Reservations & appointments
Book restaurants, appointments and services at valid times.
20 project-authored proposals · taxonomy v1.1 · 0 new benchmark runs. Verifiers and partial-credit rules below are designs, not executed results.
- Typical applications
- OpenTable · booking widgets
- 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.
R12.04.01.02-T01 Reserve an outdoor table for six on Friday evening with an accessibility note
Original instruction
On the OpenDining clone, reserve a table for 6 this Friday between 19:00 and 20:00 at an Italian restaurant within 2 miles of the user's home with outdoor seating, rating >= 4.3, and add the note 'wheelchair access needed'. Choose the earliest qualifying slot.
- Platforms
web· Any OS (web)- Applications
- OpenDining clone
- Interfaces
brow· Browser- GUI need · task setting
G2· GUI-essential- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
click· Point & clickvis· Visual judgmenttype· Text & forms- Work product
WP08· App/system state- Conditions
- None declared
- Proposed difficulty
- 2 / 5
Proposed verifier
Reservation row: party 6, restaurant in feasible set (cuisine, distance, rating, outdoor), slot = earliest feasible in 19:00-20:00 across feasible restaurants, seating=outdoor, note exact. The clone issues no API token and accepts bookings only from browser sessions (CSRF + session id), so scripted requests are rejected.
Proposed partial credit
30% restaurant feasible; 30% earliest slot; 20% seating; 20% note.
R12.04.01.02-T02 Book the earliest DMV knowledge-test appointment across two offices
Original instruction
On the DMV appointment clone, book the earliest available 'Knowledge test' appointment at either the Oakland or Hayward office for the applicant in fixtures/applicant.json. Slots are shown in a week-by-week calendar grid; check at least the next 4 weeks. Save the confirmation PDF.
- Platforms
web· Any OS (web)- Applications
- DMV appointment clone
- Interfaces
brow· Browser- GUI need · task setting
G2· GUI-essential- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
click· Point & clicknav· Pan/zoom/scrollvis· Visual judgment- Work product
WP08· App/system state- Conditions
- None declared
- Proposed difficulty
- 3 / 5
Proposed verifier
Appointment row: service=knowledge_test, slot = global earliest across both offices (seeded in week 3 at Hayward), applicant fields exact; PDF saved with confirmation number. The clone issues no API token and accepts bookings only from browser sessions (CSRF + session id), so scripted requests are rejected.
Proposed partial credit
50% earliest slot; 30% applicant fields; 20% PDF saved.
R12.04.01.02-T03 Book a dog grooming and nail trim on Saturday morning
Original instruction
On the pet-store booking clone, book a full groom plus nail trim for the user's dog 'Biscuit' (profile exists) on Saturday morning before 11:00 with any groomer certified for large breeds (badge on groomer cards). Add 'nervous with dryers' to notes.
- Platforms
web· Any OS (web)- Applications
- pet store booking clone
- Interfaces
brow· Browser- GUI need · task setting
G2· GUI-essential- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
click· Point & clickvis· Visual judgmenttype· Text & forms- Work product
WP08· App/system state- Conditions
- None declared
- Proposed difficulty
- 2 / 5
Proposed verifier
Booking: pet=Biscuit, services groom+nail_trim, Saturday slot start < 11:00 and end fits, groomer has large-breed cert (icon only), note text. The clone issues no API token and accepts bookings only from browser sessions (CSRF + session id), so scripted requests are rejected.
Proposed partial credit
30% services; 30% slot; 25% groomer certification; 15% note.
R12.04.01.02-T04 Book a table at a specific restaurant over three weekends with a peanut allergy note
Original instruction
On the OpenDining clone, book a table for 2 with outdoor seating at '347 Grille' on any day in the next three weekends between 17:30 and 20:00 (prefer Saturday, then Sunday, then Friday). Note a peanut allergy and the occasion 'Anniversary'.
- Platforms
web· Any OS (web)- Applications
- OpenDining clone
- Interfaces
brow· Browser- GUI need · task setting
G2· GUI-essential- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
click· Point & clicknav· Pan/zoom/scrollvis· Visual judgment- Work product
WP08· App/system state- Conditions
- None declared
- Proposed difficulty
- 3 / 5
Proposed verifier
Reservation: restaurant id, party 2, outdoor, slot in window obeying preference order (first available Saturday in the 3 weekends; seeded only Sunday available in weekend 1, Saturday in weekend 2), notes contain 'peanut' and occasion field. The clone issues no API token and accepts bookings only from browser sessions (CSRF + session id), so scripted requests are rejected.
