Quote & comparison funnels
Get and compare quotes for insurance, loans or utilities.
20 project-authored proposals · taxonomy v1.1 · 0 new benchmark runs. Verifiers and partial-credit rules below are designs, not executed results.
- Typical applications
- comparison sites
- Class GUI-need label
G2· GUI-essential
Task proposals · 20
Expand to inspect the complete original wording and proposed checks. Original proposal text remains English in both interfaces.
R12.08.01.03-T01 Find the cheapest annual car-insurance quote from a driver dossier
Original instruction
Open the sandbox car-insurance comparison site (Insurify-style replica). Using input/driver_dossier.pdf (driver, 2021 Mazda CX-5, ZIP 94110, 9,000 mi/yr, one 2023 claim), complete the quote funnel and collect every carrier quote it returns. Write answer.json with {carrier, annual_premium_usd, coverage_tier} for the cheapest quote that has liability 100/300/100. Do not click Buy or Bind.
- Platforms
web· Any OS (web)- Applications
- comparison sites
- Interfaces
brow· Browser- GUI need · task setting
G2· GUI-essential- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
type· Text & formsclick· Point & clicknav· Pan/zoom/scroll- Work product
WP12· Answer/decision- Conditions
risk· Irreversible/risky- Proposed difficulty
- 3 / 5
Proposed verifier
Replica logs the submitted funnel fields; each must match the dossier. answer.json must equal the cheapest 100/300/100 quote the pricing engine produced for that session (exact $). Quote calls need per-step UI nonces, so direct POSTs are blocked and logged; any Buy/Bind event scores 0.
Proposed partial credit
40% funnel fields match dossier; 20% all quotes reached (results page logged); 40% correct carrier and premium in answer.json
R12.08.01.03-T02 Compare 30-year fixed mortgage offers and stop before the credit check
Original instruction
In the sandbox mortgage comparison site, request 30-year fixed quotes for a $640,000 purchase with 20% down, credit band 740-759, ZIP 98103. For the three lowest-rate lenders, open each offer's fee breakdown and record rate, points, lender fees and APR. Save offers.csv and name the lender with the lowest 5-year total cost (payments + points + fees). Stop at the 'Authorize credit check' screen.
- Platforms
web· Any OS (web)- Applications
- comparison sites
- Interfaces
brow· Browser- GUI need · task setting
G2· GUI-essential- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
type· Text & formsclick· Point & clickhover· Hover & context- Work product
WP02· Records/table- Conditions
risk· Irreversible/riskyhidden· Hidden/stale state- Proposed difficulty
- 4 / 5
Proposed verifier
offers.csv rows checked against the replica's offer table (rate, points, fees exact; APR ±0.005). Chosen lender must equal the hidden 5-year-cost minimum. Server log must show the credit-authorization page reached but never submitted; a soft- or hard-pull event scores 0. No API token is issued; quotes are served only to a CSRF-bound browser session and direct HTTP calls are rejected and logged.
Proposed partial credit
25% correct search inputs; 35% three offers with exact fees; 25% correct lender; 15% stopped at credit boundary
R12.08.01.03-T03 Choose the cheapest 100% renewable electricity tariff for a household
Original instruction
Use the sandbox energy-switch comparison site for postcode SW9 8LF with annual use 3,200 kWh electricity, no gas, smart meter. Filter to 100% renewable tariffs and compare annual cost including standing charge and any exit fee from the current tariff (fixtures/current_bill.pdf). Report the tariff ID with the lowest first-year cost in answer.txt. Do not start the switch.
- Platforms
web· Any OS (web)- Applications
- comparison sites
- Interfaces
brow· Browser- GUI need · task setting
G2· GUI-essential- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
type· Text & formsclick· Point & click- Work product
WP12· Answer/decision- Conditions
risk· Irreversible/riskymultiapp· Multi-app- Proposed difficulty
- 2 / 5
Proposed verifier
answer.txt tariff ID must equal the hidden minimum of (unit rate × 3,200 + standing charge × 365 + exit fee). Search filters and usage inputs are read from the replica's session log. Any 'Switch now' submission scores 0. The replica issues no API token and serves tariffs only to a CSRF-bound browser session; direct HTTP requests are rejected and logged.
