Computer-Use Agent Taxonomy中文
R04.04.01.06 · TASK CLASS

Slide content & formatting edits

Insert, move and format text, images and backgrounds on existing slides, including image adjustments in place.

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 · LibreOffice Impress · Keynote · Google Slides
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.

R04.04.01.06-T01 Swap the old logo on two slides while keeping position, size and alt text

Original instruction

Open input/board_update.pptx in PowerPoint on Windows. Replace the old logo picture on slides 1 and 12 with input/logo_2026.png so each new logo keeps the old one's position and size, and set its alt text to 'Company logo'. Change nothing else. Save.

Platforms
win · Windows mac · macOS
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 hover · Hover & context
Work product
WP03 · Visual layout
Conditions
None declared
Proposed difficulty
2 / 5

Proposed verifier

python-pptx: logo pictures on slides 1 and 12 carry the new image blob (hash), xfrm equal to the old one within 0.01 in, descr 'Company logo'; all other shapes hash-identical. Saved by POWERPNT.EXE (Sysmon writer, app.xml); VBA blocked by policy and the agent has no Python or shell. Control pair: T20.

Proposed partial credit

50% both logos swapped; 30% position and size kept; 20% alt text set and nothing else changed

R04.04.01.06-T02 Set solid and picture backgrounds and restyle a title in Google Slides

Original instruction

In the Google Slides deck 'Town hall' (test account), set slide 3's background to solid #0B2545 and make its title text white, 40 pt. Set slide 4's background to the image bg_waves.png from the shared Drive folder 'Assets'. Leave every other slide unchanged.

Platforms
web · Any OS (web)
Applications
Google Slides · Google Drive
Interfaces
brow · Browser
GUI need · task setting
G2 · GUI-essential
GUI need · ordinary access
G2 · GUI-essential
Interaction primitives
click · Point & click type · Text & forms
Work product
WP03 · Visual layout
Conditions
None declared
Proposed difficulty
2 / 5

Proposed verifier

Slides API read by the grader: slide 3 pageBackgroundFill solid #0B2545, title runs #FFFFFF 40 pt; slide 4 stretchedPictureFill whose downloaded image matches bg_waves.png (pHash); other slides unchanged. Apps Script is disabled for the test domain and the agent has no OAuth token; revisions must come from the browser session.

Proposed partial credit

35% slide 3 background; 30% title formatting; 35% slide 4 picture background with others unchanged

R04.04.01.06-T03 Mask a headshot into a centered 6 cm circle in Keynote

Original instruction

Open input/team.key in Keynote on macOS. On slide 5 insert input/photos/priya.jpg, mask it with a circle so her face is centered in it, make the circle 6 cm in diameter, place it exactly over the gray placeholder circle, then delete the placeholder. Save.

Platforms
mac · macOS
Applications
Keynote
Interfaces
desk · Desktop GUI
GUI need · task setting
G3 · GUI-only
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
3 / 5

Proposed verifier

Grader exports slide 5 to PNG with its own AppleScript: circular image region centered on the placeholder position within 2 mm, diameter 6 +/-0.2 cm, detected face center within 10% of the circle radius from its center, placeholder gone. Masks are not exposed in Keynote's AppleScript dictionary and osascript is disabled for the agent.

Proposed partial credit

25% photo inserted and masked to a circle; 25% diameter; 25% position; 25% face centered with placeholder removed

R04.04.01.06-T04 Crop a site photo to 16:9 around a marked subject and correct it

Original instruction

In input/site_visit.pptx (PowerPoint), crop the photo on slide 6 to a 16:9 aspect so the crane outlined in input/crop_ref.png stays fully inside, size it 24 cm wide, and set Picture Corrections brightness +10% and contrast -10%. Do not replace or compress the image. Save.

