Job & rental applications with documents
Complete applications with uploads and consistent details.
20 project-authored proposals · taxonomy v1.1 · 0 new benchmark runs. Verifiers and partial-credit rules below are designs, not executed results.
- Typical applications
- job boards · rental portals
- Class GUI-need label
G3· GUI-only
Task proposals · 20
Expand to inspect the complete original wording and proposed checks. Original proposal text remains English in both interfaces.
R12.08.01.02-T01 Apply to a job posting with resume upload and screening answers
Original instruction
On the Greenhouse-style ATS clone, apply to 'Data Analyst II (Austin)' using fixtures/resume.pdf (upload via the OS file picker), fill contact fields and the 5 screening questions from fixtures/profile.json, and attach cover_letter.pdf. Submit the application.
- Platforms
mac· macOSwin· Windowsweb· Any OS (web)- Applications
- Greenhouse-style ATS clone
- Interfaces
brow· Browser- GUI need · task setting
G3· GUI-only- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
type· Text & formsclick· Point & clickmulti· Cross-app- Work product
WP08· App/system state- Conditions
risk· Irreversible/risky- Proposed difficulty
- 2 / 5
Proposed verifier
ATS DB: application for job id, contact fields exact, 5 answers equal profile, resume and cover letter hashes; single application. File input accepts only native picker (drag/JS set disabled).
Proposed partial credit
30% contact; 40% screening answers; 30% attachments.
R12.08.01.02-T02 Fill a rental application from a dossier and upload a pay stub
Original instruction
On the Zillow-style rental portal clone, apply for listing #R-1180 (1-bedroom) using fixtures/dossier.json (employment, income, rental history with 2 prior landlords, references) and upload the latest pay stub from fixtures/stubs/ (pick the most recent by pay date, not by filename).
- Platforms
web· Any OS (web)- Applications
- rental portal clone
- Interfaces
brow· Browser- GUI need · task setting
G3· GUI-only- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
type· Text & formsmulti· Cross-appclick· Point & click- Work product
WP08· App/system state- Conditions
risk· Irreversible/risky- Proposed difficulty
- 2 / 5
Proposed verifier
Application payload: fields match dossier (per field), 2 landlord entries ordered most recent first, pay stub = most recent pay date (hash). The clone issues no API token and accepts writes only from browser sessions (CSRF + session id), so scripted requests are rejected.
Proposed partial credit
50% dossier fields; 20% rental history; 30% correct pay stub.
R12.08.01.02-T03 Apply to three newest matching freelance jobs with tailored cover notes
Original instruction
On the TopWork-style freelance clone, find the 3 newest Python jobs with budget >= $1,000 and apply to each with resume.pdf attached and a cover note <= 150 words that names the job title and mentions one requirement from that posting. Bid each job's stated budget.
- Platforms
web· Any OS (web)- Applications
- freelance marketplace clone
- Interfaces
brow· Browser- GUI need · task setting
G3· GUI-only- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
nav· Pan/zoom/scrolltype· Text & formsmulti· Cross-app- Work product
WP08· App/system state- Conditions
- None declared
- Proposed difficulty
- 3 / 5
Proposed verifier
3 proposals on the correct job ids (filter/sort by posted date); each note <= 150 words, contains exact job title and one requirement keyword from that posting; bid equals budget; attachment hash. The clone issues no API token and accepts writes only from browser sessions (CSRF + session id), so scripted requests are rejected.
Proposed partial credit
40% correct jobs; 40% cover notes valid; 20% bids and attachments.
R12.08.01.02-T04 Complete a Workday-style multi-step application with work history parsed from a resume
Original instruction
On the Workday-style ATS clone, apply to 'Operations Manager' by uploading resume.docx (the parser fills fields with errors). Review every auto-filled section: fix the 3 wrong employment dates and a merged job entry, add the missing degree, answer voluntary disclosures with 'Decline to answer', and submit.
- Platforms
mac· macOSwin· Windowsweb· Any OS (web)- Applications
- Workday-style ATS clone
- Interfaces
brow· Browser- GUI need · task setting
G3· GUI-only- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
vis· Visual judgmenttype· Text & formsclick· Point & click- Work product
WP08· App/system state- Conditions
risk· Irreversible/riskylong· Long-horizon- Proposed difficulty
- 4 / 5
Proposed verifier
Submitted application: 4 employment entries matching resume (dates, titles), degree present, disclosures declined; seeded parser errors corrected. The clone issues no API token and accepts writes only from browser sessions (CSRF + session id), so scripted requests are rejected.
