Video-meeting session control
Run meetings: share the right window, manage participants, record, breakout rooms.
20 project-authored proposals · taxonomy v1.1 · 0 new benchmark runs. Verifiers and partial-credit rules below are designs, not executed results.
- Typical applications
- Zoom · Teams · Google Meet · Jitsi
- 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.
R11.02.01.03-T01 Join a Jitsi room muted with camera off and a display name
Original instruction
Join Jitsi Meet room 'qa-sync' on the test server meet.sandbox.test with microphone muted and camera off before entering, using display name 'Agent-1'. Stay in the room.
- Platforms
web· Any OS (web)- Applications
- Jitsi Meet
- Interfaces
brow· Browser- GUI need · task setting
G3· GUI-only- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
click· Point & clicktype· Text & forms- Work product
WP08· App/system state- Conditions
- None declared
- Proposed difficulty
- 1 / 5
Proposed verifier
Prosody/XMPP presence: participant 'Agent-1' in qa-sync with audioMuted=true and videoMuted=true from the moment of join (no unmuted media packets on the JVB for that endpoint). Browser DevTools are disabled by policy and the agent has no terminal or Prosody admin access, so room controls can only be used from the Jitsi web UI.
Proposed partial credit
30% joined right room; 20% display name; 25% muted on join; 25% camera off on join
R11.02.01.03-T02 Share only the presentation window and start the slideshow
Original instruction
In a Zoom meeting already joined (test account), share only the PowerPoint window for decks/q3_plan.pptx (not the desktop or other apps) and start the slideshow from slide 1, then advance to slide 3. Chat and email windows are open with private content and must not appear.
- Platforms
mac· macOSwin· Windows- Applications
- Zoom · PowerPoint
- Interfaces
desk· Desktop GUI- GUI need · task setting
G3· GUI-only- GUI need · ordinary access
G3· GUI-only- Interaction primitives
click· Point & clickmulti· Cross-appvis· Visual judgment- Work product
WP08· App/system state- Conditions
multiapp· Multi-apprisk· Irreversible/risky- Proposed difficulty
- 3 / 5
Proposed verifier
A verifier bot participant records the share: frames match slide 1 then slide 3 renders (SSIM ≥0.85); OCR never detects the canary strings from the chat/email windows; Zoom share type = application window. Zoom API and SDK credentials are withheld from the agent (the verifier bot is grader-only), so the share must be started from the Zoom client.
Proposed partial credit
25% window share; 25% slideshow from slide 1; 25% advanced to slide 3; 25% no private content leaked
R11.02.01.03-T03 Create and open breakout rooms from a roster
Original instruction
As host of the running Zoom meeting 'Workshop' (16 scripted participants), create 4 breakout rooms named Team A–D, assign participants per roster.csv, set rooms to close automatically after 15 minutes with a 60-second countdown, and open all rooms.
- Platforms
mac· macOSwin· Windows- Applications
- Zoom
- Interfaces
desk· Desktop GUI- GUI need · task setting
G3· GUI-only- GUI need · ordinary access
G3· GUI-only- Interaction primitives
drag· Drag & dropclick· Point & clicktype· Text & forms- Work product
WP08· App/system state- Conditions
- None declared
- Proposed difficulty
- 4 / 5
Proposed verifier
Zoom dashboard/metrics (verifier admin) and bot clients: 4 rooms with names, each bot reports its room; assignments match roster (16/16); settings auto-close 15 min and countdown 60 s; rooms opened. Compared with control pair T17. Zoom API and Server-to-Server OAuth credentials are withheld from the agent (the verifier admin token is grader-only), so actions must come from the Zoom client.
Proposed partial credit
15% rooms named; 50% assignments; 20% timing settings; 15% opened
R11.02.01.03-T04 Admit invited guests from the lobby and remove an intruder
Original instruction
As organizer of a Teams meeting in progress, admit from the lobby only the 5 people on invitees.txt, deny the others, remove the uninvited participant 'Promo Bot' who slipped in, then mute all attendees and disable their mics.
- Platforms
mac· macOSwin· Windowsweb· Any OS (web)- Applications
- Microsoft Teams
- Interfaces
desk· Desktop GUIbrow· Browser- GUI need · task setting
G3· GUI-only- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
click· Point & clicknav· Pan/zoom/scroll- Work product
WP08· App/system state- Conditions
risk· Irreversible/risky- Proposed difficulty
- 3 / 5
Proposed verifier
Teams call records / bot participants: the 5 invitees admitted, 3 non-invitees never admitted, Promo Bot removed, all attendee bots report muted with unmute disabled. Graph credentials and app registration are withheld from the agent, so meeting controls must be used in the Teams client.
