Quick utilities: calculator, sticky notes, dictionary & conversions
Get quick answers and jot information with built-in utilities: calculator, sticky notes, dictionary/thesaurus, unit and currency conversion, screenshots of results.
20 project-authored proposals · taxonomy v1.1 · 0 new benchmark runs. Verifiers and partial-credit rules below are designs, not executed results.
- Typical applications
- Calculator · Stickies · Dictionary · Spotlight
- 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.
R04.06.01.03-T01 Jot three colored Stickies notes for a desk handoff
Original instruction
On the test Mac, Stickies has no notes. Create three notes: yellow 'Parking: Level 3, bay B14', green 'Courier pickup 15:30 - dock 2', pink 'Call IT re: badge 44817'. Make only the pink note Float on Top and Translucent.
- Platforms
mac· macOS- Applications
- Stickies
- Interfaces
desk· Desktop GUI- GUI need · task setting
G2· GUI-essential- GUI need · ordinary access
G3· GUI-only- Interaction primitives
click· Point & clicktype· Text & forms- Work product
WP08· App/system state- Conditions
- None declared
- Proposed difficulty
- 1 / 5
Proposed verifier
Grader quits Stickies and reads ~/Library/Containers/com.apple.Stickies/Data/Library/Stickies: exactly 3 .rtfd notes with exact text (whitespace-trimmed), colors and floating/translucent flags from .SavedStickiesState. macOS 14 app-container protection blocks other apps from writing that container and the agent session has no terminal, so notes must be made in Stickies.
Proposed partial credit
55% three notes with exact text; 25% colors; 20% float + translucent on pink only.
R04.06.01.03-T02 Total a receipt with Windows Calculator memory and pin it on top
Original instruction
fixtures/receipt_0412.png is open in Photos: 9 line items and a sales-tax rate of 8.875%. In Windows Calculator (Standard), add the items and save the subtotal with MS, then save the tax (rounded to the cent) and the grand total as further memory entries. Leave Calculator open in Keep on top (compact overlay) mode.
- Platforms
win· Windows- Applications
- Calculator · Photos
- Interfaces
desk· Desktop GUI- GUI need · task setting
G3· GUI-only- GUI need · ordinary access
G3· GUI-only- Interaction primitives
click· Point & clicktype· Text & formsvis· Visual judgment- Work product
WP08· App/system state- Conditions
- None declared
- Proposed difficulty
- 2 / 5
Proposed verifier
Grader inspects the running Calculator via UI Automation: Memory list holds exactly subtotal, tax and total to the cent; window is the topmost compact overlay in Standard mode. Calculator has no API or CLI and the agent session has no terminal/PowerShell, so memory state can only be set through its UI; receipt values exist only in the image.
Proposed partial credit
40% subtotal in memory; 25% tax entry; 20% total entry; 15% keep-on-top overlay active.
R04.06.01.03-T03 Convert two parcels' dimensions with macOS Calculator's Convert mode
Original instruction
On macOS 15, fixtures/parcels.txt (open in TextEdit) lists 2 parcels as L x W x H in inches and weight in lb. Using Calculator's Convert mode (Length, Mass), convert all values to cm and kg rounded to 1 decimal and write 'P1: L x W x H cm, kg' style lines in a new blue Stickies note 'Parcels metric'. Leave Calculator in Convert mode set to Mass lb to kg.
- Platforms
mac· macOS- Applications
- Calculator · Stickies · TextEdit
- Interfaces
desk· Desktop GUI- GUI need · task setting
G3· GUI-only- GUI need · ordinary access
G3· GUI-only- Interaction primitives
click· Point & clicktype· Text & formsmulti· Cross-app- Work product
WP12· Answer/decision- Conditions
multiapp· Multi-app- Proposed difficulty
- 2 / 5
Proposed verifier
Note read from the Stickies container (.rtfd text + .SavedStickiesState) after the grader quits Stickies: title, blue color, 8 values within ±0.05 of references. Calculator state read via the Accessibility API: Convert on, category Mass, units lb to kg. Calculator has no AppleScript dictionary and the agent session has no terminal, so its state is set only in the GUI.