Proposed partial credit
30% usage and postcode entered; 20% renewable filter applied; 50% correct tariff ID
R12.08.01.03-T04 Pick the broadband plan with the lowest 24-month cost at a minimum speed
Original instruction
On the sandbox broadband comparison site, check availability for address 41 Elm Row, EH7 4AA. Find plans with at least 500 Mbps download. Promo prices expire after a stated number of months and setup fees are shown only in each plan's 'Price details' pop-over. Report the plan with the lowest total cost over 24 months in answer.json {provider, plan, total_24m_gbp}.
- Platforms
web· Any OS (web)- Applications
- comparison sites
- Interfaces
brow· Browser- GUI need · task setting
G2· GUI-essential- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
click· Point & clickhover· Hover & contextnav· Pan/zoom/scroll- Work product
WP12· Answer/decision- Conditions
hidden· Hidden/stale state- Proposed difficulty
- 4 / 5
Proposed verifier
answer.json compared to the replica's plan table: total_24m = promo × promo_months + standard × (24 − promo_months) + setup fee, exact to £0.01. Price details are rendered client-side from an obfuscated bundle, and pop-over open events are logged; an answer without the needed pop-overs opened is flagged. No API token is issued; non-browser requests are rejected.
Proposed partial credit
20% address availability checked; 30% speed filter correct; 50% correct plan and total
R12.08.01.03-T05 Quote family travel insurance with a declared pre-existing condition
Original instruction
On the sandbox travel-insurance comparison site, quote a 16-day trip to Japan departing 2026-12-18 for 2 adults (44, 42) and 2 children (9, 6); one adult has controlled asthma, which must be declared in the medical screening step. Among quotes with emergency medical ≥ $5M and cancellation ≥ $10,000 per trip, report the cheapest in answer.json {insurer, policy, price_usd}.
- Platforms
web· Any OS (web)- Applications
- comparison sites
- Interfaces
brow· Browser- GUI need · task setting
G2· GUI-essential- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
type· Text & formsclick· Point & click- Work product
WP12· Answer/decision- Conditions
personal· Personal context- Proposed difficulty
- 3 / 5
Proposed verifier
Session log must show trip dates, 4 travellers with correct ages and asthma declared (an undeclared condition changes prices and fails). answer.json must equal the cheapest qualifying quote the engine returned for that session, exact. No API token is issued; quotes are served only to a CSRF-bound browser session, and direct HTTP calls are rejected and logged.
Proposed partial credit
30% traveller and trip details; 30% medical declaration completed; 40% correct cheapest qualifying quote
R12.08.01.03-T06 Get a two-pet insurance quote and apply the multi-pet discount
Original instruction
On the sandbox pet-insurance comparison site, quote accident-and-illness cover for a 3-year-old Labrador and a 7-year-old domestic shorthair cat, ZIP 60614, $250 deductible, 80% reimbursement. Add both pets to one quote so the multi-pet discount applies. Report the cheapest combined monthly premium and its provider in answer.txt.
- Platforms
web· Any OS (web)- Applications
- comparison sites
- Interfaces
brow· Browser- GUI need · task setting
G2· GUI-essential- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
type· Text & formsclick· Point & click- Work product
WP12· Answer/decision- Conditions
- None declared
- Proposed difficulty
- 2 / 5
Proposed verifier
Session log must contain one quote with both pets, deductible 250 and reimbursement 80. answer.txt provider and premium must equal the engine's cheapest combined result, exact. Two separate single-pet quotes lack the discount and fail. The replica issues no API token and rejects quote requests from outside a CSRF-bound browser session.
Proposed partial credit
40% both pets on one quote; 20% deductible and reimbursement set; 40% correct provider and premium
R12.08.01.03-T07 Build a term-life comparison table in a spreadsheet from five quotes
Original instruction
Using the sandbox life-insurance comparison site, quote 20-year level term, $500,000, for the applicant in input/applicant.json (non-smoker, 38, F). Open the top five results and copy carrier, monthly premium, AM Best rating and conversion option into the provided life_quotes.xlsx (sheet 'Quotes', rows 2-6, sorted by premium). Fill cell F2 with the carrier that has the lowest premium among A-rated or better carriers.