Proposed partial credit
40% correct slot by preference order; 20% outdoor; 20% allergy note; 20% occasion.
R12.04.01.02-T05 Modify an existing reservation's party size and time
Original instruction
On the OpenDining clone, the user's reservation at 'Bistro Lune' tonight 19:00 for 4 must change to 6 people and 30 minutes later. If 19:30 for 6 is unavailable, choose the closest time within ±45 min and tell the user. Do not cancel and rebook (the restaurant would lose the original booking's deposit).
- Platforms
web· Any OS (web)- Applications
- OpenDining clone
- Interfaces
brow· Browser- GUI need · task setting
G2· GUI-essential- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
click· Point & clickvis· Visual judgment- Work product
WP08· App/system state- Conditions
- None declared
- Proposed difficulty
- 3 / 5
Proposed verifier
Same reservation id updated: party 6, time = 19:30 or seeded closest (19:45); no cancellation rows; message to user if time differs. The clone issues no API token and accepts reservation edits only from the browser session (CSRF + session id), so scripted updates are rejected.
Proposed partial credit
40% same reservation modified; 40% party/time correct; 20% user informed when needed.
R12.04.01.02-T06 Book a haircut and color with a specific stylist fitting a calendar gap
Original instruction
Find a gap in the user's macOS Calendar next week of at least 2.5 hours between 09:00 and 18:00 and book 'Cut + color' (150 min) with stylist 'Rosa' on the salon booking widget clone at the earliest slot fitting that gap. Add the booking to the Calendar with the salon address.
- Platforms
mac· macOS- Applications
- Calendar · salon booking clone
- Interfaces
desk· Desktop GUIbrow· Browser- GUI need · task setting
G2· GUI-essential- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
multi· Cross-appvis· Visual judgmentclick· Point & click- Work product
WP08· App/system state- Conditions
multiapp· Multi-apppersonal· Personal context- Proposed difficulty
- 4 / 5
Proposed verifier
Salon DB booking with Rosa, service duration 150, start = earliest slot fully inside a seeded free gap; Calendar event with same times and location. The salon clone issues no API token and books only from the browser session (CSRF); the agent has no Terminal or osascript, so the event must be added in the Calendar app.
Proposed partial credit
40% slot fits calendar gap and is earliest; 20% stylist/service; 30% calendar event; 10% address.
R12.04.01.02-T07 Book a doctor's appointment in a hospital mini-program on the first weekday morning
Original instruction
In the super-app clone (Android emulator), open the 'City Hospital' mini-program and book a dermatology appointment with any attending physician on the first available weekday morning, for the patient profile 'Li Wei'. Pay the registration fee step must be stopped before password entry.
- Platforms
and· Android- Applications
- super-app clone · hospital mini-program
- Interfaces
mob· Mobile touch- GUI need · task setting
G3· GUI-only- GUI need · ordinary access
G3· GUI-only- Interaction primitives
touch· Touch gesturesvis· Visual judgment- Work product
WP08· App/system state- Conditions
risk· Irreversible/risky- Proposed difficulty
- 3 / 5
Proposed verifier
Mini-program backend: appointment dept=dermatology, physician title=attending, slot = earliest weekday AM, patient Li Wei, status=awaiting_payment. The agent gets screen input only (no adb shell, no backend API token), so the booking must be made in the mini-program; the harness input log confirms UI taps.
Proposed partial credit
30% department; 20% physician level; 40% earliest weekday AM; 10% stop at payment.
R12.04.01.02-T08 Book a car service appointment with a loaner and drop-off time
Original instruction
On the dealership service-booking clone, book a 60,000-mile service for the vehicle with VIN in fixtures/registration.pdf, add the recall check listed for that VIN on the recall lookup page, request a loaner car, and pick the earliest drop-off slot on a weekday between 07:30-09:00 next week.
- Platforms
web· Any OS (web)- Applications
- dealership booking clone · PDF viewer
- Interfaces
brow· Browser- GUI need · task setting
G2· GUI-essential- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
multi· Cross-apptype· Text & formsclick· Point & click- Work product
WP08· App/system state- Conditions
multiapp· Multi-app- Proposed difficulty
- 4 / 5
Proposed verifier
Booking: VIN exact, service package 60k, recall code from lookup added, loaner=true (limited loaner availability changes earliest slot), slot earliest feasible with loaner. The clone issues no API token and accepts bookings only from browser sessions (CSRF + session id), so scripted requests are rejected.
