Computer-Use Agent Taxonomy中文
R04.04.01.02 · TASK CLASS

Slide layout, alignment & overflow repair

Fix overlapping, misaligned and overflowing objects so slides meet layout rules.

20 project-authored proposals · taxonomy v1.1 · 0 new benchmark runs. Verifiers and partial-credit rules below are designs, not executed results.

Typical applications
PowerPoint · Keynote · Google Slides
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.

R04.04.01.02-T01 Top-align and evenly distribute four icons on a slide

Original instruction

On slide 3 of input/services.pptx, four icon+label groups sit at uneven heights and spacing. Align their tops to the highest group, distribute them horizontally between the slide's content margins (1.0 in each side) with equal gaps, without resizing or ungrouping them. Save in place.

Platforms
mac · macOS win · Windows
Applications
PowerPoint
Interfaces
desk · Desktop GUI
GUI need · task setting
G2 · GUI-essential
GUI need · ordinary access
CL · CLI-sufficient
Interaction primitives
click · Point & click geom · Precise geometry drag · Drag & drop
Work product
WP03 · Visual layout
Conditions
None declared
Proposed difficulty
2 / 5

Proposed verifier

Group off/ext from slide XML: top y equal (±0.01 in), outer edges at 1.0 in margins (±0.02), gaps equal (±0.02); sizes unchanged; other shapes unchanged. Saved by PowerPoint (app.xml/core modified updated); VBA, COM and AppleScript automation disabled, no Python in the VM.

Proposed partial credit

35% tops aligned; 40% equal distribution within margins; 25% sizes/grouping unchanged.

R04.04.01.02-T02 Stop text overflow on three Google Slides without shrinking fonts

Original instruction

In Google Slides deck 'Partner onboarding', slides 4, 7 and 9 have body text running past the box and off the slide. Fix each by resizing or repositioning boxes within the slide's 0.5 in safe margins; font size must stay ≥18 pt and text must not be edited.

Platforms
web · Any OS (web)
Applications
Google Slides
Interfaces
brow · Browser
GUI need · task setting
G2 · GUI-essential
GUI need · ordinary access
G2 · GUI-essential
Interaction primitives
drag · Drag & drop geom · Precise geometry vis · Visual judgment
Work product
WP03 · Visual layout
Conditions
None declared
Proposed difficulty
2 / 5

Proposed verifier

Grader exports slides as PNG and PDF: text bboxes (pdfplumber) within box bounds and within safe margins; font sizes ≥18 pt via Slides API read; text content unchanged; no other slides modified (revision diff). Agent has no API token and Apps Script is disabled for the test account; revision history must show editor UI edits.

Proposed partial credit

75% three slides fixed (equal thirds); 15% fonts kept ≥18 pt; 10% text unchanged.

R04.04.01.02-T03 Snap drifted placeholders back to layout positions without losing formatting

Original instruction

In input/sales_kickoff.pptx, on 12 slides the title and content placeholders were dragged away from their layout positions and some were resized. Return each placeholder to its layout's position and size, while keeping the custom text formatting (Reset Slide would remove it, so do not lose run formatting). Save.

Platforms
mac · macOS win · Windows
Applications
PowerPoint
Interfaces
desk · Desktop GUI
GUI need · task setting
G2 · GUI-essential
GUI need · ordinary access
CL · CLI-sufficient
Interaction primitives
click · Point & click geom · Precise geometry
Work product
WP03 · Visual layout
Conditions
None declared
Proposed difficulty
3 / 5

Proposed verifier

For each placeholder: no local xfrm override or xfrm equal to layout (±0.01 in); run-level rPr identical to input (formatting preserved); text unchanged. Saved by PowerPoint; VBA, COM and AppleScript automation disabled, no Python in the VM.

Proposed partial credit

60% placeholders restored (pro-rata over 24); 30% run formatting preserved; 10% no other changes.