Proposed partial credit
50% employment entries; 20% education; 15% disclosures; 15% submitted.
R12.08.01.02-T05 Save matching job postings with salary and recency filters
Original instruction
On the job-board clone, find all 'Data Analyst' postings in Austin with salary >= $90k posted in the last 7 days (salary sometimes only in the description text) and save each to the 'Shortlist' folder. Do not apply.
- Platforms
web· Any OS (web)- Applications
- job board clone
- Interfaces
brow· Browser- GUI need · task setting
G3· GUI-only- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
nav· Pan/zoom/scrollclick· Point & clickvis· Visual judgment- Work product
WP08· App/system state- Conditions
- None declared
- Proposed difficulty
- 2 / 5
Proposed verifier
Saved jobs folder equals the seeded 3 postings (incl. one with salary only in description); no applications created. The clone issues no API token and accepts writes only from browser sessions (CSRF + session id), so scripted requests are rejected.
Proposed partial credit
80% correct saved set (F1); 20% no applications.
R12.08.01.02-T06 Update an existing application: new cover letter and phone number
Original instruction
On the ATS clone candidate portal, for application #A-33, replace the cover letter with fixtures/cover_letter_v2.pdf and update the phone number in the candidate profile to the one in fixtures/new_contact.txt. Do not withdraw or create a new application.
- Platforms
mac· macOSwin· Windowsweb· Any OS (web)- Applications
- ATS clone
- Interfaces
brow· Browser- GUI need · task setting
G3· GUI-only- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
click· Point & clickmulti· Cross-apptype· Text & forms- Work product
WP08· App/system state- Conditions
- None declared
- Proposed difficulty
- 3 / 5
Proposed verifier
Application A-33: cover letter attachment hash = v2 (v1 removed or superseded), status unchanged; candidate phone exact (E.164 normalized). The clone issues no API token and accepts writes only from browser sessions (CSRF + session id), so scripted requests are rejected.
Proposed partial credit
50% cover letter replaced; 30% phone; 20% no new/withdrawn applications.
R12.08.01.02-T07 Withdraw an application with the correct reason
Original instruction
On the ATS clone candidate portal, withdraw the application to 'Store Manager' (there are two Store Manager applications at different locations; withdraw the Riverside one), choosing reason 'Accepted another offer'. Confirm the withdrawal dialog.
- Platforms
web· Any OS (web)- Applications
- ATS clone
- Interfaces
brow· Browser- GUI need · task setting
G3· GUI-only- 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
- 1 / 5
Proposed verifier
Riverside application status WITHDRAWN with reason code; the other location's application unchanged. The clone issues no API token and accepts writes only from browser sessions (CSRF + session id), so scripted requests are rejected.
Proposed partial credit
60% correct application withdrawn; 30% reason; 10% other untouched.
R12.08.01.02-T08 Apply to a government job with screening questions from a profile and stop before submit
Original instruction
On the government jobs portal clone, apply to the 'Police Officer - Soldotna, AK' posting using resume.pdf; answer the 9 supplemental questions from profile.json (some require selecting from long dropdowns and date pickers), attach the DD-214 PDF, and stop before final submit.
- Platforms
web· Any OS (web)- Applications
- government jobs portal clone
- Interfaces
brow· Browser- GUI need · task setting
G3· GUI-only- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
type· Text & formsclick· Point & clickmulti· Cross-app- Work product
WP08· App/system state- Conditions
risk· Irreversible/risky- Proposed difficulty
- 3 / 5
Proposed verifier
Draft application with all 9 answers matching profile, attachments (resume + DD-214) in the correct slots, status=draft. The clone issues no API token and accepts writes only from browser sessions (CSRF + session id), so scripted requests are rejected.
Proposed partial credit
60% answers (equal weight per answer); 30% attachments; 10% not submitted.
R12.08.01.02-T09 Keep details consistent across a job application and a background-check form
Original instruction
After an offer, complete the background-check portal clone for the candidate using the same data as the submitted application (view it in the ATS clone): 7-year address history, employment, education. The ATS application had a typo in one employer's end date that HR corrected in a note; use the corrected date. Authorize with typed signature.
