▮ Automotive AI Skills
Doc /skills/repair-estimate-reviewer/README.htmlENJA

Repair Estimate Reviewer

English 日本語

Compare written repair estimates on scope, evidence, arithmetic, and unanswered questions before authorizing work.

Description

A cheaper total may cover different parts, labor, or follow-up work. This skill normalizes itemized estimates into a comparable table and prepares factual questions for the shop. It reviews the quote; it does not diagnose the vehicle, certify a fair market price, or approve delaying a safety repair.

Audience: Vehicle owners, service advisors, used-car businesses, and fleet purchasers reviewing repair proposals.

Capabilities

Example Prompts

2016 Honda Civic 2.0 gasoline, 120,000 km, US. Compare two written air
conditioning estimates: one includes leak diagnosis, the other only
refrigerant service. Identify missing scope before comparing price.

2018 Toyota Aqua 1.5 hybrid, 95,000 km, Japan. A synthetic quote lists
parts ¥12,000, labor 1.5 hours at ¥8,000/hour, and supplies ¥1,000.
All are before a stated 10% tax on the entire subtotal. Check the math;
do not treat these illustrative figures as a market price or tax rule.

2015 Volkswagen Golf 1.4 TSI, 140,000 km, UK. One quote lists separate
labor for two jobs that may share disassembly. Draft a question about
included operations rather than accusing the shop of duplicate billing.

2020 Nissan Leaf 40 kWh, 80,000 km, Japan. The shop proposes a traction
battery repair but the quote has no diagnostic report or warranty terms.
Prepare questions for the qualified EV specialist; no DIY instructions.

For model-specific checks, follow reference routing using the variant, production year, and original market. The browser builder can prepare the prompt. A reference match does not establish a fault, eligibility, or completion.

Standalone installation: If the repository-relative routing guide or data files are missing, use the online routing guide or browser builder. Ask for the exact variant, production year and original market; inspect the full reference and current official sources. If the resources cannot be accessed or no profile matches, say model-specific evidence is unavailable and continue only with general guidance. A match is not a diagnosis, recall eligibility, or proof of completion.

  1. Collect context and documents. Get vehicle details, mileage, market, symptoms, diagnostic findings, dates, currency, tax wording, and the complete relevant line items. Redact names, addresses, payment data, and unnecessary identifiers.
  2. Normalize without filling gaps. Return a table: item/operation; evidence; parts and quantity; labor basis; other charges; tax treatment; included/excluded; uncertainty. Keep lump-sum labor as lump sum if hours are not provided.
  3. Check arithmetic. Calculate quantity × unit price and stated labor hours × rate, then only the declared fees and tax basis. Show discrepancies and rounding assumptions. If tax inclusion or quantity is unclear, ask instead of silently choosing a total.
  4. Compare equivalent scope. Note differences in parts condition, diagnosis, calibration, alignment, fluids, disposal, and warranty where relevant. Flag possible shared labor as a question for the shop, not a proven duplicate.
  5. Prioritize questions using evidence. Ask what finding supports each repair and what the manufacturer schedule says about maintenance. Urgency comes from qualified inspection; do not recommend deferring brake, steering, SRS, fuel, or high-voltage work to save money.
  6. Return a decision brief. Provide the comparison, arithmetic, unresolved items, and up to five questions. Record what the owner has authorized separately; do not authorize, book, or send a message on their behalf unless requested.

Limitations