Computer-Use Agent Taxonomy中文
R08.01.01.02 · TASK CLASS

Sky simulation & planetarium setup

Configure planetarium software to show a specified sky and measure from it.

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

Typical applications
Stellarium · Celestia · WorldWide Telescope
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.

R08.01.01.02-T01 Report Jupiter's altitude from Mauna Kea at a set time in Stellarium

Original instruction

In Stellarium 24, set the location to Mauna Kea Observatory (19.8207 N, 155.4681 W, 4,205 m), date/time 2026-12-14 03:00 UTC, and write Jupiter's apparent altitude and azimuth (deg, 2 decimals) to output/answer.json. Leave Stellarium at that location and time.

Platforms
mac · macOS win · Windows lin · Linux
Applications
Stellarium
Interfaces
desk · Desktop GUI
GUI need · task setting
G2 · GUI-essential
GUI need · ordinary access
CL · CLI-sufficient
Interaction primitives
click · Point & click type · Text & forms
Work product
WP12 · Answer/decision
Conditions
None declared
Proposed difficulty
2 / 5

Proposed verifier

Grader reads location and JD via RemoteControl: location within 0.01 deg, time within 1 min; altitude and azimuth within 0.5 deg of ephemeris. Stellarium's script console and RemoteControl are disabled during the session; grader enables RemoteControl afterwards to read state.

Proposed partial credit

30% location; 20% time; 50% alt/az values.

R08.01.01.02-T02 Find a Mars opposition date and magnitude with AstroCalc

Original instruction

In Stellarium, use the Astronomical Calculations window (Phenomena) for 2027 to find the date and time (UTC) of Mars opposition and its visual magnitude at that moment. Write them to output/answer.json.

Platforms
mac · macOS win · Windows lin · Linux
Applications
Stellarium
Interfaces
desk · Desktop GUI
GUI need · task setting
G2 · GUI-essential
GUI need · ordinary access
CL · CLI-sufficient
Interaction primitives
click · Point & click type · Text & forms
Work product
WP12 · Answer/decision
Conditions
None declared
Proposed difficulty
3 / 5

Proposed verifier

Grader compares with Stellarium's own ephemeris (DE440): date exact, time within 1 h, magnitude within 0.05. Stellarium's script console and RemoteControl are disabled during the session; grader enables RemoteControl afterwards to read state.

Proposed partial credit

50% date; 20% time; 30% magnitude.

R08.01.01.02-T03 Find which Galilean moon transits Jupiter first after a given date

Original instruction

In Celestia, set the date to Julian Date 2460000.5, center and follow Jupiter, and advance time until the first Galilean moon begins a transit across Jupiter's disk. Write the moon's name and the transit-start time (UTC, to the minute) to output/answer.json.

Platforms
mac · macOS win · Windows lin · Linux
Applications
Celestia
Interfaces
desk · Desktop GUI
GUI need · task setting
G2 · GUI-essential
GUI need · ordinary access
CL · CLI-sufficient
Interaction primitives
click · Point & click type · Text & forms nav · Pan/zoom/scroll vis · Visual judgment
Work product
WP12 · Answer/decision
Conditions
None declared
Proposed difficulty
3 / 5

Proposed verifier

Grader computes from ephemerides: moon exact, ingress time within 5 min; final Celestia state is centered on Jupiter. Celestia's Lua/CEL console is disabled; grader reads the final state via its own cel:// URL dump.

Proposed partial credit

50% moon; 40% time; 10% final view centered on Jupiter.

R08.01.01.02-T04 Match a reference planetarium screenshot in Stellarium

Original instruction

Configure Stellarium so its view reproduces input/reference_view.png: infer and set location, date/time, viewing direction, field of view, projection, landscape, atmosphere and label/grid options (hints in input/notes.txt). Save the settings and take output/match.png.

Platforms
mac · macOS win · Windows lin · Linux
Applications
Stellarium
Interfaces
desk · Desktop GUI
GUI need · task setting
G2 · GUI-essential
GUI need · ordinary access
G2 · GUI-essential
Interaction primitives
click · Point & click type · Text & forms drag · Drag & drop vis · Visual judgment
Work product
WP08 · App/system state
Conditions
long · Long-horizon
Proposed difficulty
5 / 5

Proposed verifier

Grader re-renders from the saved state: SSIM >=0.85 vs reference at 1280x720; also checks location (0.5 deg), time (10 min), FOV (5%), direction (1 deg). Stellarium's script console and RemoteControl are disabled during the session; grader enables RemoteControl afterwards to read state.