Proposed partial credit
35% correct admits; 20% denied others; 20% intruder removed; 25% mute and lock
R11.02.01.03-T05 Record and transcribe a Meet call into Drive
Original instruction
In a Google Meet (Workspace test tenant) you host with two scripted participants, start recording with transcripts and captions in English, keep it running about 2 minutes while the bots talk, then stop. Confirm the recording and transcript land in My Drive > Meet Recordings.
- Platforms
web· Any OS (web)- Applications
- Google Meet
- Interfaces
brow· Browser- GUI need · task setting
G3· GUI-only- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
click· Point & clickav· Audio/video over time- Work product
WP05· Time-based media- Conditions
- None declared
- Proposed difficulty
- 3 / 5
Proposed verifier
Drive API (verifier): recording MP4 of 100–140 s and transcript Doc in Meet Recordings linked to this meeting code; transcript contains the bots' scripted sentences (WER ≤20%). The agent gets no OAuth client or Meet/Drive API token and DevTools are disabled by policy, so host controls must be used in the Meet browser session.
Proposed partial credit
35% recording started/stopped; 35% file in Drive with duration; 30% transcript present and accurate
R11.02.01.03-T06 Spotlight two speakers and hide non-video participants
Original instruction
In a Zoom meeting you host (12 scripted participants), spotlight the two panelists named in panel.txt for everyone, turn on 'Hide non-video participants' in gallery, and rename the participant 'iPhone (5)' to 'Guest - Lee'.
- Platforms
mac· macOSwin· Windows- Applications
- Zoom
- Interfaces
desk· Desktop GUI- GUI need · task setting
G3· GUI-only- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
click· Point & clickhover· Hover & contexttype· Text & forms- Work product
WP08· App/system state- Conditions
- None declared
- Proposed difficulty
- 3 / 5
Proposed verifier
Attendee bots record layout: spotlit video tiles equal the 2 panelists; host client state shows hide-non-video on (bot sees no avatar tiles); participant list shows 'Guest - Lee'. Zoom API and Server-to-Server OAuth credentials are withheld from the agent (the verifier admin token is grader-only), so actions must come from the Zoom client.
Proposed partial credit
50% correct spotlight; 20% hide non-video; 30% rename
R11.02.01.03-T07 Assign presenter roles and turn on live captions in Teams
Original instruction
In a Teams meeting you organize, open Meeting options and set 'Who can present' to Specific people, choose maria@ and tom@ as presenters (others attendees), turn on live captions with spoken language English, and pin maria@'s video for yourself.
- Platforms
mac· macOSwin· Windowsweb· Any OS (web)- Applications
- Microsoft Teams
- Interfaces
desk· Desktop GUIbrow· Browser- GUI need · task setting
G3· GUI-only- GUI need · ordinary access
G3· GUI-only- Interaction primitives
click· Point & clicktype· Text & forms- Work product
WP08· App/system state- Conditions
- None declared
- Proposed difficulty
- 3 / 5
Proposed verifier
Graph onlineMeeting: allowedPresenters roleIsPresenter with exactly maria and tom; participant roles in call; captions enabled event in call record; client telemetry shows pinned participant. Graph credentials and app registration are withheld from the agent, so meeting controls must be used in the Teams client.
Proposed partial credit
50% presenter roles; 30% captions; 20% pin
R11.02.01.03-T08 Share a video clip with computer audio
Original instruction
In a Zoom meeting, share the QuickTime/Media Player window playing media/safety_brief.mp4 with 'Share sound' enabled and 'Optimize for video clip' on, play it to the end (45 s), then stop sharing.
- Platforms
mac· macOSwin· Windows- Applications
- Zoom · QuickTime Player
- Interfaces
desk· Desktop GUI- GUI need · task setting
G3· GUI-only- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
click· Point & clickav· Audio/video over timemulti· Cross-app- Work product
WP08· App/system state- Conditions
multiapp· Multi-app- Proposed difficulty
- 3 / 5
Proposed verifier
Verifier bot recording: shared video frames match source (SSIM ≥0.7 on sampled frames) at ≥20 fps; received audio cross-correlates with the clip soundtrack (≥0.6) for ≥40 s; share stopped afterward. Zoom API and Server-to-Server OAuth credentials are withheld from the agent (the verifier admin token is grader-only), so actions must come from the Zoom client.