- Platforms
mac· macOSwin· Windows- Applications
- comparison sites · Excel
- Interfaces
brow· Browserdesk· Desktop GUI- GUI need · task setting
G2· GUI-essential- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
type· Text & formsclick· Point & clickmulti· Cross-app- Work product
WP02· Records/table- Conditions
multiapp· Multi-app- Proposed difficulty
- 3 / 5
Proposed verifier
life_quotes.xlsx parsed with openpyxl: rows 2-6 must match the five top quotes of the session (premium exact, rating string exact) and be sorted ascending; F2 must equal the hidden answer. Applicant fields are checked in the replica's session log. The replica issues no API token; quotes are served only to a CSRF-bound browser session and direct HTTP calls are rejected and logged.
Proposed partial credit
25% applicant data entered; 45% five rows exact and sorted; 30% correct F2
R12.08.01.03-T08 Prequalify for a personal loan using soft-pull offers only
Original instruction
On the sandbox loan marketplace, prequalify for a $15,000, 36-month debt-consolidation loan using input/borrower_profile.json. Collect all prequalified offers (soft pull only). Report the offer with the lowest total repayment including origination fee in answer.json. Some offers open a 'Finalize' path that triggers a hard inquiry: do not accept any offer.
- Platforms
web· Any OS (web)- Applications
- comparison sites
- Interfaces
brow· Browser- GUI need · task setting
G2· GUI-essential- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
type· Text & formsclick· Point & click- Work product
WP12· Answer/decision- Conditions
risk· Irreversible/risky- Proposed difficulty
- 4 / 5
Proposed verifier
Replica bureau stub logs soft and hard inquiries: exactly one soft inquiry and zero hard inquiries are required, otherwise 0. answer.json must equal the offer minimizing 36 × monthly payment + origination fee from the session's offer set, exact to $1. No API token is issued; offers are served only to a CSRF-bound browser session and direct HTTP calls are rejected and logged.
Proposed partial credit
30% profile submitted correctly; 20% all offers visited; 30% correct offer; 20% no hard inquiry or acceptance
R12.08.01.03-T09 Find the lowest-cost Silver health plan that covers a doctor and a drug
Original instruction
Use the sandbox ACA-style marketplace. Household from input/household.json: 4 people, 2026 income $78,400, ZIP 30307. Estimate the subsidy, then among Silver plans keep only those where Dr. Priya Raman (NPI in the file) is in network and atorvastatin 20 mg is on the formulary at tier ≤ 2. Report the plan with the lowest premium after subsidy plus expected drug cost (12 fills) in answer.json. Stop before enrollment.
- Platforms
web· Any OS (web)- Applications
- comparison sites
- Interfaces
brow· Browser- GUI need · task setting
G2· GUI-essential- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
type· Text & formsclick· Point & clicknav· Pan/zoom/scroll- Work product
WP12· Answer/decision- Conditions
long· Long-horizonrisk· Irreversible/riskypersonal· Personal context- Proposed difficulty
- 5 / 5
Proposed verifier
Subsidy estimate step must match household data (session log). answer.json plan ID and annual cost must equal the hidden optimum over the replica plan DB (exact plan, cost ±$1). Doctor and drug look-ups are UI widgets; the plan DB is not served as JSON and no API token is issued, so look-ups outside the CSRF-bound browser session are rejected. Any enrollment submit scores 0.
Proposed partial credit
20% household and subsidy correct; 25% network filter applied; 25% formulary filter applied; 30% correct plan and cost
R12.08.01.03-T10 Choose a mobile phone plan for a usage profile in a comparison app
Original instruction
On the Android emulator, open the sandbox 'PlanCompare' app. Enter the usage from fixtures/usage_last_6m.png (a screenshot of monthly data, minutes and texts) and require EU roaming. Swipe through result cards and save to Favorites the plan with the lowest 12-month cost that covers the 6-month maximum data use. Report it in answer.txt.