Platforms
win · Windows mac · macOS
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
3 / 5

Proposed verifier

python-pptx/XML: srcRect gives 16:9 +/-1% with the gold crane box inside the crop, width 24 +/-0.1 cm, brightnessContrast effect bright=10000 contrast=-10000, image blob hash unchanged (no compression). Saved by PowerPoint (writer provenance); VBA blocked and no Python or shell for the agent.

Proposed partial credit

35% 16:9 crop containing the crane; 20% width; 30% corrections; 15% original image retained

R04.04.01.06-T05 Restore two-level bullets and spacing on four Impress slides

Original instruction

Open input/training.odp in LibreOffice Impress. On slides 3-6 all bullet lines are level 1; lines beginning with '- ' must become level 2 with the '- ' removed. Set level-1 text to 24 pt and level-2 to 20 pt, proportional line spacing 120% and 6 pt space before each paragraph. Save.

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

Proposed verifier

odfpy: list level per paragraph equals gold, '- ' prefixes gone with text otherwise unchanged, font sizes 24/20 pt, fo:line-height 120%, margin-top 6 pt. Saved by soffice (writer provenance); macro security Very High and headless soffice/UNO runs are detected.

Proposed partial credit

40% list levels and prefixes; 30% font sizes; 30% line and paragraph spacing

R04.04.01.06-T06 Cut out headphones from a cluttered photo with Remove Background

Original instruction

Slide 2 of input/product_launch.pptx shows headphones on a cluttered desk. In PowerPoint use Remove Background, marking areas to keep or remove as needed, so only the headphones remain (including the thin headband), then add an outer soft shadow. Save.

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

Proposed verifier

XML must contain a native a14:backgroundRemoval effect on the original blob (a pre-cut PNG scores 0). Grader renders the slide via PowerPoint and compares the visible alpha region with the gold mask: IoU >= 0.90; outerShdw present. VBA blocked and no image editor or Python is installed for the agent.

Proposed partial credit

20% native background removal applied; 60% mask IoU (linear 0.75-0.90); 20% soft shadow

R04.04.01.06-T07 Add a recolored, semi-transparent crest watermark behind slide content

Original instruction

In the Google Slides deck 'Annual report' (test account), on slides 2-5 insert crest.png from Drive folder 'Brand', 10 cm wide, centered on the slide, sent behind all other objects, recolored with 'Light 2' and set to 70% transparency.

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

Proposed verifier

Slides API read by the grader on slides 2-5: image matching crest.png (pHash), width 10 +/-0.1 cm, centered within 0.2 cm, first in z-order, recolor LIGHT2, transparency 0.70 +/-0.02. Apps Script disabled for the test domain, no OAuth token for the agent; revisions from the browser session.

Proposed partial credit

30% inserted on all four slides; 30% size, centering and z-order; 40% recolor and transparency

R04.04.01.06-T08 Match six dull product photos to target looks with Keynote image adjustments

Original instruction

input/catalog.key has 6 product photos on slides 2-7 that look dull. Using Keynote's Image Adjustments (exposure, saturation, contrast, highlights, shadows, temperature, tint), make each photo match input/targets/slideN.png. Do not replace the images or edit them outside Keynote. Save.

Platforms
mac · macOS
Applications
Keynote
Interfaces
desk · Desktop GUI
GUI need · task setting
G3 · GUI-only
GUI need · ordinary access
G2 · GUI-essential
Interaction primitives
drag · Drag & drop vis · Visual judgment
Work product
WP03 · Visual layout
Conditions
None declared
Proposed difficulty
5 / 5

Proposed verifier

Grader exports slides with its own AppleScript and computes mean dE2000 inside each photo box vs target: pass <= 4, partial linear to 10. Image files inside the .key package must be unchanged (hash), proving in-app adjustments. Image Adjustments are not scriptable in Keynote and osascript is disabled for the agent; no image editor installed.

