Computer-Use Agent Taxonomy中文
R06.12.01.02 · TASK CLASS

Live switching & broadcast graphics

Run a switcher and broadcast graphics through a rundown.

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

Typical applications
ATEM Software Control · vMix · CasparCG
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.

R06.12.01.02-T01 Switch a five-cue rundown live on an ATEM

Original instruction

ATEM Software Control is connected to the test rig's ATEM Mini Pro (inputs 1–4 carry timecoded test feeds). Run rundown_A.pdf live: at the listed timecodes cut to Cam 2, mix (1 s) to Cam 3, cut to Cam 1, wipe to Cam 4, fade to black. The rundown clock starts when you press Start in the provided cue display.

Platforms
mac · macOS win · Windows
Applications
ATEM Software Control
Interfaces
desk · Desktop GUI
GUI need · task setting
G3 · GUI-only
GUI need · ordinary access
G3 · GUI-only
Interaction primitives
click · Point & click rt · Real-time control av · Audio/video over time
Work product
WP05 · Time-based media
Conditions
None declared
Proposed difficulty
3 / 5

Proposed verifier

Rig logs switcher state changes and captures program output. Each cue: correct source and transition type; timing within ±10 frames of rundown TC. ATEM SDK/TCP clients other than the GUI are firewalled; extra program changes penalised.

Proposed partial credit

40% correct sources in order; 30% transition types; 30% timing within tolerance

R06.12.01.02-T02 Configure an upstream chroma key for a presenter camera

Original instruction

In ATEM Software Control set up Upstream Key 1 as an Advanced Chroma key with fill and key from Cam 3 (presenter on green), over background Cam 4 (set graphic). Tune sample, foreground, background and edge so the presenter's hair edges hold and no green spill remains, then put the key on air.

Platforms
mac · macOS win · Windows
Applications
ATEM Software Control
Interfaces
desk · Desktop GUI
GUI need · task setting
G3 · GUI-only
GUI need · ordinary access
G2 · GUI-essential
Interaction primitives
vis · Visual judgment click · Point & click drag · Drag & drop
Work product
WP05 · Time-based media
Conditions
None declared
Proposed difficulty
3 / 5

Proposed verifier

Program capture frames vs GT alpha of the test feed: foreground IoU ≥0.94; green residue ≤0.3% of background pixels; spill metric in edge band ≤ threshold. Switcher state: USK1 type chroma, on air, sources 3/3. Agent VM has no terminal, ATEM SDK or third-party control tools, and the rig observer logs any UDP 9910 control client other than ATEM Software Control.

Proposed partial credit

25% key routing and on air; 45% matte quality; 30% spill and residue

R06.12.01.02-T03 Load a lower-third still into the media pool and key it downstream

Original instruction

In ATEM Software Control upload graphics/lt_guest.png to Media Pool still slot 3, assign Media Player 1 to it, set Downstream Key 1 fill/key from Media Player 1 (pre-multiplied), and set DSK1 rate to 1 s. Take DSK1 on air with Auto, hold 6 s, then take it off.

Platforms
mac · macOS win · Windows
Applications
ATEM Software Control
Interfaces
desk · Desktop GUI
GUI need · task setting
G3 · GUI-only
GUI need · ordinary access
AP · API-sufficient
Interaction primitives
drag · Drag & drop click · Point & click
Work product
WP08 · App/system state
Conditions
None declared
Proposed difficulty
2 / 5

Proposed verifier

Switcher state: media slot 3 hash equals PNG, MP1 source still 3, DSK1 premultiplied with rate 25/30 frames. Program capture shows lower third fade in over 1 s, held 6±1 s, fade out. Agent VM has no terminal, ATEM SDK or third-party control tools, and the rig observer logs any UDP 9910 control client other than ATEM Software Control.