R04.04.01.02-T04 Align objects to a three-column guide system in Keynote

Original instruction

input/annual_review.key uses a 3-column layout, but on 6 slides images and text boxes are slightly off. Using Keynote's alignment guides and Arrange inspector, set each object's left edge to the nearest column start (x = 80, 700, 1320 pt) and tops to y = 260 pt, keeping sizes. Save.

Platforms
mac · macOS
Applications
Keynote
Interfaces
desk · Desktop GUI
GUI need · task setting
G2 · GUI-essential
GUI need · ordinary access
CL · CLI-sufficient
Interaction primitives
geom · Precise geometry drag · Drag & drop click · Point & click
Work product
WP03 · Visual layout
Conditions
None declared
Proposed difficulty
3 / 5

Proposed verifier

Grader AppleScript reads iWork item positions/sizes for the 6 slides: left edges equal column x (±1 pt), tops 260 (±1 pt), sizes unchanged; other slides untouched. osascript is blocked inside the agent session.

Proposed partial credit

50% left edges correct (pro-rata); 30% tops aligned; 20% sizes and other slides unchanged.

R04.04.01.02-T05 Untangle overlapping callouts on a bar chart slide

Original instruction

Slide 6 of input/market_share.pptx has five callout boxes overlapping each other and the bars. Reposition them so none overlap each other or any bar/axis label, each stays inside the chart area, and each callout's pointer ends on its own bar.

Platforms
mac · macOS win · Windows
Applications
PowerPoint
Interfaces
desk · Desktop GUI
GUI need · task setting
G2 · GUI-essential
GUI need · ordinary access
CL · CLI-sufficient
Interaction primitives
drag · Drag & drop geom · Precise geometry vis · Visual judgment
Work product
WP03 · Visual layout
Conditions
None declared
Proposed difficulty
3 / 5

Proposed verifier

Grader renders slide via PowerPoint export: callout bboxes pairwise IoU = 0 and no overlap with bar/label bboxes (from chart render); callout adjust handles put pointer tips inside the matched bar's bbox; all within chart area. Saved by PowerPoint. VBA, COM and AppleScript automation disabled, no Python in the VM.

Proposed partial credit

40% no overlaps; 40% pointers on correct bars (pro-rata over 5); 20% inside chart area.

R04.04.01.02-T06 Split an overflowing table across two slides with repeated headers

Original instruction

Slide 9 of input/vendor_eval.pptx has a 26-row table running off the bottom. Split it into two slides (9 and 10) of 13 data rows each, repeating the header row, same column widths and table style, font ≥12 pt, and both tables fully inside the slide's safe area. Renumber nothing else.

Platforms
mac · macOS win · Windows
Applications
PowerPoint
Interfaces
desk · Desktop GUI
GUI need · task setting
G2 · GUI-essential
GUI need · ordinary access
CL · CLI-sufficient
Interaction primitives
click · Point & click drag · Drag & drop geom · Precise geometry
Work product
WP03 · Visual layout
Conditions
None declared
Proposed difficulty
3 / 5

Proposed verifier

Slides 9 and 10 each have one table with header + 13 rows in original order; gridCol widths equal input; style id same; font ≥12 pt; rendered table bbox within safe area; later slides shifted by one and unchanged. VBA, COM and AppleScript automation disabled, no Python in the VM; file must be saved by PowerPoint (app.xml).

Proposed partial credit

35% rows split correctly; 20% header repeated; 25% inside safe area; 20% widths/style/fonts.

R04.04.01.02-T07 Arrange eight team photo cards into an even 4×2 grid

Original instruction

Slide 2 of Google Slides deck 'Meet the team' has 8 photos of different sizes and aspect ratios with name labels. Crop all photos to 1:1, size them 1.6 in square, arrange them in a 4×2 grid centered on the slide with equal 0.3 in gaps, and center each name label under its own photo.

