Generated banner for Zoom General
Generated logo for Zoom General

Zoom General

OpenAIOfficialCreated by ZoomDistributed by OpenAI

  • Skill
  • Tool integration
  • Automation and integrations
  • MIT
  • Medium risk

From the author

Use when comparing products.
— the description by Zoom

AutoGineer summary

Written by AutoGineer from the skill’s own files, not by the author

Zoom General serves as a strategic starting point for developers and architects to navigate the Zoom ecosystem. It provides a structured workflow to classify integration requirements, identify the correct Zoom product surface (such as APIs, SDKs, or webhooks), and determine the necessary authentication model. By guiding users through a decision-making process, it ensures that developers select the right tool for their specific use case before routing to more specialized skills.

What it does

  • Classifies integration jobs based on desired outcomes like embedding meetings or automating resources.
  • Maps requirements to specific Zoom surfaces including REST APIs, WebSockets, and various SDKs.
  • Provides guidance on selecting the correct authentication and scope model for an integration.
  • Facilitates routing to specialized skills once the product context is established.
  • Offers comparative context for tasks that span multiple Zoom product boundaries.

When to use it

  • When you are in the early stages of planning a Zoom integration and need to choose the right app type.
  • When you need to compare different Zoom surfaces to determine which fits your technical requirements.
  • When a task involves multiple Zoom products and requires a high-level architectural approach.
  • When you need to confirm the correct authentication flow for a specific integration scenario.

Advantages

  • Prevents premature selection of tools by enforcing a structured classification process.
  • Clarifies the distinction between user-level OAuth, account-level OAuth, and SDK signatures.
  • Reduces complexity by routing users to narrow, specialized skills after initial planning.
  • Provides a clear decision matrix for choosing between General Apps, Server-to-Server OAuth, and Webhook Only apps.

Usage examples

“I want to build an app that automates meeting reports. Which Zoom app type should I use?”
The agent will guide you to classify the job as backend automation, leading to the recommendation of Server-to-Server OAuth for its lack of user interaction requirements.
“Should I use the Meeting SDK or the Video SDK for my project?”
The agent will help you compare the two based on your specific embedding needs and surface requirements, ensuring you choose the correct SDK for your use case.

Install

  1. Download the skill as a ZIP:

    https://autogineer.com/agent-skills/openai/zoom-general/download
  2. Unzip it into the cross-client skills folder of your project:

    mkdir -p .agents/skills && unzip zoom-general.zip -d .agents/skills

Documentation by Zoom

Zoom General

Use this skill for cross-product platform context after the user’s goal is understood. For first-step routing, prefer start, plan-zoom-product, or plan-zoom-integration.

Workflow

  1. Classify the job by outcome: embed meetings, custom video, automate resources, consume events, process media, use meeting intelligence, or publish a Marketplace app.

  2. Choose the Zoom surface that owns the behavior: REST API, webhooks, WebSockets, Meeting SDK, Video SDK, Zoom Apps SDK, Phone, Contact Center, Virtual Agent, Scribe, Rivet, or Cobrowse.

  3. Confirm auth and scope model before implementation; Zoom surfaces differ on user-level OAuth, account-level OAuth, and SDK signatures.

  4. Route to the narrow skill once the surface is chosen rather than keeping broad guidance in context.

  5. Use the preserved guide only when a task crosses product boundaries or needs detailed comparison tables.

References

Files

use-cases/meeting-bots.md

Meeting Bots

Build bots that join Zoom meetings for AI, transcription, and automation.

Overview

Meeting bots are headless applications that join Zoom meetings as participants to perform tasks like recording, transcription, real-time AI processing, or automated interactions.

Skills Needed

  • meeting-sdk/linux - Visible bot join flow, raw recording, and raw media access

  • zoom-rest-api - Meeting lookup plus OBF/ZAK retrieval, optional cloud-recording settings

  • zoom-webhooks - Optional if you want Zoom-managed cloud recording download after the meeting

  • zoom-rtms - Alternative when you need invisible media access instead of a visible participant bot

Architecture

Meeting Bot Architecture:
┌─────────────────┐     ┌─────────────────┐
│   Zoom Meeting  │────▶│  Bot (Linux)    │
│                 │     │  - Meeting SDK  │
│                 │◀────│  - Raw audio    │
└─────────────────┘     │  - Raw video    │
                        └────────┬────────┘
                                 │
                        ┌────────▼────────┐
                        │  AI Pipeline    │
                        │  - Transcription│
                        │  - Analysis     │
                        └─────────────────┘

Platform

PlatformRecommended
LinuxYes - headless, server-side
WindowsPossible but not typical
macOSPossible but not typical