- Platforms
web· Any OS (web)- Applications
- ATS clone · background check portal clone
- Interfaces
brow· Browser- GUI need · task setting
G3· GUI-only- 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-apprisk· Irreversible/risky- Proposed difficulty
- 4 / 5
Proposed verifier
Background form fields equal ATS data except the corrected end date; address history covers 7 years without gaps; signature and authorization flag set. The clone issues no API token and accepts writes only from browser sessions (CSRF + session id), so scripted requests are rejected.
Proposed partial credit
40% address history; 30% employment incl. correction; 15% education; 15% authorization.
R12.08.01.02-T10 Apply to a rental with a co-applicant and guarantor invites
Original instruction
On the rental portal clone, start an application for listing #R-2204 for the user and co-applicant 'Jamie' (email in dossier), invite the guarantor (user's parent) to complete their section, pay the application fee with the test card, and upload a photo ID and the employment letter. Report the application ID.
- Platforms
web· Any OS (web)- Applications
- rental portal clone
- Interfaces
brow· Browser- GUI need · task setting
G3· GUI-only- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
type· Text & formsmulti· Cross-appclick· Point & click- Work product
WP08· App/system state- Conditions
risk· Irreversible/risky- Proposed difficulty
- 3 / 5
Proposed verifier
Application: co-applicant and guarantor invitations (emails exact) in mail sink; fee paid (sandbox); 2 attachments correct; reported ID matches. The clone issues no API token and accepts writes only from browser sessions (CSRF + session id), so scripted requests are rejected.
Proposed partial credit
55% co-applicant and guarantor invites; 20% fee; 15% attachments; 10% ID.
R12.08.01.02-T11 Tailor a resume in Word to a posting and submit it as PDF
Original instruction
In Microsoft Word, edit resume_master.docx to create resume_acme.docx tailored to the posting in fixtures/acme_posting.pdf: keep it to 1 page, reorder skills to put the 5 posting-matched skills first, remove the 'Hobbies' section, then export to PDF and apply on the ATS clone with it.
- Platforms
mac· macOSwin· Windows- Applications
- Microsoft Word · ATS clone
- Interfaces
desk· Desktop GUIbrow· Browser- GUI need · task setting
G3· GUI-only- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
drag· Drag & droptype· Text & formsmulti· Cross-app- Work product
WP01· Document- Conditions
multiapp· Multi-applicensed· Licensed pro app- Proposed difficulty
- 4 / 5
Proposed verifier
resume_acme.pdf: 1 page; skills list first 5 items = matched set; no 'Hobbies'; other text unchanged vs master (diff); ATS application attachment hash equals that PDF. The PDF must be saved by Word (producer metadata), so python-docx or converter output is detected, and the ATS clone accepts uploads only from the browser session (no API token).
Proposed partial credit
20% one page; 30% skills order; 30% hobbies removed and other text intact (half each); 20% applied with PDF.
R12.08.01.02-T12 Fill a long-form academic job application with references and documents
Original instruction
On the academic-jobs portal clone, apply to 'Assistant Professor, Computational Biology': upload CV, research statement, teaching statement and diversity statement (all in fixtures/apply/), add 3 referees from refs.json so that reference requests go out, and answer the eligibility questions. Submit.
- Platforms
mac· macOSwin· Windowsweb· Any OS (web)- Applications
- academic jobs portal clone
- Interfaces
brow· Browser- GUI need · task setting
G3· GUI-only- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
multi· Cross-apptype· Text & formsclick· Point & click- Work product
WP08· App/system state- Conditions
risk· Irreversible/risky- Proposed difficulty
- 3 / 5
Proposed verifier
Application: 4 documents in the correct slots (hash per slot), 3 referees exact with request emails in mail sink, eligibility answers; status submitted. The clone issues no API token and accepts writes only from browser sessions (CSRF + session id), so scripted requests are rejected.
Proposed partial credit
40% documents in slots; 35% referees + requests; 15% eligibility; 10% submitted.
R12.08.01.02-T13 Apply on a mobile job app with an in-app video answer recording
Original instruction
On the Android emulator, in the job-app clone, apply to 'Barista - Downtown' with the saved profile, attach resume.pdf from Files, and answer the one video question by recording a 20-30 s clip with the virtual camera (pre-scripted feed). Set availability to weekends only and submit.