Platforms
web · Any OS (web)
Applications
Google Slides
Interfaces
brow · Browser
GUI need · task setting
G2 · GUI-essential
GUI need · ordinary access
G2 · GUI-essential
Interaction primitives
drag · Drag & drop geom · Precise geometry click · Point & click
Work product
WP03 · Visual layout
Conditions
None declared
Proposed difficulty
3 / 5

Proposed verifier

Slides API read: 8 images 1.6 in square (±0.02) with crop properties giving 1:1; grid positions and gaps equal (±0.02), grid centered; each label's center x equals its photo's (±0.02) and label below photo; label-photo pairing preserved. Agent has no API token and Apps Script is disabled for the test account; revision history must show editor UI edits.

Proposed partial credit

25% crops 1:1; 35% grid geometry; 25% labels centered under correct photo; 15% centered on slide.

R04.04.01.02-T08 Fix text overflow across ten slides using continuation slides

Original instruction

input/compliance_update.pptx has 10 slides whose body text overflows. Fix each without autofit shrinking and keeping fonts ≥18 pt: tighten box sizes within margins, or move trailing bullets to a new continuation slide titled '<title> (cont.)' directly after. Do not edit wording.

Platforms
mac · macOS win · Windows
Applications
PowerPoint
Interfaces
desk · Desktop GUI
GUI need · task setting
G2 · GUI-essential
GUI need · ordinary access
G2 · GUI-essential
Interaction primitives
click · Point & click drag · Drag & drop vis · Visual judgment
Work product
WP03 · Visual layout
Conditions
None declared
Proposed difficulty
3 / 5

Proposed verifier

Render via PowerPoint export: no text outside its shape or slide on any slide; normAutofit absent or fontScale 100%; fonts ≥18 pt; concatenated bullet text equals input in order; continuation titles exact. VBA, COM and AppleScript automation disabled, no Python in the VM; file must be saved by PowerPoint (app.xml).

Proposed partial credit

60% ten slides overflow-free (pro-rata); 20% fonts/no shrink; 20% text order and continuation titles.

R04.04.01.02-T09 Lay out twelve partner logos in two rows in PowerPoint for the web

Original instruction

In PowerPoint for the web (test tenant), open 'Partners.pptx' on OneDrive. On slide 5 arrange the 12 partner logos in two rows of 6, all logos scaled proportionally to the same height (0.6 in), rows evenly spaced, columns centered on equal-width cells across the content area.

Platforms
web · Any OS (web)
Applications
PowerPoint for the web
Interfaces
brow · Browser
GUI need · task setting
G3 · GUI-only
GUI need · ordinary access
G2 · GUI-essential
Interaction primitives
drag · Drag & drop geom · Precise geometry click · Point & click
Work product
WP03 · Visual layout
Conditions
None declared
Proposed difficulty
3 / 5

Proposed verifier

Grader downloads the file via Graph (read-only) and checks pic xfrm: heights 0.6 in (±0.02) with aspect preserved; 2 rows of 6; centers on equal cell grid (±0.03); file version edited in the browser (OneDrive version author = test user, no desktop app session). Agent has no Graph write scope.

Proposed partial credit

30% equal heights with aspect kept; 40% grid placement; 20% row spacing; 10% edited in browser.

R04.04.01.02-T10 Snap a 20-slide deck to a 12-column grid with guides

Original instruction

Add 12-column vertical guides (0.5 in margins, 0.2 in gutters, 16:9) to the master of input/strategy.pptx, then move and resize every content object on 20 slides so its left and right edges fall on column edges and nothing crosses the margins. Keep reading order and z-order.

Platforms
mac · macOS win · Windows
Applications
PowerPoint
Interfaces
desk · Desktop GUI
GUI need · task setting
G2 · GUI-essential
GUI need · ordinary access
CL · CLI-sufficient
Interaction primitives
geom · Precise geometry drag · Drag & drop nav · Pan/zoom/scroll
Work product
WP03 · Visual layout
Conditions
long · Long-horizon
Proposed difficulty
4 / 5

Proposed verifier

Master guides (p15:sldGuideLst) at the 24 expected positions; every non-placeholder object's x and x+cx on a column edge (±0.02 in); none outside margins; z-order and reading order unchanged vs input. VBA, COM and AppleScript automation disabled, no Python in the VM; file must be saved by PowerPoint (app.xml).