Proposed partial credit

25% upload; 25% DSK config; 50% on-air sequence

R06.12.01.02-T04 Build a two-up interview layout with SuperSource

Original instruction

On the rig's ATEM Constellation (SuperSource available), configure SuperSource with Box 1 = Cam 1 and Box 2 = Cam 2, each 45% size, centred vertically, at x = -8.0 and +8.0, 2-px white borders, background art from Media Player 2. Put SuperSource on program.

Platforms
mac · macOS win · Windows
Applications
ATEM Software Control
Interfaces
desk · Desktop GUI
GUI need · task setting
G3 · GUI-only
GUI need · ordinary access
CL · CLI-sufficient
Interaction primitives
geom · Precise geometry click · Point & click vis · Visual judgment
Work product
WP08 · App/system state
Conditions
None declared
Proposed difficulty
3 / 5

Proposed verifier

Switcher state: SuperSource boxes enabled, sources, size 0.45, positions ±0.1, border width and color; art source MP2; program = SuperSource. Program frame capture: box rectangles detected at expected pixel coordinates ±4 px. Agent VM has no terminal, ATEM SDK or third-party control tools, and the rig observer logs any UDP 9910 control client other than ATEM Software Control.

Proposed partial credit

40% box sources/positions; 20% borders; 15% art; 25% on program and verified in capture

R06.12.01.02-T05 Record and run an Open Show macro

Original instruction

In ATEM Software Control record macro 'Open Show' in slot 1: fade from black (1 s), mix to Cam 1, wait 2 s, DSK 1 auto on, wait 5 s, DSK 1 auto off. Save the macro and run it once.

Platforms
mac · macOS win · Windows
Applications
ATEM Software Control
Interfaces
desk · Desktop GUI
GUI need · task setting
G3 · GUI-only
GUI need · ordinary access
CL · CLI-sufficient
Interaction primitives
click · Point & click rt · Real-time control
Work product
WP08 · App/system state
Conditions
None declared
Proposed difficulty
3 / 5

Proposed verifier

Macro pool slot 1 named 'Open Show' with ordered ops (FTB, mix to input 1, wait 2 s, DSK1 auto, wait 5 s, DSK1 auto). Run log: program sequence timing within ±3 frames of waits. Agent VM has no terminal, ATEM SDK or third-party control tools, and the rig observer logs any UDP 9910 control client other than ATEM Software Control.

Proposed partial credit

40% macro steps; 20% waits; 40% runs correctly

R06.12.01.02-T06 Set up Fairlight audio for mics and camera follow

Original instruction

In ATEM Software Control's Fairlight audio page: on Mic 1 and Mic 2 enable high-pass 80 Hz, a compressor (threshold -20 dB, ratio 3:1), and set them always on; set camera inputs 1–4 to Audio Follow Video; set master limiter -3 dBFS. Save the switcher startup state.

Platforms
mac · macOS win · Windows
Applications
ATEM Software Control
Interfaces
desk · Desktop GUI
GUI need · task setting
G3 · GUI-only
GUI need · ordinary access
AP · API-sufficient
Interaction primitives
click · Point & click drag · Drag & drop
Work product
WP08 · App/system state
Conditions
None declared
Proposed difficulty
3 / 5

Proposed verifier

Switcher Fairlight state (via rig observer) per input: EQ HPF 80 Hz, dynamics settings, mix option On / AFV; master limiter -3. Startup state saved (reboot of emulator retains values). Agent VM has no terminal, ATEM SDK or third-party control tools, and the rig observer logs any UDP 9910 control client other than ATEM Software Control.

Proposed partial credit

35% mic processing; 30% AFV on cameras; 15% limiter; 20% saved startup

R06.12.01.02-T07 Direct a three-camera interview with timed lower thirds in vMix

Original instruction

vMix preset interview.vmix has CAM1–3 (playback of a recorded 4-minute interview from three angles) and title LT_Guest. Direct it live: show whoever is speaking (CAM2 host, CAM3 guest, CAM1 wide on overlap), bring LT_Guest in at the guest's first answer and out after 6 s, then once more at 3:00.