- Platforms
and· Android- Applications
- job app clone · Files
- Interfaces
mob· Mobile touch- GUI need · task setting
G3· GUI-only- GUI need · ordinary access
G3· GUI-only- Interaction primitives
touch· Touch gesturesmulti· Cross-appav· Audio/video over time- Work product
WP08· App/system state- Conditions
risk· Irreversible/risky- Proposed difficulty
- 3 / 5
Proposed verifier
Application: resume attachment hash; video answer 20-30 s duration; availability mask Sat/Sun only; submitted. The agent gets screen input only (no adb shell, no backend API token), so the application must be made in the app; the harness input log confirms UI taps.
Proposed partial credit
30% resume; 40% video answer duration; 20% availability; 10% submitted.
R12.08.01.02-T14 Complete a rental application where the portal rejects a document format
Original instruction
On the rental portal clone, the bank statement fixtures/statement.heic is rejected (PDF/JPG only, <= 5 MB). Convert it to a JPG under 5 MB using Preview (macOS) or Photos (Windows) while keeping it legible, upload it with the other docs, and complete listing #R-3307's application from dossier.json.
- Platforms
mac· macOSwin· Windows- Applications
- rental portal clone · Preview · Photos
- Interfaces
brow· Browserdesk· Desktop GUI- GUI need · task setting
G3· GUI-only- 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-apprisk· Irreversible/risky- Proposed difficulty
- 3 / 5
Proposed verifier
Uploaded JPG: size <= 5 MB, OCR of account number and closing balance matches source; application fields equal dossier; submitted. The agent has no terminal, so sips or ImageMagick are not reachable for conversion; the portal issues no API token and accepts uploads only from the browser session.
Proposed partial credit
40% converted and legible upload; 40% fields; 20% submitted.
R12.08.01.02-T15 Detect inconsistent dates between resume and LinkedIn-style profile before applying
Original instruction
Before applying to 'Product Analyst' on the ATS clone, compare resume.pdf with the user's profile on the professional-network clone. Two job dates disagree. Ask the user which is correct (simulated reply: resume is correct), update the profile to match, then apply using the resume.
- Platforms
web· Any OS (web)- Applications
- professional network clone · ATS clone
- Interfaces
brow· Browser- GUI need · task setting
G3· GUI-only- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
multi· Cross-appvis· Visual judgmenttype· Text & forms- Work product
WP08· App/system state- Conditions
multiapp· Multi-appclarify· Ask when unclear- Proposed difficulty
- 4 / 5
Proposed verifier
ASK_USER names both discrepancies before any edits; profile dates updated to resume values; application submitted with consistent dates. Neither clone issues an API token; profile edits and applications are accepted only from browser sessions (CSRF + session id), so scripted requests are rejected.
Proposed partial credit
35% discrepancies found and asked; 35% profile corrected; 30% application consistent.
R12.08.01.02-T16 Schedule an interview slot and add it to the calendar from an ATS invite
Original instruction
An interview invite email (Thunderbird test inbox) links to the ATS scheduling page. Pick the earliest slot that does not conflict with the user's calendar (Thunderbird Lightning, local calendar) and is not on a Friday, confirm it, and add the event with the video link to the calendar.
- Platforms
mac· macOSwin· Windowslin· Linux- Applications
- Thunderbird · ATS scheduling clone
- Interfaces
desk· Desktop GUIbrow· Browser- GUI need · task setting
G3· GUI-only- 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
- 3 / 5
Proposed verifier
ATS booking at the earliest feasible slot (non-Friday, no conflicts); calendar event same time with the video URL in location/description. The ATS clone issues no API token and books only from the browser session; the agent has no terminal, so the Lightning calendar is not reachable for direct writes.
Proposed partial credit
50% correct slot; 30% calendar event; 20% video link.
R12.08.01.02-T17 Apply to university housing with roommate matching and document uploads
Original instruction
On the university housing portal clone, complete the housing application for fall: choose 3 ranked residence halls by preference list (max cost per term $5,200 from the rates table image), fill the roommate-matching questionnaire from prefs.json, request roommate by student ID, upload the immunization record, and pay the deposit with the test card.