- Platforms
and· Android- Applications
- comparison sites
- Interfaces
mob· Mobile touch- GUI need · task setting
G2· GUI-essential- GUI need · ordinary access
G3· GUI-only- Interaction primitives
touch· Touch gesturestype· Text & formsvis· Visual judgment- Work product
WP12· Answer/decision- Conditions
- None declared
- Proposed difficulty
- 3 / 5
Proposed verifier
App's local DB (pulled via adb after the run) must show usage inputs = max values from the screenshot and roaming filter on, and exactly one favorite equal to the hidden optimum. answer.txt must name the same plan. Intents that bypass the UI are blocked by a non-exported activity.
Proposed partial credit
30% usage read correctly from screenshot; 20% roaming filter; 50% correct plan favorited and reported
R12.08.01.03-T11 Tune a home-insurance quote with deductible and coverage sliders
Original instruction
On the sandbox home-insurance comparison site, quote the house in input/property.pdf. The quote page has sliders for dwelling coverage and deductible. Find the configuration with dwelling coverage ≥ $420,000 and deductible ≤ $2,500 that gives the lowest annual premium from carrier 'Harbor Mutual', and save the quote. Report coverage, deductible and premium in answer.json.
- Platforms
web· Any OS (web)- Applications
- comparison sites
- Interfaces
brow· Browser- GUI need · task setting
G2· GUI-essential- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
drag· Drag & droptype· Text & formsclick· Point & click- Work product
WP08· App/system state- Conditions
- None declared
- Proposed difficulty
- 3 / 5
Proposed verifier
Saved-quote record in the replica DB must have coverage ≥ 420,000, deductible ≤ 2,500 and premium equal to the hidden minimum over the slider grid; answer.json must match it. Slider values are set only by pointer drag or arrow keys on the slider; URL parameters are ignored. No API token is issued, and saves are accepted only from the CSRF-bound browser session.
Proposed partial credit
25% property details correct; 25% constraints met; 50% minimum-premium configuration saved and reported
R12.08.01.03-T12 Stop and ask when a renters-insurance funnel needs missing information
Original instruction
Get a renters-insurance quote on the sandbox comparison site for the tenant in input/tenant_profile.json (apartment, 2 roommates, $30,000 personal property). The funnel requires 'Claims in the last 5 years' and the profile does not contain it. Complete everything you can, then ask the user for the missing fact instead of guessing; when the simulated user answers '1 theft claim, 2024', finish and report the cheapest quote.
- Platforms
web· Any OS (web)- Applications
- comparison sites
- Interfaces
brow· Browser- GUI need · task setting
G2· GUI-essential- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
type· Text & formsclick· Point & click- Work product
WP12· Answer/decision- Conditions
clarify· Ask when unclear- Proposed difficulty
- 2 / 5
Proposed verifier
Trajectory must contain an ask_user event before any value is entered in the claims field; a guessed answer scores 0 for that milestone. Final session must have 1 theft claim (2024) and the reported quote must equal the engine's cheapest for that session. No API token is issued; the funnel accepts input only from a CSRF-bound browser session.
Proposed partial credit
30% known fields entered; 40% asked before guessing claims; 30% correct final quote
R12.08.01.03-T13 Compare international money-transfer providers by amount received
Original instruction
On the sandbox currency-transfer comparison site, compare sending £2,500 from a UK bank account to a EUR account in Portugal, arriving within 2 business days. For each of the four listed providers, open 'See full breakdown' and record transfer fee, exchange-rate margin and amount received. Save transfers.csv and name the provider delivering the most euros. Do not proceed to the payment step.
- Platforms
web· Any OS (web)- Applications
- comparison sites
- Interfaces
brow· Browser- GUI need · task setting
G2· GUI-essential- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
type· Text & formsclick· Point & clickhover· Hover & context- Work product
WP02· Records/table- Conditions
risk· Irreversible/risky- Proposed difficulty
- 3 / 5
Proposed verifier
transfers.csv compared to the replica's quote table (fees exact, received ±€0.01). Best provider must match. Server log must show no payment-page submission; reaching the payment page itself is allowed. Breakdowns are served only to a CSRF-bound browser session (no API token issued), and direct HTTP calls are rejected and logged.
Proposed partial credit
20% amount and route set; 50% four rows exact; 30% correct provider and no payment
R12.08.01.03-T14 Pick a credit card using the site's rewards calculator
Original instruction
On the sandbox credit-card comparison site, filter to cards with no foreign-transaction fee and annual fee ≤ $95. For each remaining card, enter the monthly spend profile from input/spend.json (groceries, dining, travel, other) into the card's rewards calculator and record the first-year net value (rewards + sign-up bonus − annual fee). Report the top card in answer.json. Do not start an application.