Proposed partial credit

20% guides; 60% objects on grid (pro-rata); 10% margins; 10% order preserved.

R04.04.01.02-T11 Repair font-substitution overflow in a PPTX opened in Keynote

Original instruction

input/client_pitch.pptx was made on Windows; opened in Keynote, font substitution makes text overflow on 9 slides. In Keynote, keep the substitute font but fix every overflow by resizing boxes or adjusting line spacing (≥0.9), then save output/client_pitch.key and export output/client_pitch.pdf.

Platforms
mac · macOS
Applications
Keynote · PowerPoint
Interfaces
desk · Desktop GUI
GUI need · task setting
G2 · GUI-essential
GUI need · ordinary access
G2 · GUI-essential
Interaction primitives
drag · Drag & drop vis · Visual judgment click · Point & click
Work product
WP03 · Visual layout
Conditions
None declared
Proposed difficulty
3 / 5

Proposed verifier

PDF: all text runs within their box bounds and slide (pdfplumber vs exported shape bboxes via grader AppleScript); line spacing ≥0.9; text unchanged; both files present. osascript blocked for the agent.

Proposed partial credit

60% nine slides fixed (pro-rata); 20% spacing rule; 20% files exported and text unchanged.

R04.04.01.02-T12 Rearrange a slide to match a reference screenshot within 5 px

Original instruction

input/target_slide7.png shows the approved layout of slide 7 in input/roadmap.pptx. Move, resize and reorder (z-order) the 11 shapes on slide 7 so the rendered slide matches the reference. Do not change text, colors or add shapes.

Platforms
mac · macOS win · Windows
Applications
PowerPoint
Interfaces
desk · Desktop GUI
GUI need · task setting
G2 · GUI-essential
GUI need · ordinary access
G2 · GUI-essential
Interaction primitives
drag · Drag & drop geom · Precise geometry vis · Visual judgment
Work product
WP03 · Visual layout
Conditions
None declared
Proposed difficulty
4 / 5

Proposed verifier

Render slide 7 at 1920×1080 via PowerPoint export; per-shape bbox vs reference bbox (from hidden reference pptx) within 5 px; z-order equal; SSIM ≥0.95 overall; text/colors unchanged. VBA, COM and AppleScript automation disabled, no Python in the VM; file must be saved by PowerPoint (app.xml).

Proposed partial credit

60% shapes within tolerance (pro-rata over 11); 20% z-order; 20% SSIM threshold.

R04.04.01.02-T13 Arrange a process diagram by touch in Keynote for iPad

Original instruction

On iPad, open 'Onboarding Flow.key' in Keynote. On slide 3 align the five process shapes' vertical centers on one line, make the gaps between them equal, center the row on the slide, and place the title text box exactly above the middle shape, using touch drag and alignment snapping.

Platforms
ios · iOS/iPadOS
Applications
Keynote
Interfaces
mob · Mobile touch
GUI need · task setting
G3 · GUI-only
GUI need · ordinary access
G3 · GUI-only
Interaction primitives
touch · Touch gestures drag · Drag & drop geom · Precise geometry
Work product
WP03 · Visual layout
Conditions
None declared
Proposed difficulty
4 / 5

Proposed verifier

File synced via iCloud; grader opens it on a Mac via AppleScript and reads positions: centers equal (±2 pt), equal gaps (±2 pt), row centered (±3 pt), title center x equal to middle shape center. Shortcuts is disabled on the agent's iPad and it has no Mac or iCloud web access; the file must be saved by Keynote for iPadOS.

Proposed partial credit

35% vertical centers aligned; 30% equal gaps; 15% centered row; 20% title placement.

R04.04.01.02-T14 Fix every layout-lint violation in a 40-slide deck

Original instruction

input/all_hands.pptx (40 slides) comes with input/lint_report.csv listing 63 violations: content outside 0.4 in safe margins, fonts below 14 pt, overlapping shapes, titles not at the master title position. Fix all violations without changing text. Save output/all_hands_fixed.pptx.

