スキル一覧に戻る
arthurelgindell

gemini-super-gems

by arthurelgindell

Arthur's AI-powered content creation toolkit: video generation, carousel creation, image generation, and automation scripts

0🍴 0📅 2026年1月15日
GitHubで見るManusで実行

SKILL.md


name: gemini-super-gems description: This skill should be used when the user asks to "create Super Gems workflows", "generate videos with Veo", "automate Gemini video production", "use Google Gemini for video", or needs guidance on no-code AI video automation, Veo integration, multi-stage content pipelines, or bidirectional media database integration with Google Gemini. version: 1.0.0 license: MIT

Gemini Super Gems Video Production Integration

Comprehensive integration for Google Gemini Super Gems with Veo video generation and MediaDB integration.

Quick Reference

Access Points:

  • Gemini URL: https://gemini.google.com
  • Gem Manager: Accessible from Gemini sidebar
  • Requires: Google account with Gemini Ultra subscription

Super Gems Capabilities:

  • No-code workflow automation via natural language
  • Veo video generation integration
  • Multi-stage pipelines: scripting → visual → audio → export

Super Gems Overview

Super Gems are UI-based no-code automation workflows in the Gemini web interface. Unlike standard prompts, Super Gems can orchestrate multi-step processes automatically.

Key Features

FeatureDescription
No-Code AutomationBuild workflows in natural language
Veo IntegrationGenerate videos directly within workflows
Multi-Stage PipelinesChain scripting, visual, audio, export
Gem ManagerCreate, edit, organize Super Gems

Access Requirements

  1. Gemini Ultra subscription (or Google Workspace with Gemini access)
  2. Super Gems enabled (check Gemini Labs if not visible)
  3. Eligible region (primarily US, expanding globally)

Workflow Patterns

4-Stage Video Production Pipeline

Stage 1: Scripting
↓
Stage 2: Visual Generation (Veo)
↓
Stage 3: Audio Overlay
↓
Stage 4: Export & Ingest

Stage 1: Scripting

Create the narrative or script for the video.

Example prompts:

  • "Write a 30-second script for a product demo showcasing [product features]"
  • "Create a voiceover script for a travel video about [destination]"
  • "Generate a tutorial script explaining [concept] in simple terms"

Stage 2: Visual Generation (Veo)

Veo generates video content based on the script and style parameters.

Parameters:

ParameterOptions
Duration5-60 seconds typical
StyleCorporate, cinematic, animated, documentary
ResolutionUp to 1080p
Generation Time5-15 minutes typical

Style prompts:

  • "Corporate, modern, clean aesthetic"
  • "Cinematic with dramatic lighting"
  • "Animated explainer style"
  • "Documentary with natural footage feel"

Stage 3: Audio Overlay

Add narration, music, or sound effects.

Options:

  • AI-generated voiceover from script
  • Background music selection
  • Sound effect overlays
  • Custom audio upload

Stage 4: Export & Ingest

Download and optionally ingest into MediaDB.


Browser Automation via Claude-in-Chrome

Super Gems operate through the Gemini web interface. Use Claude-in-Chrome for automation.

Session Initialization

1. Check for existing Gemini tab:
   - Use tabs_context_mcp to find gemini.google.com tabs

2. Navigate if needed:
   - navigate to https://gemini.google.com
   - Wait for page load

3. Verify authentication:
   - Check for user avatar/profile indicator
   - If not authenticated, prompt user to log in manually

Gem Manager Navigation

1. From Gemini main page:
   - Locate "Gem Manager" or gems icon in sidebar
   - Click to open Gem Manager

2. Find or create gem:
   - Use find tool for "Create new gem" or existing gem name
   - Click to open gem editor

Veo Generation Monitoring

1. After triggering generation:
   - Enter polling loop (60-second intervals)
   - Check for progress indicator or completion

2. On completion:
   - Locate download/export button
   - Extract download URL
   - Trigger veo_ingest.json workflow

Element Patterns

ElementTypical Selector Pattern
Gem ManagerSidebar icon or menu item
Create GemButton with "Create" or "+"
Generate VideoButton containing "Generate"
DownloadExport icon or "Download" text
ProgressLoading indicator or percentage

Veo Integration Details

Capabilities

  • Text-to-video: Generate videos from text descriptions
  • Style transfer: Apply visual styles to generated content
  • Reference images: Use images as style guides (optional)
  • Duration control: Specify video length