Proposed partial credit

40% SSIM (pro-rata from 0.6); 20% location/time; 20% FOV/direction; 20% display options.

R08.01.01.02-T05 Configure grids, art and light pollution in Stellarium

Original instruction

In Stellarium, turn on the equatorial (J2000) grid and constellation art (art brightness 0.4), turn off atmosphere and ground, set the light-pollution Bortle class to 3, and save these as the default settings.

Platforms
mac · macOS win · Windows lin · Linux
Applications
Stellarium
Interfaces
desk · Desktop GUI
GUI need · task setting
G2 · GUI-essential
GUI need · ordinary access
CL · CLI-sufficient
Interaction primitives
click · Point & click
Work product
WP08 · App/system state
Conditions
None declared
Proposed difficulty
2 / 5

Proposed verifier

Grader parses config.ini after restart: all six settings exact; no other sections changed. Stellarium's script console and RemoteControl are disabled during the session; grader enables RemoteControl afterwards to read state.

Proposed partial credit

80% six settings correct (equal shares); 20% saved as default and nothing else changed.

R08.01.01.02-T06 Build a four-stop guided tour in WorldWide Telescope

Original instruction

In WorldWide Telescope (web client), create a tour with 4 slides: M31 (FOV 3 deg), M42 (1 deg), M51 (0.5 deg) and the Crab Nebula (0.3 deg), each 8 s with a caption naming the object, using the DSS imagery layer. Save the tour as output/messier4.wtt.

Platforms
web · Any OS (web)
Applications
WorldWide Telescope
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 nav · Pan/zoom/scroll
Work product
WP05 · Time-based media
Conditions
None declared
Proposed difficulty
4 / 5

Proposed verifier

Grader parses the .wtt XML: 4 slides in order with target RA/Dec within 1 arcmin, FOV within 5%, duration 8 s, caption text exact. WWT's JS console use is blocked by CSP in the sandbox page.

Proposed partial credit

40% slide targets; 20% FOVs; 20% durations; 20% captions.

R08.01.01.02-T07 Set Celestia's date and reverse time scale

Original instruction

In Celestia, set the simulation date to Julian Date 2400000.0 and the time rate to -20x (time running backwards), then leave the view centered on Earth.

Platforms
mac · macOS win · Windows lin · Linux
Applications
Celestia
Interfaces
desk · Desktop GUI
GUI need · task setting
G2 · GUI-essential
GUI need · ordinary access
CL · CLI-sufficient
Interaction primitives
click · Point & click type · Text & forms key · Shortcuts
Work product
WP08 · App/system state
Conditions
None declared
Proposed difficulty
2 / 5

Proposed verifier

Grader dumps the state: JD within 0.01 at dump time (accounting for elapsed time at -20x), time scale exactly -20, selection Earth. Celestia's Lua/CEL console is disabled; grader reads the final state via its own cel:// URL dump.

Proposed partial credit

40% date; 40% time scale; 20% Earth centered.

R08.01.01.02-T08 Create a custom observatory with a measured horizon in Stellarium

Original instruction

Create a custom location 'Synth Ridge Obs' (38.5121 N, 109.3002 W, 1,620 m) in Stellarium, and add a polygonal landscape from the horizon survey input/horizon_az_alt.csv with the panorama input/ridge_pano.png as a spherical landscape. Make both the defaults.

Platforms
mac · macOS win · Windows lin · Linux
Applications
Stellarium
Interfaces
desk · Desktop GUI
GUI need · task setting
G2 · GUI-essential
GUI need · ordinary access
CL · CLI-sufficient
Interaction primitives
click · Point & click type · Text & forms
Work product
WP08 · App/system state
Conditions
None declared
Proposed difficulty
4 / 5

Proposed verifier

Grader: user_locations.txt entry exact; landscape.ini valid with polygonal horizon matching the CSV (max error 0.5 deg) and spherical texture; defaults set in config.ini. Stellarium's script console and RemoteControl are disabled during the session; grader enables RemoteControl afterwards to read state.