Key Features

  • Join meetings programmatically

  • Access raw audio/video data

  • Start and stop raw recording explicitly

  • Real-time transcription

  • AI processing (sentiment, summarization)

Automatic Join + Recording Pattern

Use this chain when the user asks for a bot that automatically joins and records a meeting:

zoom-rest-api
  -> fetch meeting metadata
  -> mint OBF/ZAK token
meeting-sdk/linux
  -> join as visible participant
  -> StartRawRecording()
  -> subscribe audio/video delegates
  -> write PCM/YUV or send to downstream pipeline
optional zoom-webhooks + zoom-rest-api
  -> receive recording.completed
  -> download Zoom-managed cloud recording assets
Raw Recording Control
void onMeetingStatusChanged(MeetingStatus status, int iResult) {
    if (status != MEETING_STATUS_INMEETING) return;

    auto* recordCtrl = m_meetingService->GetMeetingRecordingController();
    if (!recordCtrl) {
        throw std::runtime_error("recording_controller_unavailable");
    }

    if (recordCtrl->CanStartRawRecording() != SDKERR_SUCCESS) {
        throw std::runtime_error("raw_recording_not_permitted");
    }

    SDKError err = recordCtrl->StartRawRecording();
    if (err != SDKERR_SUCCESS) {
        throw std::runtime_error("start_raw_recording_failed");
    }

    GetAudioRawdataHelper()->subscribe(new AudioRawDataDelegate(), true);

    IZoomSDKRenderer* renderer = nullptr;
    createRenderer(&renderer, new VideoRawDataDelegate());
    renderer->setRawDataResolution(ZoomSDKResolution_720P);
    renderer->subscribe(activeSpeakerUserId, RAW_DATA_TYPE_VIDEO);
}
Choose the Right Recording Output
RequirementCorrect path
Bot-owned audio/video files or real-time AI processingMeeting SDK Linux raw recording
Zoom-hosted MP4/M4A/transcript files after meeting endCloud recording settings + webhooks + recordings REST API

StartRawRecording() enables raw media flow. It does not create a finished MP4 by itself. You still need to persist PCM/YUV or post-process it with your own pipeline.

Bot Implementation Patterns

1. Bot Authentication Flow
// Step 1: Generate JWT for SDK authentication
void generateJWT(const string& key, const string& secret) {
    auto iat = chrono::system_clock::now();
    auto exp = iat + chrono::hours{24};

    m_jwt = jwt::create()
        .set_type("JWT")
        .set_issued_at(iat)
        .set_expires_at(exp)
        .set_payload_claim("appKey", claim(key))
        .set_payload_claim("tokenExp", claim(exp))
        .sign(algorithm::hs256{secret});
}

// Step 2: Authenticate SDK
AuthContext ctx;
ctx.jwt_token = m_jwt;
m_authService->SDKAuth(ctx);
// Wait for onAuthenticationReturn callback
2. Joining a Meeting as a Bot
JoinParam joinParam;
joinParam.userType = SDK_UT_WITHOUT_LOGIN;
JoinParam4WithoutLogin& param = joinParam.param.withoutloginuserJoin;
param.meetingNumber = meetingNumber;
param.userName = "My Transcription Bot";  // Display name
param.psw = password;
param.isVideoOff = true;   // Bots typically don't need video
param.isAudioOff = false;  // Need audio for transcription

// For own meetings: Use ZAK token
param.userZAK = zakToken;

// For external meetings (after Feb 2026): Use OBF token
param.onBehalfToken = obfToken;

err = m_meetingService->Join(joinParam);

Token Requirements:

Meeting TypeRequired Tokens
Your own meetingsJWT + ZAK
External meetings (before Feb 2026)JWT only
External meetings (after Feb 2026)JWT + OBF
3. Capturing Audio Streams
class AudioRawDataDelegate : public IZoomSDKAudioRawDataDelegate {
public:
    void onMixedAudioRawDataReceived(AudioRawData *data) override {
        // Mixed audio from all participants
        // Format: 16-bit PCM, 16kHz or 32kHz, mono

        // Send to transcription service
        transcriptionService.process(
            data->GetBuffer(),
            data->GetBufferLen(),
            data->GetSampleRate()
        );
    }

    void onOneWayAudioRawDataReceived(AudioRawData* data, uint32_t node_id) override {
        // Individual participant audio - useful for speaker identification
        speakerIdentifier.process(node_id, data);
    }
};

// Subscribe after joining meeting
auto* pRawDataHelper = GetAudioRawdataHelper();
pRawDataHelper->subscribe(new AudioRawDataDelegate());
4. Processing Video Frames
class VideoRawDataDelegate : public IZoomSDKRendererDelegate {
public:
    void onRawDataFrameReceived(YUVRawDataI420 *data) override {
        // Format: I420 (YUV 4:2:0) - contiguous planar data
        int width = data->GetStreamWidth();
        int height = data->GetStreamHeight();