Platforms
win · Windows
Applications
vMix
Interfaces
desk · Desktop GUI
GUI need · task setting
G3 · GUI-only
GUI need · ordinary access
G2 · GUI-essential
Interaction primitives
click · Point & click rt · Real-time control av · Audio/video over time
Work product
WP05 · Time-based media
Conditions
licensed · Licensed pro app
Proposed difficulty
4 / 5

Proposed verifier

Program recording vs speaker-diarisation ground truth: ≥80% of time on correct camera, no shots <1.5 s. Lower third appearances within ±1 s of targets, held 6±1 s. vMix HTTP API port blocked to the agent; activators logged.

Proposed partial credit

50% camera-speaker agreement; 15% no flash cuts; 35% lower thirds timed

R06.12.01.02-T08 Drive a live scoreboard from a data source during a match

Original instruction

In vMix create a scoreboard GT title with Home/Away names and scores, bind fields to data source data/score.csv (vMix Data Sources, auto-refresh 1 s) and keep it on overlay 1 during the 10-minute match playback. The CSV will change 5 times; also manually correct one wrong value flagged by the referee message at 06:00.

Platforms
win · Windows
Applications
vMix
Interfaces
desk · Desktop GUI
GUI need · task setting
G3 · GUI-only
GUI need · ordinary access
G3 · GUI-only
Interaction primitives
click · Point & click type · Text & forms rt · Real-time control
Work product
WP05 · Time-based media
Conditions
licensed · Licensed pro app interrupt · Interruptions
Proposed difficulty
4 / 5

Proposed verifier

OCR on program recording every second: displayed scores match expected timeline within 2 s of each CSV change; corrected value appears within 30 s of 06:00 and persists. Title bindings exist in preset. vMix Web Controller (HTTP API) and TCP API are disabled, and no vMix Scripts may be added (Settings > Scripting list checked).

Proposed partial credit

25% title and bindings; 45% scores track changes; 30% manual correction

R06.12.01.02-T09 Cut a slow-motion replay of a goal and return to live

Original instruction

In vMix with Replay enabled on CAM1/CAM2 ISO feeds of a match playback, when the goal occurs (around 02:10) mark in 4 s before and out 2 s after on CAM2, play the replay at 50% with the Replay stinger, then return to live CAM1 within 20 s of the goal.

Platforms
win · Windows
Applications
vMix
Interfaces
desk · Desktop GUI
GUI need · task setting
G3 · GUI-only
GUI need · ordinary access
G3 · GUI-only
Interaction primitives
click · Point & click rt · Real-time control av · Audio/video over time
Work product
WP05 · Time-based media
Conditions
licensed · Licensed pro app
Proposed difficulty
4 / 5

Proposed verifier

Program recording: replay segment content matches CAM2 source frames from goal−4 s to goal+2 s (frame matching) at 0.5x speed (12 s ±1); stinger precedes it; program back on live CAM1 within 20 s of goal plus replay duration. vMix Web Controller (HTTP API) and TCP API are disabled, and no vMix Scripts may be added (Settings > Scripting list checked).

Proposed partial credit

35% correct in/out; 25% half-speed playback; 15% stinger; 25% return to live

R06.12.01.02-T10 Build and run a CasparCG Client rundown with templates

Original instruction

In CasparCG Client connected to the test CasparCG Server, build rundown show.xml: play clips/opener.mov on channel 1 layer 10 with MIX 25 frames, an HTML lower-third template on layer 20 with f0='Dr. Sample', f1='Guest Researcher', a looping logo bug on layer 30, then clear layer 20 after 8 s. Save and run it.

Platforms
win · Windows lin · Linux
Applications
CasparCG Client
Interfaces
desk · Desktop GUI
GUI need · task setting
G3 · GUI-only
GUI need · ordinary access
G3 · GUI-only
Interaction primitives
drag · Drag & drop type · Text & forms click · Point & click
Work product
WP05 · Time-based media
Conditions
None declared
Proposed difficulty
3 / 5

Proposed verifier

Saved rundown XML items with channels/layers/params. Server log (AMCP received only from the client app) shows commands in order; channel output capture: opener mix-in, lower-third text OCR matches, logo bug loops, layer 20 cleared after 8±1 s.