Limitations

  • Generation time: 5-15 minutes per video
  • Quota: Subject to Gemini Ultra subscription limits
  • Quality: Best for short-form content (under 60 seconds)
  • Control: Less precise than frame-by-frame editing

Output Formats

FormatResolutionTypical Size
MP4720p-1080p10-50 MB
WebM720p-1080p8-40 MB

Director's Formula (Veo Prompting)

For optimal Veo results, structure prompts using all six elements:

[Camera: movement, lens, focus]
[Subject: detailed description, clothing, expression]
[Action: concrete verbs, specific movements]
[Setting: environment, lighting, time of day]
[Style: aesthetic, color grading, production quality]
[Audio: SFX tags, dialogue in quotes, ambient sounds]

Example:

[Camera: Slow dolly zoom, 35mm lens, shallow depth of field]
[Subject: Professional woman in business casual, confident posture]
[Action: She reaches for the smart water bottle, LED display glows]
[Setting: Modern minimalist kitchen, morning golden hour light]
[Style: Corporate modern aesthetic, warm color grading, clean lines]
[Audio: Ambient kitchen sounds, soft "whoosh" on bottle pickup, subtle corporate music]

This formula ensures Veo receives unambiguous directorial instructions.


Bidirectional Media Database Integration

Inbound: Veo → MediaDB

Automatically ingest Veo outputs into the LanceDB media database.

Workflow:

Veo completion → Download video → Generate metadata →
Extract thumbnail → Add to LanceDB → Slack notification

Metadata mapping:

Veo OutputMediaDB Field
video filemedia_blob
frame at 5sthumbnail_blob
generation promptdescription
style parametersstyle_tags
"veo", "ai_generated"subjects
auto-detectedmood
1 (default)quality_class

Use script: scripts/veo_ingest.py

Outbound: MediaDB → Super Gems

Export MediaDB assets for use as Veo inputs.

Use cases:

  • Reference images for style transfer
  • Existing clips for extension/remixing
  • Audio assets for overlay

Workflow:

Query MediaDB → Select assets → Export to staging →
Upload to Veo reference panel → Generate

Use script: scripts/media_export.py


N8N Workflow Integration

Webhook Endpoints

EndpointMethodPurpose
/webhook/veo-ingestPOSTIngest completed Veo video
/webhook/super-gems-triggerPOSTTrigger full workflow
/webhook/super-gems-status/{id}GETCheck workflow status

veo-ingest Request Body

{
  "download_url": "https://gemini.google.com/...",
  "filename": "veo_output_001.mp4",
  "session_id": "gems_session_abc123",
  "quality_hint": 1,
  "subjects": ["marketing", "product demo"]
}

super-gems-trigger Request Body

{
  "workflow_type": "video_production",
  "script_prompt": "Create a 30-second product demo...",
  "visual_style": "corporate, modern, clean",
  "audio_style": "upbeat background music",
  "source_assets": ["media_id_1", "media_id_2"]
}

Since Super Gems cannot directly call external APIs or webhooks, we use Google Sheets as a bridge to connect to N8N workflows.

Integration Pattern

Super Gem → Writes to Google Sheet → N8N monitors Sheet → Triggers veo_ingest.py → MediaDB

Why Use Sheet Bridge?

FeatureWebhook ApproachSheet Bridge Approach
Direct API callsNot supportedN/A (uses native Sheets)
Browser automationRequiredOptional
ReliabilityDepends on UI stabilityVery stable
Setup complexityMedium (Claude-in-Chrome)Low (native integration)

Sheet Structure

The "Super Gems Workflow Queue" spreadsheet tracks all production workflows:

ColumnPurpose
session_idUnique workflow identifier
statuspendingready_for_ingestprocessingcomplete
drive_urlLink to exported video in Google Drive
metadataJSON with subjects, mood, style_tags
created_atTimestamp of creation
processed_atTimestamp when N8N processed

N8N Workflow

Use n8n_workflows/super_gems_sheet_monitor.json which:

  1. Polls the Sheet for status = "ready_for_ingest"
  2. Downloads video from Drive URL
  3. Runs veo_ingest.py to ingest to MediaDB
  4. Updates Sheet status
  5. Sends Slack notification

Setup Guide

See references/google-sheet-bridge.md for complete setup instructions.


Session State Management

Track multi-stage workflow progress with session state.