Proposed partial credit
60% eight converted values (pro-rata); 15% note title and color; 25% Calculator left in Convert mode lb to kg.
R04.06.01.03-T04 Extract a register field with Calculator Programmer-mode masking and bit toggles
Original instruction
fixtures/dma_status.txt gives DMA status word 0x8F3A19C4; bits 12-23 hold the transfer count. In Windows Calculator Programmer mode (DWORD), first clear bit 31 of the word using the bit-toggling keypad, then from the original word AND the field mask and shift right by 12, leaving the field value displayed. Record both results in hex and decimal in a new Sticky Note 'DMA'.
- Platforms
win· Windows- Applications
- Calculator · Sticky Notes
- Interfaces
desk· Desktop GUI- GUI need · task setting
G3· GUI-only- GUI need · ordinary access
G3· GUI-only- Interaction primitives
click· Point & clicktype· Text & forms- Work product
WP12· Answer/decision- Conditions
multiapp· Multi-app- Proposed difficulty
- 3 / 5
Proposed verifier
Grader reads running Calculator via UI Automation: Programmer mode, DWORD, HEX/DEC readouts equal the field value. plum.sqlite (after Sticky Notes closes) has note 'DMA' with both results in hex and decimal. Calculator exposes no API and the agent session has no terminal/PowerShell, so readouts and word size can only come from UI input.
Proposed partial credit
35% field value displayed in Programmer DWORD; 25% bit-31-cleared word correct in note; 40% note has both results in hex and decimal.
R04.06.01.03-T05 Compute contract end dates with Windows Calculator date calculation
Original instruction
fixtures/contracts.csv lists 6 service contracts with start date and term (months or days; one starts on Jan 31). Using Windows Calculator's Date calculation mode, find each end date (start + term, minus 1 day) and the days from the VM's fixed date 2026-10-14 to it. Create a Sticky Note 'Renewals' with lines 'ID | YYYY-MM-DD | days left', sorted by days left.
- Platforms
win· Windows- Applications
- Calculator · Sticky Notes
- Interfaces
desk· Desktop GUI- GUI need · task setting
G2· GUI-essential- GUI need · ordinary access
G3· GUI-only- Interaction primitives
click· Point & clicktype· Text & formsmulti· Cross-app- Work product
WP12· Answer/decision- Conditions
multiapp· Multi-app- Proposed difficulty
- 3 / 5
Proposed verifier
Grader reads plum.sqlite after Sticky Notes closes: note 'Renewals', 6 lines whose end dates and day counts equal references computed with Calculator's month arithmetic (month-end clamping), correct sort. Agent session has no terminal/PowerShell or sqlite tool, and rows must carry Sticky Notes' own Id/CreatedAt fields, so direct DB writes are detected.
Proposed partial credit
50% end dates (pro-rata over 6); 30% days-left values; 20% title, line format and sort order.
R04.06.01.03-T06 Convert a multi-currency expense list with Calculator's cached exchange rates
Original instruction
fixtures/trip_expenses.csv has 10 expenses in EUR, JPY, GBP and CHF. The VM is offline, so Windows Calculator's Currency converter uses its cached rates; do not press 'Update rates'. Convert each amount to USD with that converter, round to cents, and create a Sticky Note 'Trip USD' with lines 'row#: amount USD' and a final 'TOTAL: x USD'.
- Platforms
win· Windows- Applications
- Calculator · Sticky Notes
- Interfaces
desk· Desktop GUI- GUI need · task setting
G2· GUI-essential- GUI need · ordinary access
G3· GUI-only- Interaction primitives
click· Point & clicktype· Text & formsmulti· Cross-app- Work product
WP12· Answer/decision- Conditions
multiapp· Multi-app- Proposed difficulty
- 3 / 5
Proposed verifier
Before the episode the grader records each pair's cached rate from the converter footer via UI Automation; lines must match within ±0.01 USD and TOTAL within ±0.05, read from plum.sqlite after app close. No network and no terminal/PowerShell in the agent session, so the rates are reachable only through the converter UI.