Proposed partial credit
30% window shared; 35% audio shared; 20% video optimized smoothness; 15% stopped after
R11.02.01.03-T09 Secure a Jitsi room with lobby and moderation
Original instruction
As moderator of Jitsi room 'board-q4' on the test server, enable the lobby, set the room password from room_pw.txt, admit only the 3 named board members waiting, kick the participant spamming chat, and enable 'Everyone starts muted' and 'Everyone starts hidden'.
- Platforms
web· Any OS (web)- Applications
- Jitsi Meet
- Interfaces
brow· Browser- GUI need · task setting
G3· GUI-only- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
click· Point & clicktype· Text & forms- Work product
WP08· App/system state- Conditions
risk· Irreversible/risky- Proposed difficulty
- 4 / 5
Proposed verifier
Prosody room config: members_only (lobby) on, password set; occupants equal moderator + 3 members; spammer kicked event; startAudioMuted/startVideoMuted policies set (new bot joins muted/hidden). Browser DevTools are disabled by policy and the agent has no terminal or Prosody admin access, so room controls can only be used from the Jitsi web UI.
Proposed partial credit
20% lobby; 15% password; 30% correct admits; 35% kick and start-muted policies
R11.02.01.03-T10 Launch a prepared poll and share the results
Original instruction
In a Zoom meeting you host, launch the prepared poll 'Pick the date' (3 options), wait until at least 8 of the 10 scripted participants have voted, end the poll, and share the results with participants.
- Platforms
mac· macOSwin· Windows- Applications
- Zoom
- Interfaces
desk· Desktop GUI- GUI need · task setting
G3· GUI-only- GUI need · ordinary access
G3· GUI-only- Interaction primitives
click· Point & click- Work product
WP08· App/system state- Conditions
- None declared
- Proposed difficulty
- 3 / 5
Proposed verifier
Zoom poll report (verifier API): poll launched and ended once with ≥8 responses; bot clients received the shared-results view; poll not relaunched. Zoom API and Server-to-Server OAuth credentials are withheld from the agent (the verifier admin token is grader-only), so actions must come from the Zoom client.
Proposed partial credit
30% launched; 30% ended after ≥8 votes; 40% results shared
R11.02.01.03-T11 Run Teams breakout rooms with a mid-session reassignment
Original instruction
In a Teams meeting with 12 scripted attendees, create 3 breakout rooms (Red/Green/Blue), assign per groups.csv, set a 10-minute time limit, open them, move 'Ana Ortiz' from Red to Blue after she asks in chat, send an announcement to all rooms, then close rooms and return everyone.
- Platforms
mac· macOSwin· Windows- Applications
- Microsoft Teams
- Interfaces
desk· Desktop GUI- GUI need · task setting
G3· GUI-only- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
drag· Drag & dropclick· Point & clicktype· Text & forms- Work product
WP08· App/system state- Conditions
interrupt· Interruptions- Proposed difficulty
- 4 / 5
Proposed verifier
Bot clients report room membership timeline: initial assignment matches CSV; Ana moved to Blue after her chat request; all bots receive announcement text; all back in main meeting after close; time limit set 10 min. Graph credentials and app registration are withheld from the agent, so meeting controls must be used in the Teams client.
Proposed partial credit
30% initial assignment; 25% reassignment; 20% announcement; 25% close/return and timer
R11.02.01.03-T12 Keep confidential pop-ups out of a live screen share
Original instruction
You are sharing your screen in a Zoom meeting to demo the expense app. During the demo, a chat app will pop up notifications with confidential text. Keep the demo going (open three expense reports in order) while ensuring none of the notification content is visible to attendees.
- Platforms
mac· macOSwin· Windows- Applications
- Zoom · Expense app
- Interfaces
desk· Desktop GUI- GUI need · task setting
G3· GUI-only- GUI need · ordinary access
G3· GUI-only- Interaction primitives
click· Point & clickvis· Visual judgmentmulti· Cross-app- Work product
WP08· App/system state- Conditions
interrupt· Interruptionsrisk· Irreversible/riskymultiapp· Multi-app- Proposed difficulty
- 4 / 5
Proposed verifier
Verifier bot recording: OCR never shows the notification canary strings; three report pages appear in order; acceptable strategies: window-only share, Focus/DND on, or pause share during pop-up (gap ≤10 s). Zoom API and Server-to-Server OAuth credentials are withheld from the agent (the verifier admin token is grader-only), so actions must come from the Zoom client.