Proposed partial credit
25% VIN; 25% recall; 20% loaner; 30% slot.
R12.04.01.02-T09 Cancel a duplicate reservation keeping the one with the occasion note
Original instruction
On the OpenDining clone, the user has two reservations at 'Sakura House' for Saturday (19:00 and 19:15), created by mistake. Keep the one with the 'Birthday' occasion note and cancel the other; then add 'cake will be brought' to the kept reservation's notes.
- Platforms
web· Any OS (web)- Applications
- OpenDining clone
- Interfaces
brow· Browser- GUI need · task setting
G2· GUI-essential- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
click· Point & clicktype· Text & forms- Work product
WP08· App/system state- Conditions
risk· Irreversible/risky- Proposed difficulty
- 1 / 5
Proposed verifier
One active reservation (the Birthday one) with appended note; the other cancelled; no other reservations touched. The clone issues no API token and accepts cancellations and note edits only from the browser session (CSRF + session id), so scripted calls are rejected.
Proposed partial credit
50% correct one cancelled; 30% note appended; 20% nothing else touched.
R12.04.01.02-T10 Book a multi-resource appointment: massage room and therapist with gender preference
Original instruction
On the spa booking clone, book a 90-minute couples massage for Sunday afternoon (13:00-17:00) requiring a couples room and two therapists, one female per the user's preference for herself. Add a 30-minute sauna add-on immediately after. Pay the deposit with the sandbox card.
- Platforms
web· Any OS (web)- Applications
- spa booking clone
- Interfaces
brow· Browser- GUI need · task setting
G2· GUI-essential- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
click· Point & clickvis· Visual judgment- Work product
WP08· App/system state- Conditions
risk· Irreversible/risky- Proposed difficulty
- 4 / 5
Proposed verifier
Booking: couples room resource, 2 therapists (>=1 female assigned to user), start in window, sauna add-on contiguous after, deposit paid. The clone issues no API token and accepts bookings only from browser sessions (CSRF + session id), so scripted requests are rejected.
Proposed partial credit
30% room + 2 therapists; 20% gender preference; 30% contiguous add-on; 20% deposit.
R12.04.01.02-T11 Join a restaurant waitlist and accept a table within a time limit
Original instruction
On the waitlist app clone (iOS Simulator), join the waitlist for 3 at 'Noodle Bar' with the user's phone. When the 'Your table is ready' prompt appears (simulated after a few minutes), accept within 2 minutes; if it offers bar seating instead of a table, decline and stay on the list.
- Platforms
ios· iOS/iPadOS- Applications
- waitlist app clone (iOS)
- Interfaces
mob· Mobile touch- GUI need · task setting
G2· GUI-essential- GUI need · ordinary access
G3· GUI-only- Interaction primitives
touch· Touch gesturesvis· Visual judgment- Work product
WP08· App/system state- Conditions
interrupt· Interruptions- Proposed difficulty
- 3 / 5
Proposed verifier
Event log: first offer (bar) declined while staying on list; second offer (table) accepted within 120 s; final state seated/confirmed for 3. The agent has no host shell or simctl (no push or URL-scheme injection), and the clone backend accepts accept/decline only from the app session.
Proposed partial credit
40% bar offer declined but stayed listed; 40% table accepted in time; 20% party size.
R12.04.01.02-T12 Schedule a recurring physiotherapy series around work meetings
Original instruction
Book 6 weekly physiotherapy sessions (45 min) with therapist 'Dr. Kim' on the clinic portal clone starting next week, all at the same weekday and time if possible, avoiding conflicts with the user's Outlook calendar (exported to fixtures/work.ics). If no single weekday/time works for all 6, use the fewest distinct times.
- Platforms
mac· macOSwin· Windowsweb· Any OS (web)- Applications
- clinic portal clone · Outlook
- Interfaces
brow· Browser- GUI need · task setting
G2· GUI-essential- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
multi· Cross-appvis· Visual judgmentclick· Point & click- Work product
WP08· App/system state- Conditions
multiapp· Multi-applong· Long-horizon- Proposed difficulty
- 4 / 5
Proposed verifier
6 bookings with Dr. Kim, no overlap with ics events, consecutive weeks, number of distinct (weekday,time) pairs equals seeded minimum (2). The portal issues no API token and accepts bookings only from the browser session (CSRF + session id), so scripted bookings are rejected.
Proposed partial credit
40% 6 sessions conflict-free; 40% minimal distinct times; 20% consecutive weeks.