- Platforms
web· Any OS (web)- Applications
- housing portal clone
- Interfaces
brow· Browser- GUI need · task setting
G3· GUI-only- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
drag· Drag & dropvis· Visual judgmenttype· Text & forms- Work product
WP08· App/system state- Conditions
risk· Irreversible/risky- Proposed difficulty
- 4 / 5
Proposed verifier
Application: ranked halls (drag-to-rank) are the 3 top preferences within cost cap; questionnaire answers equal prefs; roommate request ID; attachment; deposit paid. The clone issues no API token and accepts writes only from browser sessions (CSRF + session id), so scripted requests are rejected.
Proposed partial credit
30% ranking within cap; 30% questionnaire; 30% roommate request and document (half each); 10% deposit.
R12.08.01.02-T18 Batch-apply to five jobs with per-application custom questions and tracking sheet
Original instruction
Apply to the 5 postings listed in targets.xlsx across two ATS clones (Greenhouse-style and Lever-style). Each has 1-3 custom questions; answer from profile.json or, if not covered, ask the user once with all uncovered questions batched. Record application IDs and dates in targets.xlsx.
- Platforms
mac· macOSwin· Windows- Applications
- Greenhouse-style ATS clone · Lever-style ATS clone · Excel
- Interfaces
brow· Browserdesk· Desktop GUI- GUI need · task setting
G3· GUI-only- 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-applong· Long-horizonclarify· Ask when unclear- Proposed difficulty
- 5 / 5
Proposed verifier
5 applications with answers (profile-derived exact; user-provided from simulated reply); exactly one ASK_USER containing the 2 uncovered questions; targets.xlsx IDs/dates match ATS DBs. Neither ATS clone issues an API token; applications are accepted only from browser sessions (CSRF + session id), so scripted submissions are rejected.
Proposed partial credit
50% applications (equal weight per application); 20% single batched ask; 30% tracking sheet.
R12.08.01.02-T19 Respond to a landlord's request for more documents within a rental portal thread
Original instruction
The landlord on the rental portal clone requests, in the application message thread, (1) a reference letter from the previous landlord and (2) proof of renter's insurance quote. Find the reference letter in Mail (Thunderbird attachment), get a quote PDF from the insurance-quote clone using the dossier, upload both into the thread, and reply confirming.
- Platforms
mac· macOSwin· Windowslin· Linux- Applications
- rental portal clone · Thunderbird · insurance quote clone
- Interfaces
brow· Browserdesk· Desktop GUI- GUI need · task setting
G3· GUI-only- 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-applong· Long-horizon- Proposed difficulty
- 4 / 5
Proposed verifier
Thread contains 2 attachments: reference letter hash = email attachment; quote PDF generated for dossier address/coverage; reply message present. No binding policy purchased in quote clone. The web clones issue no API token and accept uploads and quotes only from browser sessions (CSRF + session id), so scripted requests are rejected.
Proposed partial credit
35% reference letter; 35% quote PDF correct; 15% reply; 15% no policy purchased.
R12.08.01.02-T20 Assemble a complete rental application packet across portals with consistent details
Original instruction
For listing #R-4410, assemble and submit: credit report request from the credit-report clone (soft pull, PDF), employment verification letter request via the HR portal clone, latest 3 pay stubs from the payroll clone, and the rental application on the portal with all docs and data consistent (name, SSN last 4, income). Stop at final submit for user review.
- Platforms
web· Any OS (web)- Applications
- rental portal clone · credit report clone · HR portal clone · payroll portal clone
- Interfaces
brow· Browser- GUI need · task setting
G3· GUI-only- 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-applong· Long-horizonrisk· Irreversible/risky- Proposed difficulty
- 5 / 5
Proposed verifier
Draft application with 5 attachments (hash-matched to each source portal output), monthly income = pay-stub-derived value, identity fields consistent across all; status=ready_for_submit; ASK_USER present; only soft pull logged. None of the clones issues an API token; documents and drafts are served only to browser sessions (CSRF + session id), so scripted requests are rejected.
Proposed partial credit
20% credit report (soft pull); 40% employment letter and pay stubs (half each); 25% consistent fields; 15% stopped for review.
Mapped research references
Existing benchmark-to-class links from the research catalog. They are not record-level annotations, executions or product endorsements.