Proposed partial credit

85% color match across 6 photos (pro-rata by dE2000); 15% original images untouched

R04.04.01.06-T09 Give six team photos rounded square crops, borders and shadows

Original instruction

Slide 8 of input/org_overview.pptx has 6 team photos of different shapes. In PowerPoint make each a rounded-rectangle crop 4 x 4 cm with the face centered, add a 3 pt white border and an 'Offset: Bottom Right' outer shadow, and keep each photo's current center point. Save.

Platforms
win · Windows mac · macOS
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 click · Point & click
Work product
WP03 · Visual layout
Conditions
None declared
Proposed difficulty
3 / 5

Proposed verifier

python-pptx/XML: prstGeom roundRect, ext 4x4 +/-0.05 cm, ln 38100 EMU FFFFFF, outerShdw matching the preset, center within 0.1 cm of input; gold face boxes inside the crop and near its center. Saved by PowerPoint (writer provenance); VBA blocked, no Python or shell for the agent.

Proposed partial credit

30% shape and size; 30% border and shadow; 25% centers kept; 15% faces centered

R04.04.01.06-T10 Move a quote and photo onto another slide's guide boxes and drop the old slide

Original instruction

In input/strategy.pptx, move the customer quote text box and the photo from slide 7 onto slide 3, fitting them inside the two dashed guide boxes on its right half; set the photo 9 cm wide with aspect kept and keep the quote's formatting. Delete the guide boxes and then slide 7. Save.

Platforms
win · Windows mac · macOS
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 click · Point & click
Work product
WP03 · Visual layout
Conditions
None declared
Proposed difficulty
3 / 5

Proposed verifier

python-pptx: 11 slides; slide 3 holds the quote with identical runs/rPr and the photo with the same blob, each inside its guide box bbox (+/-0.2 cm), width 9 cm, aspect within 0.5%; guide shapes gone. Saved by PowerPoint (writer provenance); VBA blocked, no Python or shell.

Proposed partial credit

35% objects moved intact; 25% placement in guide boxes; 20% photo size and aspect; 20% guides and slide 7 removed

R04.04.01.06-T11 Replace eight screenshots by visual match while keeping geometry and crop

Original instruction

input/release_notes.odp has 8 slides with app screenshots; input/screens_v2/ has the 8 updated screenshots under unrelated file names. In LibreOffice Impress replace each old screenshot with its updated counterpart (match by content) so the new one keeps the old one's position, size and crop. Save.

Platforms
win · Windows lin · Linux
Applications
LibreOffice Impress
Interfaces
desk · Desktop GUI
GUI need · task setting
G2 · GUI-essential
GUI need · ordinary access
G2 · GUI-essential
Interaction primitives
click · Point & click vis · Visual judgment
Work product
WP03 · Visual layout
Conditions
None declared
Proposed difficulty
4 / 5

Proposed verifier

odfpy: each draw:frame's embedded image equals the gold-mapped new file (hash), svg:x/y/width/height unchanged, fo:clip unchanged. Saved by soffice (writer provenance); macro security Very High and headless soffice/UNO runs are detected.

Proposed partial credit

60% correct mapping (pro-rata over 8); 25% geometry kept; 15% crop kept

R04.04.01.06-T12 Color positive and negative KPI deltas in green and red across six slides

Original instruction

In the Google Slides deck 'Q3 KPIs' (test account), slides 2-7 contain callouts with changes like '+4.2%' and '-1.8%'. Make every positive value bold #1E8E3E and every negative value bold #D93025; leave all other text unchanged.

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
click · Point & click type · Text & forms vis · Visual judgment
Work product
WP03 · Visual layout
Conditions
None declared
Proposed difficulty
3 / 5

Proposed verifier

Slides API read by the grader: the 31 gold value ranges carry the right color and bold; all other runs keep their original style and text. Apps Script disabled for the test domain and no OAuth token for the agent; revisions from the browser session.