Proposed partial credit

35% rundown items correct; 40% playout as specified; 25% saved rundown

R06.12.01.02-T11 Update a live lower third with corrected data mid-show

Original instruction

During CasparCG playout of rundown news.xml, the producer message at cue 4 says the guest title is wrong. Without taking the template off air, use the client's Update with f1='Chief Hydrologist' at cue 4, then at cue 6 swap to the second guest's data and play the template out.

Platforms
win · Windows lin · Linux
Applications
CasparCG Client
Interfaces
desk · Desktop GUI
GUI need · task setting
G3 · GUI-only
GUI need · ordinary access
CL · CLI-sufficient
Interaction primitives
type · Text & forms click · Point & click rt · Real-time control
Work product
WP05 · Time-based media
Conditions
interrupt · Interruptions
Proposed difficulty
3 / 5

Proposed verifier

Server AMCP log: CG UPDATE on the right layer with f1 value at cue 4 (no CG STOP before it); CG ADD/UPDATE for guest 2 at cue 6 then CG STOP. Output OCR shows text change without a blank frame. The agent has no terminal or telnet; the server accepts AMCP only from the CasparCG Client host, and its log flags any other connection.

Proposed partial credit

40% update without off-air; 35% guest 2 swap; 25% clean out

R06.12.01.02-T12 Configure a 10-window multiview with meters and safe areas

Original instruction

In ATEM Software Control set Multiview 1 to the layout in mv_layout.png: preview/program top, then Cam 1–4, Media Player 1, SuperSource, Clean Feed 1 and Color Bars in the small windows; enable audio meters on all windows and safe-area markers on preview.

Platforms
mac · macOS win · Windows
Applications
ATEM Software Control
Interfaces
desk · Desktop GUI
GUI need · task setting
G3 · GUI-only
GUI need · ordinary access
G2 · GUI-essential
Interaction primitives
click · Point & click drag · Drag & drop
Work product
WP08 · App/system state
Conditions
None declared
Proposed difficulty
2 / 5

Proposed verifier

Switcher multiview state: layout orientation, 8 window sources in exact positions, meters on per window, safe area on. Multiview output capture labels OCR confirm positions. Agent VM has no terminal, ATEM SDK or third-party control tools, and the rig observer logs any UDP 9910 control client other than ATEM Software Control.

Proposed partial credit

60% window sources; 20% meters; 20% safe area

R06.12.01.02-T13 Stream and record a show from an ATEM Mini Pro

Original instruction

In ATEM Software Control output settings, stream to custom server rtmp://127.0.0.1/live (key show42) at 1080p30 'Streaming High' quality, and record program to the attached USB disk with file name Show42. Start both, run the show for 60 s, stop both.

Platforms
mac · macOS win · Windows
Applications
ATEM Software Control
Interfaces
desk · Desktop GUI
GUI need · task setting
G3 · GUI-only
GUI need · ordinary access
AP · API-sufficient
Interaction primitives
click · Point & click type · Text & forms
Work product
WP05 · Time-based media
Conditions
None declared
Proposed difficulty
3 / 5

Proposed verifier

Local RTMP ingest recorded 60±5 s at 1080p30 with bitrate in the Streaming High range; USB disk image holds Show42*.mp4 of matching duration. Streaming config XML exported equals settings. Agent VM has no terminal, ATEM SDK or third-party control tools, and the rig observer logs any UDP 9910 control client other than ATEM Software Control.

Proposed partial credit

45% streaming verified at ingest; 40% recording file; 15% settings

R06.12.01.02-T14 Set up a vMix virtual set with presenter zoom presets

Original instruction

In vMix add input VirtualSet 'NewsDesk', key the presenter camera CAM1 (green screen) into it, and create zoom presets Wide, Medium and Close; show them on program in the order in cue_vs.txt at roughly 10 s intervals during a 40 s segment.