Proposed partial credit
60% per-row USD amounts (pro-rata over 10); 20% TOTAL line; 20% note title and line format.
R04.06.01.03-T07 Arrange four Stickies notes into a 2x2 grid with colors and flags
Original instruction
Main display is 1440x900 points. Stickies has 4 notes: 'Today', 'Waiting on', 'Ideas', 'Numbers'. Resize each to 320x380 pt and place their top-left corners at (740,50), (1100,50), (740,450), (1100,450) in that order. Colors: Today yellow, Waiting on pink, Ideas green, Numbers blue. Make 'Numbers' translucent and 'Today' float on top. Do not edit note text.
- Platforms
mac· macOS- Applications
- Stickies
- Interfaces
desk· Desktop GUI- GUI need · task setting
G2· GUI-essential- GUI need · ordinary access
G3· GUI-only- Interaction primitives
drag· Drag & dropgeom· Precise geometryclick· Point & click- Work product
WP08· App/system state- Conditions
- None declared
- Proposed difficulty
- 3 / 5
Proposed verifier
Grader quits Stickies and reads .SavedStickiesState: frames converted to top-left points within ±12 pt, colors, translucent/floating flags; .rtfd texts byte-identical. Frames are only written by Stickies itself; app-container protection blocks other processes from editing the container and the agent has no terminal.
Proposed partial credit
50% four frames within tolerance (pro-rata); 25% colors; 15% translucent and float flags; 10% texts unchanged.
R04.06.01.03-T08 Triage a cluttered Sticky Notes board: recolor, format and delete finished notes
Original instruction
Sticky Notes (local, signed out) holds 14 notes seeded from fixtures/plum_seed.sqlite; 4 are closed and appear only in the notes list. Recolor notes starting 'URGENT' to Pink and 'IDEA' to Purple; in 'Release checklist' bold the first line and strike through the 3 items marked [x]; delete the 3 notes ending 'DONE', confirming each. Change nothing else.
- Platforms
win· Windows- Applications
- Sticky Notes
- Interfaces
desk· Desktop GUI- GUI need · task setting
G2· GUI-essential- GUI need · ordinary access
G3· GUI-only- Interaction primitives
click· Point & clickkey· Shortcutstype· Text & forms- Work product
WP08· App/system state- Conditions
risk· Irreversible/riskyhidden· Hidden/stale state- Proposed difficulty
- 3 / 5
Proposed verifier
Grader closes the app and reads plum.sqlite: Theme per note, exactly the 3 DONE notes deleted, 11 others with unchanged text; it reopens 'Release checklist' and reads bold/strikethrough ranges via UI Automation TextPattern. Agent session has no terminal, PowerShell or sqlite tool, so edits must go through the app.
Proposed partial credit
30% recolors incl. closed notes; 30% bold line and 3 strikethroughs; 30% exactly the 3 DONE notes deleted; 10% no collateral changes.
R04.06.01.03-T09 Set Dictionary sources and their order for bilingual editing
Original instruction
An editor wants Dictionary.app to show only these sources, in this order: Oxford American Writer's Thesaurus, New Oxford American Dictionary, French - English (Oxford-Hachette), Apple Dictionary. Wikipedia and every other source must be off. Configure it in Dictionary > Settings (reorder rows by dragging), then look up 'moribund' in a new window.
- Platforms
mac· macOS- Applications
- Dictionary
- Interfaces
desk· Desktop GUI- GUI need · task setting
G2· GUI-essential- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
drag· Drag & dropclick· Point & click- Work product
WP08· App/system state- Conditions
- None declared
- Proposed difficulty
- 3 / 5
Proposed verifier
Grader quits Dictionary and reads DCSActiveDictionaries from com.apple.DictionaryServices preferences: exact ordered list of the 4 bundles, nothing else; then checks the source tab order in a fresh window via the Accessibility API. The agent session has no terminal, so 'defaults write' is unavailable and order must come from the Settings list.
