Auto Repair Estimate Decoder Skill

Auto Repair Estimate Decoder Skill

Repair estimates mix three different things on one page: what your car needs now, what it will need eventually, and what the shop would like to sell you this week. This skill separates them — decoding each line into plain language, triaging by "what happens if I don't," and flagging the patterns (fluid-flush bundles, replace-instead-of-repair, labor-hour padding) that distinguish an estimate from an invoice-in-waiting. It works from the estimate text; it never diagnoses the car.

What This Skill Produces

• A line-by-line decode: what each item is, what it does, what happens if deferred
• Urgency triage: safety-now / soon-with-timeframe / eventually / optional-upsell
• Sanity checks: labor hours vs. the job's typical shape, parts pricing patterns, diagnostic-fee handling
• Ranked red flags and the authorize/decline/second-opinion recommendation per line, plus the questions to ask

Required Inputs

Ask for these only if they aren't already provided:

• The estimate text — every line with prices; photos of the writeup work. Partial estimates get decoded with the missing pieces named.
• The car and the symptom — year/make/model/mileage, and what brought it in (the estimate should connect to the symptom; lines that don't are the interesting ones).
• Context — how long they plan to keep the car, and whether this shop has history with them.

Framework: Severity Scale

• 🔴 Challenge before authorizing — repairs unrelated to the presenting symptom appearing without explanation, "replace" where "repair/resurface" is standard practice for the condition described, fluid-flush bundles on intervals far shorter than typical service schedules, labor hours that imply the job twice ("remove and replace" charged for overlapping operations), vague lines ("front end work") with specific prices, diagnostic fees charged and not credited toward authorized work (shop-dependent — ask).
• 🟡 Verify — reasonable questions to ask — OEM vs. aftermarket parts pricing unstated, "while we're in there" additions (sometimes genuinely economical — labor overlap is real; make them show the overlap math), preventive recommendations stated without the wear measurement that justifies them ("brakes at 4mm" is information; "brakes soon" is a mood).
• 🟢 Looks standard — lines that match the symptom, itemized parts+labor, wear items at plausible mileage; say so — most shops are honest, and knowing which lines are fine is half the value.

Triage every line by the deferral question: what specifically happens if this isn't done today? Safety items (brakes at spec limits, steering, tires at cords) get named as safety items. "Eventually" items get a mileage/timeframe. Anything the shop can't attach a consequence to is optional by definition. Frame typical-cost comparisons as ranges to verify locally, never as fact — labor rates vary wildly by region.

Output Format

Repair Estimate Decode: [vehicle — shop, date]

1. The verdict — total quoted vs. the defensible-now subset, in two sentences.

2. Line-by-line decode

3. 🚩 Red flags, ranked — the quoted line, the pattern it matches, and the question that resolves it.