Platforms
win · Windows
Applications
vMix
Interfaces
desk · Desktop GUI
GUI need · task setting
G3 · GUI-only
GUI need · ordinary access
G3 · GUI-only
Interaction primitives
vis · Visual judgment click · Point & click rt · Real-time control
Work product
WP05 · Time-based media
Conditions
licensed · Licensed pro app
Proposed difficulty
3 / 5

Proposed verifier

Preset: virtual set input with CAM1 keyed (chroma key settings present). Program recording: presenter keyed cleanly (residual green ≤0.5%); three framings identified by presenter bbox size in expected order with ≥8 s holds. vMix Web Controller (HTTP API) and TCP API are disabled, and no vMix Scripts may be added (Settings > Scripting list checked).

Proposed partial credit

30% virtual set and key; 20% key quality; 50% preset framings in order

R06.12.01.02-T15 Follow the ball live by switching between six game cameras

Original instruction

vMix preset match6.vmix plays six synchronized camera angles of a 5-minute 5-a-side match. Direct it live so the program shows a camera where the ball is visible and reasonably large; avoid cuts faster than 2 s and jump cuts between adjacent angles. No pausing or slow-motion.

Platforms
win · Windows
Applications
vMix
Interfaces
desk · Desktop GUI
GUI need · task setting
G3 · GUI-only
GUI need · ordinary access
G2 · GUI-essential
Interaction primitives
vis · Visual judgment rt · Real-time control click · Point & click
Work product
WP05 · Time-based media
Conditions
licensed · Licensed pro app
Proposed difficulty
5 / 5

Proposed verifier

Program recording aligned to sources; ball-visibility ground truth per camera per frame. Score: % of frames where program camera shows the ball (≥70% to pass), mean ball size rank, cut-rate violations. API port blocked; playback cannot be paused (rig enforces).

Proposed partial credit

60% ball-visible coverage; 20% best-angle rank; 20% cut discipline

R06.12.01.02-T16 Recover from a camera failure during a live show

Original instruction

Run rundown_B.pdf on the ATEM (Cam 1 main, Cam 3 backup of the same shot). At an unannounced moment Cam 1 drops to black; switch to Cam 3 within 2 s. If both main and backup fail, put Media Player 2 slate 'Technical difficulties' on air with DSK off and log the time in incident.txt.

Platforms
mac · macOS win · Windows
Applications
ATEM Software Control
Interfaces
desk · Desktop GUI
GUI need · task setting
G3 · GUI-only
GUI need · ordinary access
G2 · GUI-essential
Interaction primitives
vis · Visual judgment rt · Real-time control click · Point & click
Work product
WP05 · Time-based media
Conditions
interrupt · Interruptions
Proposed difficulty
5 / 5

Proposed verifier

Program capture: black-on-air time after Cam 1 failure ≤2 s; after double failure slate on program within 3 s and DSK off; remaining rundown cues followed (±10 frames). incident.txt time within 5 s of injected failure. Agent VM has no terminal, ATEM SDK or third-party control tools, and the rig observer logs any UDP 9910 control client other than ATEM Software Control.

Proposed partial credit

35% backup switch latency; 30% slate on double failure; 20% rundown continuity; 15% incident log

R06.12.01.02-T17 Script the interview rundown via the vMix HTTP API (control pair of T07)

Original instruction

Control pair of T07. Given the speaker-turn timestamps in turns.csv, write rundown.py that drives vMix via its HTTP API (Cut, OverlayInput1In/Out) to run the same interview with lower-third timing, starting when the playback begins.

Platforms
win · Windows
Applications
vMix
Interfaces
api · API/SDK
GUI need · task setting
AP · API-sufficient
GUI need · ordinary access
AP · API-sufficient
Interaction primitives
code · Code & commands
Work product
WP05 · Time-based media
Conditions
licensed · Licensed pro app
Proposed difficulty
2 / 5

Proposed verifier

Same program-recording scoring as T07; vMix API log shows calls from the script only.