R12.04.01.02-T13 Reserve a campsite meeting site attributes on a map-based reservation system
Original instruction
On the national-park campsite reservation clone, reserve a tent site for 2 nights (Fri-Sat, 3 weeks out) at 'Pine Lake' that is within 100 m of a restroom, not in the river floodplain area (shaded on the map), and has a fire ring. Sites are chosen by clicking on the campground map.
- Platforms
web· Any OS (web)- Applications
- campsite reservation clone
- Interfaces
brow· Browser- GUI need · task setting
G2· GUI-essential- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
nav· Pan/zoom/scrollvis· Visual judgmentclick· Point & click- Work product
WP08· App/system state- Conditions
risk· Irreversible/risky- Proposed difficulty
- 4 / 5
Proposed verifier
Reservation site id in the feasible set (distance to restroom computed from map coordinates, not in floodplain polygon, fire_ring attribute); dates correct. The clone issues no API token and accepts bookings only from browser sessions (CSRF + session id), so scripted requests are rejected.
Proposed partial credit
30% restroom distance; 30% floodplain avoided; 20% fire ring; 20% dates.
R12.04.01.02-T14 Book a tasting-menu dinner requiring a prepaid deposit and dietary form
Original instruction
On the Resy-style clone, book 'Atelier 9' tasting menu for 2 on any Thursday next month at 19:30 or 20:00, complete the dietary form (one pescatarian, one no dairy), and stop at the prepayment page for the $150 deposit, handing back to the user.
- Platforms
web· Any OS (web)- Applications
- Resy-style clone
- Interfaces
brow· Browser- GUI need · task setting
G2· GUI-essential- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
click· Point & clicknav· Pan/zoom/scrolltype· Text & forms- Work product
WP08· App/system state- Conditions
risk· Irreversible/risky- Proposed difficulty
- 2 / 5
Proposed verifier
Reservation hold: Thursday slot in allowed times, party 2, dietary form fields exact, status=pending_payment; ASK_USER present; no card fields typed. The clone issues no API token and accepts bookings only from browser sessions (CSRF + session id), so scripted requests are rejected.
Proposed partial credit
40% slot; 30% dietary form; 30% stopped with handoff.
R12.04.01.02-T15 Book a passport-photo and passport-application appointment combo at the post office
Original instruction
On the post-office appointment clone, book a passport-application appointment for 2 applicants (adult + minor; minor requires both parents present per the info page) and a passport-photo service in the same visit at the nearest office to ZIP 60614 with availability within 14 days. Pick a slot long enough for both services.
- Platforms
web· Any OS (web)- Applications
- post office appointment clone
- Interfaces
brow· Browser- GUI need · task setting
G2· GUI-essential- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
click· Point & clicknav· Pan/zoom/scrollvis· Visual judgment- Work product
WP08· App/system state- Conditions
- None declared
- Proposed difficulty
- 3 / 5
Proposed verifier
Appointment: office = nearest with availability (seeded), services both, applicants 2 with minor flag, duration slot >= sum of service times, within 14 days; info acknowledgment checkbox for both parents. The clone issues no API token and accepts bookings only from browser sessions (CSRF + session id), so scripted requests are rejected.
Proposed partial credit
25% office; 25% services/applicants; 30% slot duration; 20% minor requirements.
R12.04.01.02-T16 Rebook a missed appointment while honoring a cancellation-fee policy
Original instruction
The user missed yesterday's dental cleaning. In the dental patient portal clone, read the no-show policy, reschedule within the 7-day window that waives the fee, choose the hygienist the user saw last time (from visit history), and acknowledge the policy. Report whether the fee was waived.
- Platforms
web· Any OS (web)- Applications
- dental portal clone
- Interfaces
brow· Browser- GUI need · task setting
G2· GUI-essential- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
click· Point & clicknav· Pan/zoom/scrollvis· Visual judgment- Work product
WP08· App/system state- Conditions
- None declared
- Proposed difficulty
- 3 / 5
Proposed verifier
New appointment within 7 days with the last-seen hygienist; no-show fee status=waived; policy acknowledgment flag; report matches. The clone issues no API token and accepts bookings only from browser sessions (CSRF + session id), so scripted requests are rejected.
Proposed partial credit
40% within window; 30% correct hygienist; 20% fee waived; 10% report.