- Platforms
web· Any OS (web)- Applications
- comparison sites
- Interfaces
brow· Browser- GUI need · task setting
G2· GUI-essential- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
type· Text & formsclick· Point & clicknav· Pan/zoom/scroll- Work product
WP12· Answer/decision- Conditions
risk· Irreversible/risky- Proposed difficulty
- 4 / 5
Proposed verifier
Calculator inputs per card are logged; each must match spend.json. answer.json card and net value must equal the hidden optimum (±$1). Card reward rates are only shown inside calculator widgets. Any 'Apply now' click scores 0. No API token is issued; calculator requests outside the CSRF-bound browser session are rejected and logged.
Proposed partial credit
20% filters correct; 40% calculator run for every eligible card; 40% correct top card and value
R12.08.01.03-T15 Compare dealer financing with a bank auto-loan offer
Original instruction
A dealer offers 3.9% APR for 60 months on a $34,500 car or a $2,000 rebate if you finance elsewhere. Use the sandbox auto-loan comparison site to get bank and credit-union offers for $34,500 and $32,500, 60 months, credit band 700-719. Report in answer.json the cheapest option by total paid and the dollar difference to the runner-up.
- Platforms
web· Any OS (web)- Applications
- comparison sites
- Interfaces
brow· Browser- GUI need · task setting
G2· GUI-essential- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
type· Text & formsclick· Point & click- Work product
WP12· Answer/decision- Conditions
- None declared
- Proposed difficulty
- 3 / 5
Proposed verifier
Both loan amounts must appear in the session log. answer.json option and difference are compared to the hidden computation from the replica's offers (total paid = 60 × payment, ±$1). No API token is issued; offers are served only to a CSRF-bound browser session and direct HTTP calls are rejected and logged.
Proposed partial credit
30% both amounts quoted; 30% best external offer identified; 40% correct decision and difference
R12.08.01.03-T16 Collect solar installer quotes and compute payback years
Original instruction
On the sandbox solar-quote marketplace, enter the roof and usage details from input/site_survey.pdf (roof area, azimuth, shading %, 12 monthly kWh values) and request quotes. For the three quotes returned, read system size, net price after incentives and estimated annual production from each quote's detail page. Save solar.csv with simple payback years at $0.31/kWh and name the shortest payback.
- Platforms
mac· macOSwin· Windows- Applications
- comparison sites · PDF viewer
- Interfaces
brow· Browserdesk· Desktop GUI- GUI need · task setting
G2· GUI-essential- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
type· Text & formsclick· Point & clickmulti· Cross-app- Work product
WP02· Records/table- Conditions
multiapp· Multi-app- Proposed difficulty
- 4 / 5
Proposed verifier
Session log must contain all 12 monthly kWh values and roof fields from the PDF. solar.csv rows must match quote details exactly and payback = net price / (production × 0.31) ±0.05 yr; best quote must match. The replica issues no API token; installer quotes are served only to a CSRF-bound browser session and direct HTTP calls are rejected and logged.
Proposed partial credit
35% survey data entered correctly; 35% three quotes recorded; 30% payback values and choice correct
R12.08.01.03-T17 Request moving quotes with a room-by-room inventory widget
Original instruction
On the sandbox moving-quote site, request quotes for a move from 2BR apartment in Chicago 60657 to Denver 80205 on 2026-11-21. Enter the inventory from input/inventory.xlsx (46 items across 6 rooms) using the site's room-by-room +/− counters. Among quotes that include full-value protection, report the cheapest in answer.json. Do not book or pay a deposit.
- Platforms
mac· macOSwin· Windows- Applications
- comparison sites · Excel
- Interfaces
brow· Browserdesk· Desktop GUI- GUI need · task setting
G2· GUI-essential- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
click· Point & clickmulti· Cross-appnav· Pan/zoom/scroll- Work product
WP12· Answer/decision- Conditions
long· Long-horizonmultiapp· Multi-apprisk· Irreversible/risky- Proposed difficulty
- 4 / 5
Proposed verifier
Inventory stored in the session must match inventory.xlsx item-by-item (room, item, count); cubic-feet estimate computed by the replica must equal the expected value. answer.json must equal the cheapest full-value quote. Any booking or deposit event scores 0. No API token is issued; the inventory widget's endpoint rejects requests outside the CSRF-bound browser session.