Proposed partial credit

50% camera-speaker agreement; 20% no flash cuts; 30% lower thirds

R06.12.01.02-T18 Program a 30-minute CasparCG playout block with timed breaks

Original instruction

In CasparCG Client build a rundown for a 30-minute block from schedule.xlsx: programme segments, two ad breaks starting at exactly 10:00 and 20:00 (each 3 spots), station ident before each break, and a continuous clock bug template. Run it from 00:00 against the client clock.

Platforms
win · Windows lin · Linux
Applications
CasparCG Client
Interfaces
desk · Desktop GUI
GUI need · task setting
G3 · GUI-only
GUI need · ordinary access
G3 · GUI-only
Interaction primitives
drag · Drag & drop type · Text & forms rt · Real-time control
Work product
WP05 · Time-based media
Conditions
long · Long-horizon multiapp · Multi-app
Proposed difficulty
4 / 5

Proposed verifier

Output capture and server log: item order matches schedule; breaks start at 10:00 and 20:00 ±2 frames (auto-step with timecodes, not manual); idents precede breaks; clock bug present throughout; total duration 30:00 ±1 s. The agent has no terminal or telnet; the server accepts AMCP only from the CasparCG Client host, and its log flags any other connection.

Proposed partial credit

30% order; 35% break timing; 15% idents; 20% clock bug and duration

R06.12.01.02-T19 Build DVE and stinger transitions and use them on air

Original instruction

In ATEM Software Control create a DVE push transition (left, 20-frame rate, 4-px yellow border), and a stinger using clip 1 from the Media Pool (clip duration 45 frames, trigger at frame 15, mix rate 10). Use DVE for Cam1→Cam2 and stinger for Cam2→Cam3 on air.

Platforms
mac · macOS win · Windows
Applications
ATEM Software Control
Interfaces
desk · Desktop GUI
GUI need · task setting
G3 · GUI-only
GUI need · ordinary access
CL · CLI-sufficient
Interaction primitives
click · Point & click type · Text & forms vis · Visual judgment
Work product
WP05 · Time-based media
Conditions
None declared
Proposed difficulty
4 / 5

Proposed verifier

Switcher transition settings match; program capture: first transition shows push motion with yellow border over 20 frames; second shows stinger clip frames with source switch at frame 15±1. Agent VM has no terminal, ATEM SDK or third-party control tools, and the rig observer logs any UDP 9910 control client other than ATEM Software Control.

Proposed partial credit

30% DVE settings; 30% stinger settings; 40% both executed correctly on air

R06.12.01.02-T20 Run a ten-minute live show from a rundown across switcher and graphics

Original instruction

Run show_rundown.xlsx (24 cues, 10 minutes) live: switching on the ATEM via ATEM Software Control, lower thirds and full-screen graphics from CasparCG Client fed into ATEM input 5, audio mic changes on Fairlight, and one producer change mid-show delivered as a message at 05:30 (swap cues 17 and 18).

Platforms
mac · macOS win · Windows
Applications
ATEM Software Control · CasparCG Client
Interfaces
desk · Desktop GUI
GUI need · task setting
G3 · GUI-only
GUI need · ordinary access
G2 · GUI-essential
Interaction primitives
click · Point & click rt · Real-time control multi · Cross-app av · Audio/video over time
Work product
WP05 · Time-based media
Conditions
multiapp · Multi-app long · Long-horizon interrupt · Interruptions
Proposed difficulty
5 / 5

Proposed verifier

Rig logs both systems and records program. Per cue: right element, timing ±15 frames; swapped cues executed in new order; no black or wrong-source frames >0.5 s. Score = weighted cue accuracy; API ports firewalled.

Proposed partial credit

50% cue accuracy; 20% timing; 15% mid-show change; 15% no on-air errors

Mapped research references

No benchmark reference is assigned to this class in the current mapping.

In this workflow family