Proposed partial credit

70% values formatted correctly (pro-rata); 30% no collateral changes

R04.04.01.06-T13 Apply a semi-transparent photo background to five slides and fix text contrast

Original instruction

In input/conference.pptx (PowerPoint), give slides 2-6 only a picture background from input/bg_city.jpg at 40% transparency, without changing the slide master or layouts. Then make every text box on those slides meet WCAG 4.5:1 contrast against the rendered background by changing text colors only. Save.

Platforms
win · Windows mac · macOS
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 vis · Visual judgment
Work product
WP03 · Visual layout
Conditions
None declared
Proposed difficulty
4 / 5

Proposed verifier

XML: slide-level bg blipFill on slides 2-6 only, alphaModFix 60000, master/layouts hash-identical. Grader renders each slide and computes contrast of each text box vs sampled background (>= 4.5); only color attributes may differ from input. Saved by PowerPoint; VBA blocked, no Python or shell.

Proposed partial credit

40% backgrounds on exactly slides 2-6; 20% master and layouts untouched; 40% contrast met (pro-rata over text boxes)

R04.04.01.06-T14 Remove studio backgrounds with Instant Alpha and line up three products

Original instruction

In input/lineup.key (Keynote), slide 4 has 3 product photos on white studio backgrounds. Use Instant Alpha to remove each background, set the slide background to a linear gradient #1B2735 to #090A0F at 90 degrees, scale the products to the same 7 cm height and sit them on the baseline guide with equal gaps. Save.

Platforms
mac · macOS
Applications
Keynote
Interfaces
desk · Desktop GUI
GUI need · task setting
G3 · GUI-only
GUI need · ordinary access
G2 · GUI-essential
Interaction primitives
path · Continuous path drag · Drag & drop geom · Precise geometry
Work product
WP03 · Visual layout
Conditions
None declared
Proposed difficulty
5 / 5

Proposed verifier

Grader renders slide 4 with its own AppleScript: per-product alpha IoU vs gold masks >= 0.9, gradient endpoints dE2000 < 3, heights 7 +/-0.1 cm, bottoms on the baseline +/-1 mm, gaps equal +/-1 mm; package images derive from the originals (pHash). osascript is disabled and no image editor is installed.

Proposed partial credit

45% background removal IoU (pro-rata); 15% gradient; 20% heights; 20% baseline and equal gaps

R04.04.01.06-T15 Apply a 22-item change list to a 30-slide sales deck

Original instruction

Open input/sales_deck.pptx (30 slides) in PowerPoint and input/change_list.xlsx in Excel. Apply all 22 listed edits: text replacements, font and color fixes, 6 image swaps from input/new_images/, 3 background changes and 2 crops, exactly as specified. Nothing else in the deck may change. Save.

Platforms
win · Windows mac · macOS
Applications
PowerPoint · Excel
Interfaces
desk · Desktop GUI
GUI need · task setting
G2 · GUI-essential
GUI need · ordinary access
CL · CLI-sufficient
Interaction primitives
click · Point & click multi · Cross-app drag · Drag & drop
Work product
WP03 · Visual layout
Conditions
multiapp · Multi-app long · Long-horizon
Proposed difficulty
5 / 5

Proposed verifier

Grader runs one python-pptx check per edit (22) plus a diff of every other shape and slide vs input (must be identical). Saved by PowerPoint (writer provenance); VBA blocked in both apps and the agent has no Python or shell.

Proposed partial credit

40% text and formatting edits; 30% image swaps and crops; 15% background changes; 15% no collateral changes

R04.04.01.06-T16 Swap speaker headshots into oval frames with faces centered in Google Slides

Original instruction

In the Google Slides deck 'Summit speakers' (test account), slides 3-7 each show a speaker headshot in an oval frame. Replace each with the matching photo from Drive folder 'Speakers 2026' (file names = speaker names), keep the oval frame size and position, and adjust each crop so the face is centered in the oval.