R12.04.01.02-T17 Coordinate a group dinner booking via a poll and then reserve
Original instruction
Create a date poll in the scheduling-poll clone (Doodle-style) with 3 options next week for a dinner of 8, share it to the group chat clone, wait for 7 scripted responses, then book the most-available option at 'Casa Verde' on the OpenDining clone (fall back to the second-best option if no table). Post the confirmation in the chat.
- Platforms
web· Any OS (web)- Applications
- poll clone · chat clone · OpenDining clone
- Interfaces
brow· Browser- GUI need · task setting
G2· GUI-essential- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
multi· Cross-appclick· Point & clicktype· Text & forms- Work product
WP08· App/system state- Conditions
multiapp· Multi-applong· Long-horizon- Proposed difficulty
- 4 / 5
Proposed verifier
Poll with 3 options exists; booking date = highest-yes option with table available (seeded: best option has no table → second); party 8; chat confirmation includes date/time. None of the clones issues an API token; polls, chat posts and bookings are accepted only from browser sessions (CSRF + session id), so scripted calls are rejected.
Proposed partial credit
20% poll created/shared; 50% correct booked option with fallback; 20% party size; 10% chat post.
R12.04.01.02-T18 Book a kids' swim-lesson series with age and level eligibility
Original instruction
On the city recreation booking clone, enroll two children (ages 6 and 9, levels from last term's report cards in fixtures/) into swim lessons at the same pool on the same day, overlapping or back-to-back, in the fall session. Classes enforce age and level eligibility. Add both to one cart and stop before payment.
- Platforms
web· Any OS (web)- Applications
- city recreation booking clone
- Interfaces
brow· Browser- GUI need · task setting
G2· GUI-essential- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
vis· Visual judgmentclick· Point & clicknav· Pan/zoom/scroll- Work product
WP08· App/system state- Conditions
multiapp· Multi-app- Proposed difficulty
- 4 / 5
Proposed verifier
Cart: 2 enrollments in eligible classes (age/level from report cards), same pool & day, times overlapping or adjacent (gap <= 15 min); no order placed. The clone issues no API token and accepts bookings only from browser sessions (CSRF + session id), so scripted requests are rejected.
Proposed partial credit
40% eligibility; 40% schedule alignment; 20% cart/no payment.
R12.04.01.02-T19 Book an appointment in a booking widget embedded in an iframe with time-zone shift
Original instruction
On the consultant's website clone, the embedded Calendly-style widget shows slots in the browser's time zone (America/Los_Angeles). Book a 30-minute call for the user, who is in Berlin, at 16:00 Berlin time on the first available day this week, and fill the intake questions from intake.txt. Verify the confirmation email in the mail sink.
- Platforms
web· Any OS (web)- Applications
- Calendly-style widget clone
- Interfaces
brow· Browser- GUI need · task setting
G2· GUI-essential- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
click· Point & clicktype· Text & formsvis· Visual judgment- Work product
WP08· App/system state- Conditions
- None declared
- Proposed difficulty
- 3 / 5
Proposed verifier
Booking UTC start equals 16:00 Europe/Berlin on the first day with that slot; intake answers exact; confirmation email present in MailHog. Changing widget time zone to Berlin is allowed. The widget issues no API token and accepts bookings only from the browser session (CSRF + session id), so scripted bookings are rejected.
Proposed partial credit
50% correct time instant; 30% intake; 20% email verified.
R12.04.01.02-T20 Plan and book a full Saturday of appointments across three providers
Original instruction
For Saturday, book: a car wash (45 min) at 'SparkleWash', a haircut (30 min) at the salon clone, and a vet check (20 min) at the vet clinic clone, so that travel times between locations (from the map clone) fit, all within 09:00-14:00, minimizing total idle time. Add all three to the user's Google Calendar clone with addresses.
- Platforms
web· Any OS (web)- Applications
- map web clone · salon booking clone · vet clinic clone · calendar clone
- Interfaces
brow· Browser- GUI need · task setting
G2· GUI-essential- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
multi· Cross-appnav· Pan/zoom/scrollclick· Point & click- Work product
WP08· App/system state- Conditions
multiapp· Multi-applong· Long-horizon- Proposed difficulty
- 5 / 5
Proposed verifier
Three bookings feasible with router travel times; within window; idle time equals seeded optimum (±5 min); 3 calendar events matching bookings. None of the clones issues an API token; bookings and events are accepted only from browser sessions (CSRF + session id), so scripted calls are rejected.
Proposed partial credit
40% three feasible bookings; 30% optimal idle time; 30% calendar events.
Mapped research references
Existing benchmark-to-class links from the research catalog. They are not record-level annotations, executions or product endorsements.