Proposed partial credit
50% no leak; 35% three reports demoed; 15% share continuity
R11.02.01.03-T13 Set host controls in Google Meet and end for all
Original instruction
In a Google Meet you host, turn on Host management, turn off chat messages and screen sharing for everyone except hosts, then end the call for everyone at exactly the 5-minute mark of the meeting.
- Platforms
web· Any OS (web)- Applications
- Google Meet
- Interfaces
brow· Browser- GUI need · task setting
G3· GUI-only- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
click· Point & click- Work product
WP08· App/system state- Conditions
- None declared
- Proposed difficulty
- 2 / 5
Proposed verifier
Meet REST/report (verifier): host management on; chat and present locks on; bot participant's attempt to chat/share fails; conference end time 5:00 ±20 s after start, all participants disconnected. The agent gets no OAuth client or Meet/Drive API token and DevTools are disabled by policy, so host controls must be used in the Meet browser session.
Proposed partial credit
25% host management; 40% locks; 35% ended for all on time
R11.02.01.03-T14 Moderate a webinar Q&A and promote a panelist
Original instruction
In a Zoom Webinar (test account) as host, answer live the 5 attendee questions that match faq.md (type answers publicly), dismiss the 2 off-topic spam questions, and promote attendee 'Dr. Okafor' to panelist when she asks in Q&A.
- Platforms
mac· macOSwin· Windows- Applications
- Zoom
- Interfaces
desk· Desktop GUI- GUI need · task setting
G3· GUI-only- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
type· Text & formsclick· Point & clicknav· Pan/zoom/scroll- Work product
WP08· App/system state- Conditions
interrupt· Interruptions- Proposed difficulty
- 4 / 5
Proposed verifier
Webinar Q&A report: 5 questions answered publicly with text containing FAQ key facts; 2 spam dismissed; no other dismissals; participant report shows Okafor role changed to panelist after her request. Zoom API and Server-to-Server OAuth credentials are withheld from the agent (the verifier admin token is grader-only), so actions must come from the Zoom client.
Proposed partial credit
50% answers; 20% spam dismissed; 30% promotion
R11.02.01.03-T15 Present with PowerPoint Live and hand over control
Original instruction
In a Teams meeting, share decks/roadmap.pptx via PowerPoint Live, use presenter view to go to slide 4, turn off attendees' ability to browse slides privately, then give control to co-presenter tom@ and take it back after he advances two slides.
- Platforms
mac· macOSwin· Windows- Applications
- Microsoft Teams · PowerPoint
- Interfaces
desk· Desktop GUI- GUI need · task setting
G3· GUI-only- GUI need · ordinary access
G3· GUI-only- Interaction primitives
click· Point & clicknav· Pan/zoom/scrollmulti· Cross-app- Work product
WP08· App/system state- Conditions
multiapp· Multi-app- Proposed difficulty
- 4 / 5
Proposed verifier
Attendee bot captures: PowerPoint Live share at slide 4; private navigation disabled (bot browsing blocked); control granted to tom@ (bot advances to slide 6) then reclaimed; final presenter is agent. Graph credentials and app registration are withheld from the agent, so meeting controls must be used in the Teams client.
Proposed partial credit
25% PowerPoint Live; 15% slide 4; 20% private browsing off; 40% control handoff and return
R11.02.01.03-T16 Record locally with separate audio tracks per participant
Original instruction
In Zoom desktop settings enable local recording with 'Record a separate audio file for each participant' and save location ~/Zoom/recordings. In the running meeting, record 60 s while three scripted participants speak in turn, then stop and wait for conversion.
- Platforms
mac· macOSwin· Windows- Applications
- Zoom
- Interfaces
desk· Desktop GUI- GUI need · task setting
G3· GUI-only- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
click· Point & clickav· Audio/video over time- Work product
WP05· Time-based media- Conditions
- None declared
- Proposed difficulty
- 3 / 5
Proposed verifier
Recording folder contains MP4 (55–70 s) and per-participant audio files (≥3), each dominated by its speaker's scripted audio (energy only in their turns, correlation ≥0.6). No terminal, ffmpeg or Zoom API credentials for the agent; recording files written by any process other than Zoom are caught by a file-write audit log.