State file: ~/.gemini-super-gems/sessions.json

Session structure:

{
  "session_id": "gems_abc123",
  "workflow_name": "product_demo_v1",
  "created_at": "2026-01-14T12:00:00Z",
  "stages": {
    "scripting": {"status": "completed", "output": "..."},
    "visual": {"status": "in_progress", "started_at": "..."},
    "audio": {"status": "pending"},
    "export": {"status": "pending"}
  },
  "outputs": {}
}

Use script: scripts/super_gems_state.py


Best Practices

Session Management

  1. Single active session: Run one Veo generation at a time
  2. Tab persistence: Keep Gemini tab open during generation
  3. Auth monitoring: Check for session expiry periodically
  4. Screenshot verification: Capture screenshots at each stage

Generation Optimization

  1. Clear prompts: Specific, detailed descriptions yield better results
  2. Style consistency: Define style once, reference throughout
  3. Reasonable duration: Start with 15-30 second videos
  4. Batch planning: Queue multiple videos, process sequentially

Error Recovery

ErrorRecovery Action
Session expiredPrompt user re-authentication
Generation timeoutRetry with simplified prompt
Quota exceededQueue for later, notify user
Network errorRetry with exponential backoff
UI changedUpdate selectors, screenshot debug

Context Poisoning Prevention (Gemini 3)

Gemini 3 can experience "context drift" where it forgets instructions over long conversations. Prevent this with:

  1. Sandwich Structure Prompts:

    • Top: Objective Definition (the "Why")
    • Middle: Context and Data
    • Bottom: Output Protocol (forces execution mode)
  2. Structured Delimiters:

    • Use <context>, <task>, <rules> XML tags
    • Clear boundaries prevent instruction confusion
  3. Objective-Driven Constraints:

    • Instead of: "Don't deviate from brand"
    • Use: "Objective: Maintain brand consistency"
  4. Reset Techniques:

    • Settings > Personal Context > Toggle OFF/ON
    • Clear old memories in Settings > Saved Info
    • Start fresh sessions for complex workflows

See templates/super-gem-prompt-template.md for the full optimized prompt.


Workflow Templates

Product Demo Template

workflow: product_demo
stages:
  scripting:
    prompt: "Write a 30-second script highlighting [features]"
  visual:
    style: "corporate, modern, product-focused"
    duration: 30
  audio:
    type: "voiceover + background"
    music_style: "upbeat, professional"

Content Repurposing Template

workflow: content_repurpose
input: existing_blog_post
stages:
  scripting:
    prompt: "Convert blog post to video script"
  visual:
    style: "educational, animated"
    reference_images: from_mediadb
  audio:
    type: "ai_voiceover"

See templates/ directory for full workflow templates.


Bundled Resources

Scripts

  • scripts/veo_ingest.py - Download and ingest Veo outputs
  • scripts/super_gems_state.py - Session state management
  • scripts/media_export.py - Export MediaDB assets for Veo
  • scripts/workflow_monitor.py - Monitor long-running generations

References

  • references/super-gems-patterns.md - Detailed workflow patterns
  • references/veo-integration.md - Veo capabilities deep dive
  • references/browser-automation.md - Claude-in-Chrome patterns
  • references/google-sheet-bridge.md - Sheet-based N8N integration guide
  • references/drive-asset-setup.md - Google Drive asset library setup

Templates

  • templates/video-production-workflow.md - Full production template
  • templates/content-repurposing.md - Content transformation template
  • templates/super-gem-prompt-template.md - Full Sandwich Structure prompt for Gem configuration

N8N Workflows

  • n8n_workflows/veo_ingest.json - Veo ingestion webhook
  • n8n_workflows/super_gems_trigger.json - Full workflow trigger (webhook-based)
  • n8n_workflows/super_gems_sheet_monitor.json - Sheet-based workflow monitor (recommended)

スコア

総合スコア

50/100

リポジトリの品質指標に基づく評価

SKILL.md

SKILL.mdファイルが含まれている

+20
LICENSE

ライセンスが設定されている

0/10
説明文

100文字以上の説明がある

+10
人気

GitHub Stars 100以上

0/15
最近の活動

3ヶ月以内に更新がある

0/10
フォーク

10回以上フォークされている

0/5
Issue管理

オープンIssueが50未満

+5
言語

プログラミング言語が設定されている

+5
タグ

1つ以上のタグが設定されている

0/5

レビュー

💬

レビュー機能は近日公開予定です