Proposed partial credit

30% location; 40% horizon polygon; 15% panorama; 15% defaults.

R08.01.01.02-T09 Frame a globular cluster with the Oculars plugin

Original instruction

In Stellarium's Oculars plugin, add telescope 'SYN 200/1200' (aperture 200 mm, focal 1,200 mm) and eyepiece 'SYN 25' (25 mm, 52 deg AFOV), select them, center M13 in the ocular view, and write the true field of view (arcmin) and magnification to output/answer.json.

Platforms
mac · macOS win · Windows lin · Linux
Applications
Stellarium
Interfaces
desk · Desktop GUI
GUI need · task setting
G2 · GUI-essential
GUI need · ordinary access
G3 · GUI-only
Interaction primitives
click · Point & click type · Text & forms
Work product
WP12 · Answer/decision
Conditions
None declared
Proposed difficulty
3 / 5

Proposed verifier

Grader: ocular.ini contains the telescope and eyepiece with exact values; ocular view active on M13; TFOV within 0.5' and magnification exact. Stellarium's script console and RemoteControl are disabled during the session; grader enables RemoteControl afterwards to read state.

Proposed partial credit

40% equipment defined; 20% M13 framed; 40% values.

R08.01.01.02-T10 Plan an observing sequence of ten Messier objects for one night

Original instruction

For 2026-09-12 from the location in input/site.json, use Stellarium (and its Observing List) to choose 10 Messier objects from input/candidates.txt that stay above 30 deg altitude for at least 30 min between 21:00 and 01:00 local, order them by transit time and save the list; export output/plan.csv with windows.

Platforms
mac · macOS win · Windows lin · Linux
Applications
Stellarium
Interfaces
desk · Desktop GUI
GUI need · task setting
HY · Hybrid
GUI need · ordinary access
G2 · GUI-essential
Interaction primitives
click · Point & click type · Text & forms vis · Visual judgment
Work product
WP02 · Records/table
Conditions
None declared
Proposed difficulty
5 / 5

Proposed verifier

Grader computes altitude windows: all 10 objects satisfy the constraint, order by transit time correct, windows within 5 min; Stellarium observing list saved with the same 10.

Proposed partial credit

40% valid objects; 30% order; 20% windows; 10% list saved.

R08.01.01.02-T11 Recreate the view of Saturn from Titan's surface in Celestia

Original instruction

In Celestia, go to the surface of Titan at latitude 0, longitude 0, set the date 2026-07-01 00:00 UTC, point at Saturn with a 40 deg field of view, turn off atmosphere haze, and save a screenshot to output/titan_view.png.

Platforms
mac · macOS win · Windows lin · Linux
Applications
Celestia
Interfaces
desk · Desktop GUI
GUI need · task setting
G2 · GUI-essential
GUI need · ordinary access
CL · CLI-sufficient
Interaction primitives
click · Point & click type · Text & forms nav · Pan/zoom/scroll
Work product
WP03 · Visual layout
Conditions
None declared
Proposed difficulty
3 / 5

Proposed verifier

Grader dumps the state: observer body Titan, surface lat/lon within 1 deg, time within 1 min, FOV within 2 deg, Saturn within 2 deg of view center; screenshot SSIM >=0.8 vs grader render. Celestia's Lua/CEL console is disabled; grader reads the final state via its own cel:// URL dump.

Proposed partial credit

30% observer position; 20% time; 30% pointing/FOV; 20% screenshot.

R08.01.01.02-T12 Time maximum eclipse and totality duration at a site

Original instruction

In Stellarium, for the 2027-08-02 total solar eclipse, set the location to Luxor, Egypt (25.6872 N, 32.6396 E) and find the local time of maximum eclipse and the duration of totality by stepping time with the GUI. Write both to output/eclipse.json.