        // Option 1: Save raw YUV for later processing
        yuvFile.write(data->GetYBuffer(), width * height);
        yuvFile.write(data->GetUBuffer(), (width/2) * (height/2));
        yuvFile.write(data->GetVBuffer(), (width/2) * (height/2));

        // Option 2: Convert to OpenCV for real-time processing
        // (requires copying planes into contiguous buffer first)
    }
};

// Subscribe to specific user's video
auto* pVideoHelper = GetRawdataRendererHelper();
pVideoHelper->setRawDataResolution(ZoomSDKResolution_720P);
pVideoHelper->subscribe(userId, RAW_DATA_TYPE_VIDEO, new VideoRawDataDelegate());
5. Handling Participant Events
class MeetingParticipantsDelegate : public IMeetingParticipantsCtrlEvent {
public:
    void onUserJoin(IList<unsigned int>* lstUserID, const wchar_t* strUserList) override {
        // New participants joined
        for (int i = 0; i < lstUserID->GetCount(); i++) {
            unsigned int userId = lstUserID->GetItem(i);
            auto userInfo = m_participantsCtrl->GetUserByUserID(userId);
            log("User joined: " + userInfo->GetUserName());

            // Subscribe to their video if needed
            subscribeToUserVideo(userId);
        }
    }

    void onUserLeft(IList<unsigned int>* lstUserID, const wchar_t* strUserList) override {
        // Participants left
    }

    void onHostChangeNotification(unsigned int userId) override {
        // Host changed
    }
};
6. Graceful Disconnection
void Bot::leaveMeeting() {
    // Stop raw data subscriptions
    GetAudioRawdataHelper()->unSubscribe();

    // Stop recording if active
    auto recCtl = m_meetingService->GetMeetingRecordingController();
    recCtl->StopRawRecording();

    // Leave meeting
    m_meetingService->Leave(LEAVE_MEETING);

    // Wait for onMeetingStatusChanged(MEETING_STATUS_ENDED)
}

// Handle unexpected disconnection
void onMeetingStatusChanged(MeetingStatus status, int iResult) {
    if (status == MEETING_STATUS_ENDED || status == MEETING_STATUS_FAILED) {
        cleanup();
        // Optionally reconnect
        if (shouldReconnect) {
            scheduleReconnect();
        }
    }
}

Scaling Considerations

ConsiderationRecommendation
Bot per meeting1 bot instance per meeting (SDK limitation)
Container deploymentUse Kubernetes with 1 pod per bot
Resource allocation2-4 GB RAM, 1-2 CPU cores per bot
Queue managementUse message queue (Redis, RabbitMQ) for bot assignments
Health monitoringImplement heartbeat checks for bot instances

Meeting SDK vs Video SDK for Bots

AspectMeeting SDKVideo SDK
JoinsZoom meetingsVideo SDK sessions
FeaturesFull meeting featuresCore video features
Use caseMeeting bots, recordingCustom video apps, telehealth

Choose Meeting SDK for: Joining existing Zoom meetings, transcription bots Choose Video SDK for: Custom video sessions you control, BYOS recording

Detailed Platform Guides

  • Meeting SDK Linux - Quick Start - Complete setup guide

  • High-Level Bot Scenarios - Production architectures

  • Resilient Bot Pattern - Retry logic, OBF tokens

  • Linux Platform Reference - Dependencies, Docker, troubleshooting

Specific Use Cases

Resources

Requirements and compatibility

Written for: OpenAI Codex

Works with: Claude Code, Claude apps, OpenAI Codex, GitHub Copilot, Cursor, Gemini CLI, Kiro, OpenCode, Goose, Amp, Roo Code, JetBrains Junie, Factory Droid, Devin (ex-Windsurf), Mistral Vibe, OpenHands, Letta

“Works with” means the tool reads the open Agent Skills format; a skill written for one tool may use features others lack.

Security

What our scanner noticed· Medium risk

  • Possible data-exfiltration endpoints
  • Instructions that use the network

A static scan is not a guarantee. Skills are third-party content: review a skill before you let an agent use it.

Attribution and licence

Created by
Zoom
Publisher
OpenAI (Official)
Distributed by
OpenAI
Source
openai/plugins · plugins/zoom/skills/general · commit d420d13
Licence
MIT License (redistribution allowed)
  • LICENSE (added by AutoGineer from plugins/zoom/LICENSE)
  • ATTRIBUTION.md (written by AutoGineer)
Synced

AutoGineer is not affiliated with or endorsed by Zoom or OpenAI.

Download ZIP

The ZIP holds the skill’s files, its licence files and an ATTRIBUTION.md. Download zoom-general.zip