Proposed partial credit
35% the 4 sources enabled; 40% exact order; 25% Wikipedia and all other sources disabled.
R04.06.01.03-T10 Insert exact Unicode lab symbols into Stickies with Character Viewer
Original instruction
input/symbol_spec.csv lists 10 characters by Unicode name (e.g., MICRO SIGN, DEGREE CELSIUS, INCREMENT, PER MILLE SIGN, RIGHTWARDS HARPOON OVER LEFTWARDS HARPOON). Create a purple Stickies note titled 'Lab symbols' with one line per row 'NAME: <char>', finding each character via Character Viewer search. Then add the first 4 characters to Character Viewer's Favorites.
- Platforms
mac· macOS- Applications
- Character Viewer · Stickies
- Interfaces
desk· Desktop GUI- GUI need · task setting
G2· GUI-essential- GUI need · ordinary access
G3· GUI-only- Interaction primitives
click· Point & clicktype· Text & formsnav· Pan/zoom/scroll- Work product
WP01· Document- Conditions
- None declared
- Proposed difficulty
- 3 / 5
Proposed verifier
Note read from the Stickies container (.rtfd text + .SavedStickiesState) after the grader quits Stickies; code points compared exactly with no normalization, so lookalikes (Greek mu for MICRO SIGN, Greek Delta for INCREMENT) fail. Favorites read from the open Character Viewer via the Accessibility API. Harness typing is ASCII-only and the agent has no terminal, so non-ASCII characters must come from Character Viewer.
Proposed partial credit
60% exact code points (pro-rata over 10); 15% note title, color and line format; 25% four favorites added.
R04.06.01.03-T11 Build a currency and key-symbol cheat sheet with the emoji panel and Character Map
Original instruction
Create a Sticky Note 'Currency & keys' with one item per line: the signs ₹ € ¥ ₩ ₺ ₽ inserted from the emoji panel's Symbols tab (Win+.), the shrug kaomoji from its kaomoji tab, and U+2318, U+21E7, U+238B and U+2325 copied from Character Map (Segoe UI Symbol, Advanced view, Go to Unicode), each followed by its 'U+XXXX' label.
- Platforms
win· Windows- Applications
- Emoji panel · Character Map · Sticky Notes
- Interfaces
desk· Desktop GUI- GUI need · task setting
G2· GUI-essential- GUI need · ordinary access
G3· GUI-only- Interaction primitives
click· Point & clickkey· Shortcutsmulti· Cross-app- Work product
WP01· Document- Conditions
- None declared
- Proposed difficulty
- 3 / 5
Proposed verifier
Grader reads plum.sqlite after app close: title plus 11 lines with exact code points and correct U+ labels. Harness typing is ASCII-only and the agent session has no terminal/PowerShell, so non-ASCII characters must come from the emoji panel or Character Map; clipboard-history sources are logged.
Proposed partial credit
35% six currency signs; 15% kaomoji; 40% four key symbols with correct U+ labels; 10% title and one-per-line layout.
R04.06.01.03-T12 Answer quick site questions with Spotlight or Windows Search and flag ambiguous units
Original instruction
fixtures/site_questions.txt has 9 questions from a site lead (e.g., '45 psi in kPa', 'sqrt(2)*3.6 m', '20 gallons in L', '2 tons in kg'). Answer each with the system search box only (Spotlight on macOS, Windows Search on Windows), 2 decimals, in a note 'Site answers' (Stickies / Sticky Notes) as 'Q#: value unit'. If a unit is ambiguous (US vs imperial gallon, short vs metric ton), write 'Q#: ASK - <reason>' instead.