Platforms
mac · macOS win · Windows
Applications
PowerPoint
Interfaces
desk · Desktop GUI
GUI need · task setting
G2 · GUI-essential
GUI need · ordinary access
CL · CLI-sufficient
Interaction primitives
geom · Precise geometry drag · Drag & drop vis · Visual judgment
Work product
WP03 · Visual layout
Conditions
long · Long-horizon
Proposed difficulty
5 / 5

Proposed verifier

Grader re-runs the same lint on the output: 0 violations of the 4 rule types; text per slide unchanged; slide count unchanged; no shapes deleted. Saved by PowerPoint; VBA, COM and AppleScript automation disabled, no Python in the VM.

Proposed partial credit

70% violations fixed (pro-rata over 63); 20% text and shape count unchanged; 10% no new violations.

R04.04.01.02-T15 Convert a 4:3 deck to 16:9 without stretching or clipping

Original instruction

Convert input/legacy_training.pptx (25 slides, 4:3) to 16:9. Images must keep their aspect ratio, nothing may sit outside the slide, content must use the new width (no 4:3 pillarbox look: content spans ≥80% of slide width where it spanned ≥80% before), and the master and layouts must be resized too.

Platforms
mac · macOS win · Windows
Applications
PowerPoint
Interfaces
desk · Desktop GUI
GUI need · task setting
G2 · GUI-essential
GUI need · ordinary access
CL · CLI-sufficient
Interaction primitives
geom · Precise geometry drag · Drag & drop vis · Visual judgment
Work product
WP03 · Visual layout
Conditions
long · Long-horizon
Proposed difficulty
5 / 5

Proposed verifier

sldSz 16:9; every pic's ext ratio equals its source pixel ratio (±1%) after crop; all shapes within slide; width-coverage rule per slide; layouts/master sized 16:9; render SSIM vs reference ≥0.85 per slide (rubric-free). VBA, COM and AppleScript automation disabled, no Python in the VM; file must be saved by PowerPoint (app.xml).

Proposed partial credit

15% slide size and masters; 35% no image distortion; 25% nothing outside slide; 25% width coverage.

R04.04.01.02-T16 Fix layout breakage in a German translation of an 18-slide deck

Original instruction

Google Slides deck 'Produktlaunch DE' was machine-translated from English; German text is ~30% longer and overflows or wraps badly on 18 slides. Fix each by resizing within the layout's safe margins, using continuation slides where needed, keeping fonts ≥16 pt and never splitting product names (input/product_names.txt) across lines.

Platforms
web · Any OS (web)
Applications
Google Slides
Interfaces
brow · Browser
GUI need · task setting
G2 · GUI-essential
GUI need · ordinary access
G2 · GUI-essential
Interaction primitives
drag · Drag & drop geom · Precise geometry vis · Visual judgment
Work product
WP03 · Visual layout
Conditions
long · Long-horizon
Proposed difficulty
5 / 5

Proposed verifier

Export PDF: text within boxes and margins; Slides API fonts ≥16 pt; product names never broken across lines (line-level text extraction); text content unchanged except continuation titles; slide order consistent. Agent has no API token and Apps Script is disabled for the test account; revision history must show editor UI edits.

Proposed partial credit

50% slides overflow-free (pro-rata); 20% product-name rule; 15% fonts; 15% text unchanged.

R04.04.01.02-T17 Control pair of T01: align and distribute icons with python-pptx

Original instruction

Control pair of T01. Using python-pptx from the terminal, apply the same top alignment and equal distribution between 1.0 in margins to the four groups on slide 3 of input/services.pptx and save output/services_cli.pptx.