Platforms
web · Any OS (web)
Applications
Google Slides · Google Drive
Interfaces
brow · Browser
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
4 / 5

Proposed verifier

Slides API: correct source image per slide (pHash), element size/position unchanged within 0.1 cm, ellipse mask kept; grader maps gold face boxes through cropProperties and requires the face center within 8% of the frame center. Apps Script disabled for the domain, no OAuth token; revisions from the browser session.

Proposed partial credit

35% correct photos placed; 25% frame geometry kept; 40% faces centered (pro-rata)

R04.04.01.06-T17 Insert four SVG service icons, recolor to brand blue and size them

Original instruction

In input/services.pptx on PowerPoint for Mac, slide 9 has 4 empty icon slots above service names. Insert the matching icon from input/icons/*.svg (file name = service) into each slot, recolor each to #0057B8 with Graphics Fill, and size each 1.5 cm high with aspect kept, centered in its slot. 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

XML: 4 pictures with svgBlip, each matched to its slot by name and position (center within 0.1 cm); grader renders the slide and requires icon pixels within dE2000 3 of #0057B8; height 1.5 +/-0.05 cm, aspect kept. Saved by PowerPoint (writer provenance); AppleScript and VBA blocked for the agent.

Proposed partial credit

40% icons in the correct slots; 30% recolor; 30% size, aspect and centering

R04.04.01.06-T18 Crop an archival photo to guide lines and set grayscale and gamma in Impress

Original instruction

On slide 4 of input/history.odp (LibreOffice Impress), crop the archival photo by dragging its crop handles to the four red guide lines drawn on the slide, then delete the guide lines. Set the photo's color mode to Grayscale and gamma to 1.2, and keep its top-left position. Save.

Platforms
win · Windows mac · macOS lin · Linux
Applications
LibreOffice Impress
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
Work product
WP03 · Visual layout
Conditions
None declared
Proposed difficulty
3 / 5

Proposed verifier

odfpy: fo:clip values within 1 mm of gold, draw:color-mode greyscale, draw:gamma 120%, svg:x/y unchanged, guide shapes gone. Saved by soffice (writer provenance); macro security Very High and headless soffice/UNO runs are detected.

Proposed partial credit

45% crop to guides; 30% grayscale and gamma; 25% position kept and guides removed

R04.04.01.06-T19 Fill the word GROW with a photo while keeping it editable text

Original instruction

On the title slide of input/grow_campaign.pptx (PowerPoint), make the word 'GROW' (Arial Black 120 pt) show input/field.jpg through its letters using a picture fill on the text, so the photo appears only inside the glyphs. The word must remain editable text, centered horizontally on the slide. Save.

Platforms
win · Windows mac · macOS
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 vis · Visual judgment
Work product
WP03 · Visual layout
Conditions
None declared
Proposed difficulty
4 / 5

Proposed verifier

XML: run rPr holds blipFill with field.jpg's blob; text 'GROW' in Arial Black 120 pt remains a text run; box centered within 0.2 cm. Grader renders the slide: photo pixels only inside glyph mask (SSIM >= 0.8 vs reference). Saved by PowerPoint; VBA blocked and no Python or image editor for the agent.

Proposed partial credit

40% picture fill on the text run; 25% still editable text with font/size; 15% centered; 20% render matches

R04.04.01.06-T20 Control pair of T01: swap the logos with python-pptx

Original instruction

Control pair of T01. Using Python 3 with python-pptx in a terminal, replace the logo picture on slides 1 and 12 of input/board_update.pptx with input/logo_2026.png, keeping each picture's position and size, set alt text 'Company logo', and save to output/board_update.pptx.

Platforms
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 checks as T01 on output/board_update.pptx: new image blob on both slides, xfrm within 0.01 in, descr set, all other shapes hash-identical. CLI control for the GUI task T01.

Proposed partial credit

50% both logos swapped; 30% position and size kept; 20% alt text set and nothing else changed

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