Platforms
mac · macOS win · Windows lin · Linux
Applications
Stellarium
Interfaces
desk · Desktop GUI
GUI need · task setting
G2 · GUI-essential
GUI need · ordinary access
CL · CLI-sufficient
Interaction primitives
click · Point & click key · Shortcuts vis · Visual judgment
Work product
WP12 · Answer/decision
Conditions
None declared
Proposed difficulty
3 / 5

Proposed verifier

Grader compares with Besselian-element computation: maximum within 20 s, duration within 10 s. The VM has no internet and no ephemeris libraries (skyfield, astropy, PyEphem removed); Stellarium's script console and RemoteControl are disabled, and both answers must match the times Stellarium displayed per its UI log (grader enables RemoteControl afterwards to read state).

Proposed partial credit

20% location; 40% maximum time; 40% duration.

R08.01.01.02-T13 Build and export a sorted observing list in KStars

Original instruction

In KStars, set the location to the site in input/site.json and the date 2026-10-20, add the 8 targets in input/targets.txt to the Observation Planner's session list, sort by transit time, and export output/session.obs.

Platforms
mac · macOS win · Windows lin · Linux
Applications
KStars
Interfaces
desk · Desktop GUI
GUI need · task setting
G2 · GUI-essential
GUI need · ordinary access
CL · CLI-sufficient
Interaction primitives
click · Point & click type · Text & forms
Work product
WP02 · Records/table
Conditions
None declared
Proposed difficulty
3 / 5

Proposed verifier

Grader parses the .obs: 8 targets present in transit-time order computed by grader; site and date exact. KStars' D-Bus interface and script builder are disabled for the agent and the VM has no terminal, so the list must be built in the Observation Planner.

Proposed partial credit

40% targets; 40% order; 20% site/date.

R08.01.01.02-T14 Compute Jupiter's alt-az with Skyfield (control pair of T01)

Original instruction

Control pair of T01: with Skyfield and the DE440s ephemeris in input/, compute Jupiter's apparent altitude and azimuth from Mauna Kea (19.8207 N, 155.4681 W, 4,205 m) at 2026-12-14 03:00 UTC, with standard refraction, and write output/answer.json.

Platforms
mac · macOS win · Windows lin · Linux
Applications
Skyfield · Python
Interfaces
term · Terminal
GUI need · task setting
CL · CLI-sufficient
GUI need · ordinary access
CL · CLI-sufficient
Interaction primitives
code · Code & commands
Work product
WP12 · Answer/decision
Conditions
None declared
Proposed difficulty
3 / 5

Proposed verifier

Control pair of T01: values within 0.05 deg of reference.

Proposed partial credit

50% altitude; 50% azimuth.

R08.01.01.02-T15 Register a FITS image as a sky layer in WorldWide Telescope

Original instruction

In the WorldWide Telescope web client, load input/m101_synth.fits as an image layer, confirm it registers on the sky using its WCS (the galaxy aligned with the DSS background), set opacity 70% and a linear stretch 0-500, and save the layer set as output/m101_layers.wwtl.

Platforms
web · Any OS (web)
Applications
WorldWide Telescope
Interfaces
brow · Browser
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
WP04 · 3D/spatial model
Conditions
None declared
Proposed difficulty
4 / 5

Proposed verifier

Grader parses the layer file: image center within 2 arcsec of the WCS reference, rotation within 0.2 deg, opacity and stretch exact. WWT's JS console use is blocked by CSP in the sandbox page and the agent has no terminal, so the layer must be built in the web client.

Proposed partial credit

50% registration; 25% opacity/stretch; 25% saved layer set.

R08.01.01.02-T16 Load a spacecraft trajectory add-on and set synchronized split views

Original instruction

In Celestia, enable the add-on fixtures/addons/syn_probe/ (spacecraft with an xyz trajectory), split the window into two views: left following the probe from 50 km, right showing the Earth-Moon system from above the ecliptic. Set time to the probe's lunar flyby (find closest approach) and pause.

Platforms
mac · macOS win · Windows lin · Linux
Applications
Celestia
Interfaces
desk · Desktop GUI
GUI need · task setting
G2 · GUI-essential
GUI need · ordinary access
CL · CLI-sufficient
Interaction primitives
click · Point & click key · Shortcuts nav · Pan/zoom/scroll vis · Visual judgment
Work product
WP08 · App/system state
Conditions
None declared
Proposed difficulty
5 / 5

Proposed verifier

Grader dumps the state: two views with the specified targets/distances (10%); time within 2 min of closest approach computed from the xyz file; time paused. Celestia's Lua/CEL console is disabled; grader reads the final state via its own cel:// URL dump.