- Platforms
mac· macOSwin· Windows- Applications
- Spotlight · Windows Search · Stickies · Sticky Notes
- Interfaces
desk· Desktop GUI- GUI need · task setting
G2· GUI-essential- GUI need · ordinary access
G3· GUI-only- Interaction primitives
key· Shortcutstype· Text & formsmulti· Cross-app- Work product
WP12· Answer/decision- Conditions
clarify· Ask when unclearmultiapp· Multi-app- Proposed difficulty
- 3 / 5
Proposed verifier
Grader reads the note store (Stickies container or plum.sqlite): 7 numeric answers within ±0.01 and exactly the 2 ambiguous items marked ASK with the right reason. Calculator apps, Terminal/PowerShell and Python are removed from the VM image, so the search box is the only calculator; the note must be written by the notes app.
Proposed partial credit
60% seven numeric answers (pro-rata); 30% both ambiguous items flagged ASK with the correct reason; 10% note title and format.
R04.06.01.03-T13 Work a loan quote on macOS Calculator's Paper Tape and save the tape
Original instruction
On a macOS 14 test Mac, fixtures/loan_quote.txt asks for the monthly payment on $38,500 at 6.24% APR over 60 months, plus total interest. In Calculator's Scientific mode compute P*r/(1-(1+r)^-n) with r = APR/12, show the Paper Tape and save it with File > Save Tape As to ~/Desktop/loan_tape.txt. Put payment and total interest (2 dp) in a yellow Stickies note.
- Platforms
mac· macOS- Applications
- Calculator · Stickies
- Interfaces
desk· Desktop GUI- GUI need · task setting
G3· GUI-only- GUI need · ordinary access
G3· GUI-only- Interaction primitives
click· Point & clicktype· Text & formskey· Shortcuts- Work product
WP12· Answer/decision- Conditions
multiapp· Multi-app- Proposed difficulty
- 4 / 5
Proposed verifier
Grader parses loan_tape.txt: includes a power step and a line equal to the payment ±0.01; Stickies values ±0.01. File-system event provenance must attribute the tape write to Calculator; TextEdit is removed and the agent session has no terminal, so a forged tape is detected.
Proposed partial credit
40% tape saved by Calculator with a power step; 35% monthly payment correct; 25% total interest correct in the note.
R04.06.01.03-T14 Build a live shared-flat budget in Calculator Math Notes with variables
Original instruction
On macOS 15, open Calculator's Math Notes and create note 'Flat budget' from fixtures/flat_costs.txt: define variables rent, power, internet and water (monthly), then result lines ending in '=' for the per-person share among 3 people, the annual total, and Mia's monthly share if she pays 40 percent of rent plus a third of the other costs. Results must reference the variables, not hard-coded numbers.
- Platforms
mac· macOS- Applications
- Calculator · Notes
- Interfaces
desk· Desktop GUI- GUI need · task setting
G2· GUI-essential- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
type· Text & formsclick· Point & click- Work product
WP01· Document- Conditions
- None declared
- Proposed difficulty
- 4 / 5
Proposed verifier
Grader decodes the note from NoteStore.sqlite (must be in the Math Notes folder), checks the 4 assignments and that result lines reference variables, reads rendered results via the Accessibility API, then edits rent through AX and re-checks results. Notes automation (osascript) is denied to the agent session, which has no terminal.
Proposed partial credit
30% four variables equal the fixture; 40% per-person, annual and Mia results correct; 30% results update after the grader changes rent.
R04.06.01.03-T15 Find where two cost curves cross using Windows Calculator's Graphing mode
Original instruction
fixtures/cost_models.txt gives f(x)=0.04x^2+3x+120 and g(x)=18x+15 (x in hundreds of units). In Windows Calculator's Graphing mode plot f, g and h(x)=f(x)-g(x); set the window to x 0..400, y -200..7000 and make g dashed. Use Analyze function on h to find its x-intercepts and record both crossings as (x, f(x)) to 3 decimals in a Sticky Note 'Break-even'.