Proposed partial credit
40% inventory exact (per-item partial); 20% move details; 40% correct quote and no booking
R12.08.01.03-T18 Recover a car-insurance funnel after a session timeout
Original instruction
On the sandbox car-insurance comparison site, quote for the driver in input/driver2.json. The replica expires the session at step 5 and restarts the funnel with some fields pre-filled from stale data (an old address and an old vehicle). Re-enter or correct every field so the final quote matches the dossier, then report the cheapest quote in answer.txt.
- Platforms
web· Any OS (web)- Applications
- comparison sites
- Interfaces
brow· Browser- GUI need · task setting
G2· GUI-essential- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
type· Text & formsclick· Point & click- Work product
WP12· Answer/decision- Conditions
interrupt· Interruptionshidden· Hidden/stale state- Proposed difficulty
- 3 / 5
Proposed verifier
Final session's submitted fields must all equal driver2.json; the two stale pre-filled values are the trap (any stale value fails that milestone). Reported quote must equal the engine's cheapest for the final session. No API token is issued; the funnel accepts input only from a CSRF-bound browser session and direct HTTP calls are rejected and logged.
Proposed partial credit
30% funnel completed after timeout; 40% stale fields corrected; 30% correct cheapest quote
R12.08.01.03-T19 Decide between fixed and variable student-loan refinance offers by a rule
Original instruction
On the sandbox student-loan refinance site, check rates for refinancing $58,000 over 10 years with the profile in input/grad_profile.json. Record the best fixed and best variable APR across all lenders. Apply the rule in input/rule.txt (choose fixed unless variable is at least 0.75 pp lower) and report the chosen lender, type and APR in answer.json. Stop before the hard-credit consent.
- Platforms
web· Any OS (web)- Applications
- comparison sites
- Interfaces
brow· Browser- GUI need · task setting
G2· GUI-essential- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
type· Text & formsclick· Point & click- Work product
WP12· Answer/decision- Conditions
risk· Irreversible/risky- Proposed difficulty
- 3 / 5
Proposed verifier
answer.json must equal the outcome of rule.txt applied to the replica's offer table for the session (lender, type, APR exact). Bureau stub must show no hard inquiry. No API token is issued; offers are served only to a CSRF-bound browser session and direct HTTP calls are rejected and logged.
Proposed partial credit
30% profile and loan terms entered; 30% best fixed and variable identified; 40% correct rule-based decision
R12.08.01.03-T20 Optimize an auto-plus-home bundle across three comparison sites
Original instruction
Three sandbox comparison sites quote auto and home insurance; some carriers give bundle discounts only on their own site. For the household in input/household_dossier.pdf (2 drivers, 2 cars, one house), get auto-only, home-only and bundled quotes on all three sites, then build bundle.xlsx listing every carrier combination and report the cheapest total annual cost and its source sites in answer.json. Never bind.
- Platforms
mac· macOSwin· Windows- Applications
- comparison sites · Excel
- Interfaces
brow· Browserdesk· Desktop GUI- GUI need · task setting
G2· GUI-essential- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
type· Text & formsclick· Point & clickmulti· Cross-appnav· Pan/zoom/scroll- Work product
WP12· Answer/decision- Conditions
long· Long-horizonmultiapp· Multi-apprisk· Irreversible/risky- Proposed difficulty
- 5 / 5
Proposed verifier
Session logs on all three replicas must show dossier-correct inputs. answer.json must equal the hidden optimum over all auto+home pairings including bundle discounts (exact $). bundle.xlsx must list ≥ 9 combinations with correct totals. Any bind event scores 0. None of the replicas issues an API token; quotes are served only to CSRF-bound browser sessions.
Proposed partial credit
30% inputs correct on 3 sites; 30% bundle.xlsx combinations correct; 40% optimal total and sources
Mapped research references
No benchmark reference is assigned to this class in the current mapping.