Proposed partial credit

30% add-on loaded and probe followed; 30% second view; 40% flyby time.

R08.01.01.02-T17 Time a lunar occultation of Regulus to the minute

Original instruction

In Stellarium, from the location in input/observer.json, find the next occultation of Regulus by the Moon after 2026-11-01, and determine disappearance and reappearance times (UTC) to within 1 minute by stepping time and zooming. Write both and the lunar limb (bright/dark) to output/occultation.json.

Platforms
mac · macOS win · Windows lin · Linux
Applications
Stellarium
Interfaces
desk · Desktop GUI
GUI need · task setting
G2 · GUI-essential
GUI need · ordinary access
CL · CLI-sufficient
Interaction primitives
click · Point & click key · Shortcuts nav · Pan/zoom/scroll vis · Visual judgment
Work product
WP12 · Answer/decision
Conditions
None declared
Proposed difficulty
5 / 5

Proposed verifier

Grader compares with an occultation prediction from ephemerides for that site: both times within 1 min, limb classification exact. Stellarium's script console and RemoteControl are disabled during the session; grader enables RemoteControl afterwards to read state.

Proposed partial credit

20% correct event identified; 30% disappearance; 30% reappearance; 20% limbs.

R08.01.01.02-T18 Predict a visible satellite pass from TLEs in Stellarium

Original instruction

In Stellarium's Satellites plugin, import the TLE file input/syn_sats.tle, select 'SYN-SAT-1', and from the location in input/observer.json find the next pass after 2026-10-15 18:00 UTC that is sunlit while the observer is in darkness. Report rise, culmination time, max elevation and set times in output/pass.json.

Platforms
mac · macOS win · Windows lin · Linux
Applications
Stellarium
Interfaces
desk · Desktop GUI
GUI need · task setting
G2 · GUI-essential
GUI need · ordinary access
CL · CLI-sufficient
Interaction primitives
click · Point & click type · Text & forms vis · Visual judgment
Work product
WP12 · Answer/decision
Conditions
None declared
Proposed difficulty
4 / 5

Proposed verifier

Grader propagates with SGP4: correct pass chosen; times within 30 s; max elevation within 1 deg; satellites.json contains the imported TLE. Stellarium's script console and RemoteControl are disabled during the session; grader enables RemoteControl afterwards to read state.

Proposed partial credit

20% TLE imported; 30% correct pass; 30% times; 20% max elevation.

R08.01.01.02-T19 Identify the planet closest to the Moon in Stellarium Web

Original instruction

In Stellarium Web (browser), set the observer to Cape Town and time to 2026-11-20 19:30 local, find which planet is angularly closest to the Moon, and write its name and separation (deg) to output/answer.json.

Platforms
web · Any OS (web)
Applications
Stellarium Web
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
WP12 · Answer/decision
Conditions
None declared
Proposed difficulty
3 / 5

Proposed verifier

Grader computes from ephemerides: planet exact, separation within 0.2 deg. The agent has no terminal or Python and browser devtools are blocked, so the separation must be read in Stellarium Web.

Proposed partial credit

70% planet; 30% separation.

R08.01.01.02-T20 Compare the Sun's angular size from Mars at perihelion and aphelion

Original instruction

In Celestia, place the observer on Mars and use the info text to measure the Sun's apparent angular diameter at Mars perihelion and aphelion in 2026-2027 (find the dates by advancing time). Write dates and diameters (arcmin) to output/sun_mars.json.

Platforms
mac · macOS win · Windows lin · Linux
Applications
Celestia
Interfaces
desk · Desktop GUI
GUI need · task setting
G2 · GUI-essential
GUI need · ordinary access
CL · CLI-sufficient
Interaction primitives
click · Point & click type · Text & forms key · Shortcuts
Work product
WP12 · Answer/decision
Conditions
None declared
Proposed difficulty
3 / 5

Proposed verifier

Grader computes from Mars' orbit: dates within 2 days, diameters within 0.1 arcmin. Celestia's Lua/CEL console is disabled and the agent has no terminal or Python, so values must be read from Celestia's info text.

Proposed partial credit

40% dates; 60% diameters.

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