- Platforms
win· Windows- Applications
- Calculator · Sticky Notes
- Interfaces
desk· Desktop GUI- GUI need · task setting
G3· GUI-only- GUI need · ordinary access
G3· GUI-only- Interaction primitives
click· Point & clicktype· Text & formsvis· Visual judgmentnav· Pan/zoom/scroll- Work product
WP12· Answer/decision- Conditions
multiapp· Multi-app- Proposed difficulty
- 4 / 5
Proposed verifier
Grader inspects the running Calculator via UI Automation: the three equations, graph-options window bounds and g's line style; plum.sqlite note values within ±0.001 (x) and ±0.01 (f). Calculator has no API and the agent session has no terminal/PowerShell, so equations, styles and window can only be set in its UI.
Proposed partial credit
30% three functions plotted; 20% window bounds and dashed g; 50% both crossings correct in the note (pro-rata).
R04.06.01.03-T16 Record dictionary sense counts and thesaurus synonyms for a style sheet
Original instruction
input/style_terms.txt lists 10 words (e.g., temper, buttress, quixotic). For each, using Dictionary.app's installed New Oxford American Dictionary and Oxford American Writer's Thesaurus, add a line to a green Stickies note 'Style terms': word | part of speech of the first entry | count of numbered senses in that entry | first three synonyms of the thesaurus's first sense.
- Platforms
mac· macOS- Applications
- Dictionary · Stickies
- Interfaces
desk· Desktop GUI- GUI need · task setting
G2· GUI-essential- GUI need · ordinary access
G3· GUI-only- Interaction primitives
type· Text & formsclick· Point & clicknav· Pan/zoom/scrollmulti· Cross-app- Work product
WP02· Records/table- Conditions
multiapp· Multi-app- Proposed difficulty
- 4 / 5
Proposed verifier
Grader derives references from the installed bundles on its side (DictionaryServices text) and compares per field; synonyms exact and ordered. The agent session has no terminal or Python, so DictionaryServices/pyobjc lookups are unavailable; the note is read from the Stickies container (.rtfd text + .SavedStickiesState) after the grader quits Stickies.
Proposed partial credit
20% parts of speech; 40% sense counts; 40% thesaurus synonyms (each pro-rata over 10 words).
R04.06.01.03-T17 Decode a little-endian IMU frame in macOS Calculator's Programmer mode
Original instruction
fixtures/imu_frame.txt holds 6 little-endian 16-bit two's-complement words as hex byte pairs (e.g., 'C4 FF'), scaled 0.061 mg/LSB. In macOS Calculator's Programmer mode byte-swap each word and read it as signed 16-bit; scale in Basic mode. Write 'axis: raw_dec, mg (3 dp)' for all 6 in a blue Stickies note, then save a window screenshot of Calculator showing the last word in binary as ~/Desktop/last_word.png.
- Platforms
mac· macOS- Applications
- Calculator · Stickies · Screenshot
- Interfaces
desk· Desktop GUI- GUI need · task setting
G2· GUI-essential- GUI need · ordinary access
G3· GUI-only- Interaction primitives
click· Point & clicktype· Text & formskey· Shortcutsmulti· Cross-app- Work product
WP12· Answer/decision- Conditions
multiapp· Multi-applong· Long-horizon- Proposed difficulty
- 5 / 5
Proposed verifier
Note read from the Stickies container (.rtfd text + .SavedStickiesState) after the grader quits Stickies: raw signed values exact, mg ±0.001. last_word.png size must match the Calculator window's AX bounds (shadow allowed), carry screencapture metadata, and OCR of its binary row must equal the last word. Agent session has no terminal or Python, so decoding goes through Calculator.
Proposed partial credit
50% six raw signed values (pro-rata); 30% scaled mg values; 20% window screenshot with matching binary readout.
R04.06.01.03-T18 Audit text contrast with Digital Color Meter and Calculator
Original instruction
input/mockup_v3.png (sRGB-tagged, flat colors) is open in Preview at 100% and shows 6 labeled text/background pairs. With Digital Color Meter set to 'Display in sRGB' and hex values, sample each pair's text and background, compute the WCAG 2.x contrast ratio with Calculator's Scientific mode, and record 'pair: #text #bg ratio PASS|FAIL' (ratio 2 dp, AA normal text 4.5:1) in a pink Stickies note.