Proposed partial credit
30% setting enabled; 30% recording made; 40% separate per-participant tracks
R11.02.01.03-T17 Pre-assign breakout rooms via the Zoom API (control pair of T03)
Original instruction
Control pair of T03. With Zoom Server-to-Server OAuth test credentials, write prebreakout.py that updates meeting 'Workshop' with 4 pre-assigned breakout rooms per roster.csv (by email) before it starts.
- Platforms
mac· macOSwin· Windowslin· Linux- Applications
- Zoom API
- 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
WP08· App/system state- Conditions
- None declared
- Proposed difficulty
- 2 / 5
Proposed verifier
GET meeting settings.breakout_room: 4 rooms named Team A–D with participant emails matching roster; bot clients land in assigned rooms when host opens them.
Proposed partial credit
40% room config; 60% correct assignment
R11.02.01.03-T18 Sketch on the meeting whiteboard during a Jitsi call
Original instruction
In Jitsi room 'arch-review', open the shared whiteboard and draw three labeled boxes 'Client', 'API', 'DB' connected left-to-right by arrows, so participants can see it, then close the whiteboard for everyone.
- Platforms
web· Any OS (web)- Applications
- Jitsi Meet
- Interfaces
brow· Browser- GUI need · task setting
G3· GUI-only- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
drag· Drag & droptype· Text & formspath· Continuous path- Work product
WP03· Visual layout- Conditions
- None declared
- Proposed difficulty
- 3 / 5
Proposed verifier
Whiteboard (Excalidraw collab) scene snapshot captured by bot: 3 rectangles with exact labels ordered left→right, 2 arrows binding Client→API and API→DB; whiteboard closed event afterwards. Browser DevTools are disabled by policy and the agent has no terminal or Prosody admin access, so room controls can only be used from the Jitsi web UI.
Proposed partial credit
40% three labeled boxes; 40% arrows bound correctly; 20% closed after
R11.02.01.03-T19 Run a moderated town hall end to end
Original instruction
Run 'October Town Hall' (Teams town hall, test tenant) per runbook.md: start on time, spotlight the CEO, run moderated Q&A (publish 6 approved questions, reject 3), show the prepared slide deck, record, end at the scheduled time, and download the attendance report to reports/.
- Platforms
mac· macOSwin· Windows- Applications
- Microsoft Teams
- Interfaces
desk· Desktop GUI- GUI need · task setting
G3· GUI-only- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
click· Point & clicknav· Pan/zoom/scrollmulti· Cross-app- Work product
WP08· App/system state- Conditions
long· Long-horizoninterrupt· Interruptions- Proposed difficulty
- 5 / 5
Proposed verifier
Event logs: started within 1 min; spotlight on CEO; Q&A states: 6 published, 3 rejected per runbook criteria; deck shared; recording exists; ended ±1 min; attendance CSV in reports/ matches bot roster. Graph credentials and app registration are withheld from the agent, so meeting controls must be used in the Teams client.
Proposed partial credit
30% start/end timing and spotlight; 35% Q&A moderation; 15% deck; 20% recording and report
R11.02.01.03-T20 Host a timed training session following a rundown
Original instruction
Host a Zoom training per rundown.csv: share window A (terminal) at 00:00 for 3 min, switch share to window B (browser docs) without stopping the meeting, at 06:00 launch 3 breakout rooms (auto-assign) for 5 min, bring everyone back, run poll 'Recap', and end at 15:00.
- Platforms
mac· macOSwin· Windows- Applications
- Zoom · Terminal · Browser
- Interfaces
desk· Desktop GUI- GUI need · task setting
G3· GUI-only- GUI need · ordinary access
G3· GUI-only- Interaction primitives
click· Point & clickrt· Real-time controlmulti· Cross-app- Work product
WP08· App/system state- Conditions
multiapp· Multi-applong· Long-horizon- Proposed difficulty
- 5 / 5
Proposed verifier
Bot recording and Zoom metrics: share content switches at 3:00 ±20 s with no desktop shown; breakout opened at 6:00 ±30 s for 5 min ±30 s; poll launched after return; meeting ends at 15:00 ±30 s. Zoom API and Server-to-Server OAuth credentials are withheld from the agent (the verifier admin token is grader-only), so actions must come from the Zoom client.
Proposed partial credit
25% share sequence; 25% breakout timing; 20% poll; 30% end time and no desktop exposure
Mapped research references
No benchmark reference is assigned to this class in the current mapping.