats "any problems?" — everyone answers the second with "nothing."
prices) set the anchor; each finding gets its rough repair-cost class (small/medium/engine-money — classes, not fake precise quotes) as negotiation material. The walk-away list is absolute: finance owing unresolved, write-off category undisclosed, no-inspection sellers, mileage that doesn't add up. And the standing rule: the deposit buffer for year-one surprises is part of the budget, not optional.
## Listing decode
| Phrase / tell | Reading | 🔴🟡🟢 | Test question |
## Before you travel (verify-local items)
[ ] History check (search: …) [ ] Finance owing [ ] Mileage consistency
[ ] Recalls [ ] Seller = keeper — each with country-flag
## At the viewing (cost-of-miss order)
[The checklist with novice-friendly descriptions of each sign]
## Ask the seller
[The questions + what evasive answers look like]
## The money frame
[Comp anchor · findings → negotiation classes · the walk-away list ·
the year-one buffer line]
⚠ A pre-purchase inspection by a mechanic beats every checklist here for
engine/transmission health — budget for one on any car you're serious about.suspicion without a receipt
search terms, never asserted procedure
pro-inspection items say so
negotiation is explicitly the wrong response
"that's definitely the clutch," no
registers, deposit law) — flag and route
the framing
exists to identify which kind is across the table
buyer is excited
mechanic-quote-decoder]] for after you own it; car-lease-decoder]] for the leasing route; car-tco]] for what this car really costs per year; franklin-decision-ledger]] when it's down to two cars.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-12 | pass→pass | 27,643 | 50,817 | +84% | 1 | 1 | 0% | 3,315 | 4,650 | +40% | 0 | 0 | — |
case-01 | fail→fail | 59,196 | 49,954 | -16% | 1 | 1 | 0% | 6,989 | 7,536 | +8% | 0 | 0 | — |
case-02 | fail→fail | 53,351 | 46,445 | -13% | 1 | 1 | 0% | 6,093 | 7,002 | +15% | 0 | 0 | — |
case-03 | fail→fail | 56,985 | 40,226 | -29% | 1 | 1 | 0% | 7,472 | 6,353 | -15% | 0 | 0 | — |
case-04 | pass→pass | 13,854 | 19,886 | +44% | 1 | 1 | 0% | 2,243 | 4,098 | +83% | 0 | 0 | — |
case-11 | pass→pass | 22,184 | 33,504 | +51% | 1 | 1 | 0% | 2,592 | 5,739 | +121% | 0 | 0 | — |
case-05 | pass→pass | 18,411 | 26,164 | +42% | 1 | 1 | 0% | 2,209 | 4,420 | +100% | 0 | 0 | — |
case-06 | pass→pass | 22,757 | 25,972 | +14% | 1 | 1 | 0% | 3,243 | 5,514 | +70% | 0 | 0 | — |
case-07 | fail→pass | 25,939 | 26,187 | +1% | 1 | 1 | 0% | 3,092 | 5,550 | +79% | 0 | 0 | — |
case-08 | pass→pass | 20,574 | 38,323 | +86% | 1 | 1 | 0% | 2,407 | 6,345 | +164% | 0 | 0 | — |
case-09 | pass→pass | 24,732 | 33,935 | +37% | 1 | 1 | 0% | 3,095 | 5,916 | +91% | 0 | 0 | — |
case-10 | fail→pass | 23,362 | 31,556 | +35% | 1 | 1 | 0% | 2,906 | 5,731 | +97% | 0 | 0 | — |
case-13 | pass→pass | 21,632 | 27,214 | +26% | 1 | 1 | 0% | 2,407 | 4,563 | +90% | 0 | 0 | — |
case-14 | pass→fail | 20,139 | 35,816 | +78% | 1 | 1 | 0% | 2,215 | 5,427 | +145% | 0 | 0 | — |
case-15 | pass→pass | 20,849 | 15,474 | -26% | 1 | 1 | 0% | 2,248 | 4,027 | +79% | 0 | 0 | — |
case-16 | pass→pass | 22,025 | 39,614 | +80% | 1 | 1 | 0% | 2,388 | 4,992 | +109% | 0 | 0 | — |
case-17 | pass→pass | 24,703 | 25,011 | +1% | 1 | 1 | 0% | 2,597 | 4,570 | +76% | 0 | 0 | — |
case-18 | pass→pass | 28,145 | 23,381 | -17% | 1 | 1 | 0% | 2,961 | 5,218 | +76% | 0 | 0 | — |
case-19 | pass→pass | 26,580 | 38,773 | +46% | 1 | 1 | 0% | 2,848 | 4,011 | +41% | 0 | 0 | — |
case-20 | pass→pass | 17,260 | 18,282 | +6% | 1 | 1 | 0% | 2,536 | 4,237 | +67% | 0 | 0 | — |
case-21 | pass→pass | 39,079 | 30,336 | -22% | 1 | 1 | 0% | 3,959 | 5,303 | +34% | 0 | 0 | — |
case-22 | pass→pass | 22,427 | 24,550 | +9% | 1 | 1 | 0% | 2,742 | 4,411 | +61% | 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 +5 percentage points is the difference between those two pass rates over the 22 comparable cases. 1 case got worse with the skill loaded, and it is included in that figure.
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.