- Platforms
mac· macOS- Applications
- Digital Color Meter · Calculator · Preview · Stickies
- Interfaces
desk· Desktop GUI- GUI need · task setting
G2· GUI-essential- GUI need · ordinary access
G3· GUI-only- Interaction primitives
hover· Hover & contextgeom· Precise geometryvis· Visual judgmentmulti· Cross-app- Work product
WP12· Answer/decision- Conditions
multiapp· Multi-applong· Long-horizon- Proposed difficulty
- 5 / 5
Proposed verifier
Grader knows the fixture's exact sRGB colors: hex within ±1 per channel, ratios ±0.02, PASS/FAIL exact; note read from the Stickies container (.rtfd text + .SavedStickiesState) after the grader quits Stickies. The agent has no terminal or image tooling and Preview shows no pixel values, so colors must be sampled with Digital Color Meter.
Proposed partial credit
35% twelve hex colors (pro-rata); 45% six contrast ratios; 20% PASS/FAIL verdicts.
R04.06.01.03-T19 Fill a chiller commissioning sheet with Windows Calculator converters
Original instruction
fixtures/chiller_readings.txt has 8 readings in psi, °F, BTU/h, US gal/min and ft. Using Windows Calculator only, convert each to kPa, °C, kW, L/s and m with Converter modes (chain steps where no direct unit exists), compute COP = cooling kW / input kW in Scientific mode, and post Sticky Note 'Chiller C-2' with each SI value (3 significant figures) and COP. Snip the Calculator window showing COP to Pictures\cop.png.
- Platforms
win· Windows- Applications
- Calculator · Sticky Notes · Snipping Tool
- Interfaces
desk· Desktop GUI- GUI need · task setting
G2· GUI-essential- GUI need · ordinary access
G3· GUI-only- Interaction primitives
click· Point & clicktype· Text & formsmulti· Cross-app- Work product
WP02· Records/table- Conditions
multiapp· Multi-applong· Long-horizon- Proposed difficulty
- 5 / 5
Proposed verifier
Grader reads plum.sqlite: 8 SI values within 3-sig-fig tolerance, COP ±0.01; Scientific History (UI Automation) contains the COP division; cop.png size matches Calculator's window rect and OCR shows COP. Calculator has no API and the agent session has no terminal/PowerShell, so values come from its UI.
Proposed partial credit
45% eight SI conversions (pro-rata); 25% COP in note and History; 15% note title and format; 15% window snip showing COP.
R04.06.01.03-T20 Identify homoglyphs in flagged sender names with Character Viewer
Original instruction
input/flagged_senders.txt (8 lines, open in TextEdit) holds synthetic sender names and domains that resemble 'examplebank.com' or 'contoso.com'; some use non-Latin lookalike letters. Paste suspect characters into Character Viewer to identify them, and in a yellow Stickies note write per line 'L#: pos N U+XXXX NAME' (N 1-based) for every non-ASCII character, or 'L#: CLEAN'.
- Platforms
mac· macOS- Applications
- Character Viewer · TextEdit · Stickies
- Interfaces
desk· Desktop GUI- GUI need · task setting
G2· GUI-essential- GUI need · ordinary access
G3· GUI-only- Interaction primitives
click· Point & clicktype· Text & formsvis· Visual judgmentmulti· Cross-app- Work product
WP12· Answer/decision- Conditions
multiapp· Multi-app- Proposed difficulty
- 4 / 5
Proposed verifier
Grader compares with the fixture's true code points: per-line (position, U+code) sets exact, names case-insensitive, CLEAN lines exact; note read from the Stickies container (.rtfd text + .SavedStickiesState) after the grader quits Stickies. The agent has no terminal or hex viewer and lookalikes are indistinguishable in screenshots, so identification needs Character Viewer's info pane.
Proposed partial credit
60% non-ASCII characters with positions and code points (pro-rata); 25% Unicode names; 15% CLEAN lines correct.
Mapped research references
Existing benchmark-to-class links from the research catalog. They are not record-level annotations, executions or product endorsements.