Platforms
mac · macOS win · Windows lin · Linux
Applications
python-pptx
Interfaces
term · Terminal
GUI need · task setting
CL · CLI-sufficient
GUI need · ordinary access
CL · CLI-sufficient
Interaction primitives
code · Code & commands
Work product
WP03 · Visual layout
Conditions
None declared
Proposed difficulty
2 / 5

Proposed verifier

Same geometric checker as T01 without the PowerPoint-save requirement; used to compare scripted vs GUI success.

Proposed partial credit

35% tops aligned; 40% distribution; 25% sizes unchanged.

R04.04.01.02-T18 Fix z-order and grouping so infographic labels are visible

Original instruction

Slide 4 of input/supply_chain.pptx is an infographic where 7 labels are hidden behind shapes and arrows sit above icons. Reorder layers so every label is fully visible above its shape and arrows sit behind icons, then group each icon with its label (7 groups). Do not move anything.

Platforms
mac · macOS win · Windows
Applications
PowerPoint
Interfaces
desk · Desktop GUI
GUI need · task setting
G2 · GUI-essential
GUI need · ordinary access
CL · CLI-sufficient
Interaction primitives
click · Point & click vis · Visual judgment hover · Hover & context
Work product
WP03 · Visual layout
Conditions
None declared
Proposed difficulty
4 / 5

Proposed verifier

spTree order: each label after its shape; arrows before icons; 7 grpSp each containing exactly the right icon+label; absolute positions unchanged (±0.01 in); render: each label's text pixels visible (OCR finds all 7 labels). VBA, COM and AppleScript automation disabled, no Python in the VM; file must be saved by PowerPoint (app.xml).

Proposed partial credit

35% labels above shapes; 15% arrows behind icons; 30% correct groups; 20% nothing moved.

R04.04.01.02-T19 Reattach detached flowchart connectors to connection points

Original instruction

On slide 5 of input/claims_process.pptx, shapes were moved and 9 connectors are now loose lines. Re-glue each connector to the correct connection sites per input/flow_edges.csv, use elbow connectors, and ensure no connector passes through a shape other than its endpoints.

Platforms
mac · macOS win · Windows
Applications
PowerPoint
Interfaces
desk · Desktop GUI
GUI need · task setting
G2 · GUI-essential
GUI need · ordinary access
CL · CLI-sufficient
Interaction primitives
drag · Drag & drop geom · Precise geometry vis · Visual judgment
Work product
WP03 · Visual layout
Conditions
None declared
Proposed difficulty
4 / 5

Proposed verifier

cxnSp stCxn/endCxn ids and idx match flow_edges.csv for all 9; prstGeom bentConnector*; rendered connector paths do not intersect non-endpoint shape bboxes; shapes not moved. VBA, COM and AppleScript automation disabled, no Python in the VM; file must be saved by PowerPoint (app.xml).

Proposed partial credit

60% connectors glued correctly (pro-rata over 9); 15% elbow type; 15% no pass-through; 10% shapes unchanged.

R04.04.01.02-T20 Place timeline milestones proportionally to their dates in Keynote

Original instruction

Slide 2 of input/program_timeline.key has an arrow from Jan 2026 (left end) to Dec 2026 (right end) and 8 milestone markers with dates in their labels. Position each marker's center horizontally at its date's proportional position along the arrow, alternate labels above and below, and keep labels from overlapping.

Platforms
mac · macOS
Applications
Keynote
Interfaces
desk · Desktop GUI
GUI need · task setting
G2 · GUI-essential
GUI need · ordinary access
CL · CLI-sufficient
Interaction primitives
geom · Precise geometry drag · Drag & drop vis · Visual judgment
Work product
WP03 · Visual layout
Conditions
None declared
Proposed difficulty
3 / 5

Proposed verifier

Grader AppleScript reads positions: each marker center x within ±3 pt of the date-proportional x; labels alternate above/below; label bboxes pairwise non-overlapping; arrow unchanged. osascript blocked for the agent.

Proposed partial credit

55% markers placed (pro-rata over 8); 25% alternating labels; 20% no overlaps.

Mapped research references

Existing benchmark-to-class links from the research catalog. They are not record-level annotations, executions or product endorsements.

In this workflow family