glasses-to-social（眼镜社交）

Glasses-to-Social

Transform photos from smart glasses into social media posts with AI-generated captions.

Overview

This skill creates a pipeline from smart glasses (Meta Ray-Ban, etc.) to social media:

1. User snaps photo with glasses
2. Photo syncs to Google Drive folder
3. Agent detects new photo, analyzes with vision
4. Agent drafts post matching user's voice/style
5. User approves, agent publishes

Setup

1. Configure Google Drive Folder

Create a shared Google Drive folder for glasses photos:

# User creates folder "Glasses-to-Social" in Google Drive
# Share with "Anyone with link can view"
# Copy the folder URL

2. Set Up Config

Create config file at glasses-to-social/config.json:

{
  "googleDriveFolderUrl": "https://drive.google.com/drive/folders/YOUR_FOLDER_ID",
  "folderId": "YOUR_FOLDER_ID",
  "downloadPath": "./glasses-to-social/downloads",
  "processedFile": "./glasses-to-social/data/processed.json",
  "defaultHashtags": ["#MedicalAI", "#HealthTech"],
  "autoPost": false
}

3. Configure Glasses Auto-Sync

For Meta Ray-Ban glasses:

1. Open Meta View app on phone
2. Settings > Gallery > Enable "Import Automatically"
3. iOS: Enable Google Photos backup (syncs Camera Roll)
4. Create iOS Shortcut to copy new Meta photos to Google Drive folder

Usage

Manual Check

Ask the agent to check for new photos:

Check my glasses folder for new photos

Automated Monitoring

Set up a cron job to check periodically:

{
  "name": "Glasses-to-Social: Check photos",
  "schedule": {"kind": "cron", "expr": "*/15 * * * *", "tz": "UTC"},
  "payload": {
    "message": "Check the Glasses-to-Social folder for new photos. If found, analyze and draft a tweet."
  }
}

Processing Flow

When a new photo is detected:

1. Download from Google Drive using gdown:

   gdown --folder "FOLDER_URL" -O ./downloads/ --remaining-ok

1. Compare against processed list in data/processed.json

1. For new photos, analyze with vision:

- Describe the scene/subject - Identify relevant context for social post - Note any text, people, or notable elements

1. Draft post matching user's voice:

- Keep it concise and authentic - Add relevant hashtags - First-person perspective works well for glasses content

1. Send draft to user for approval:

- Include image preview - Show proposed caption - Wait for "POST" confirmation or edits

1. On approval, publish to configured platform (X/Twitter, etc.)

1. Mark photo as processed in data/processed.json

Scripts

check-new-photos.sh

Checks Google Drive folder for new images:

scripts/check-new-photos.sh

Output format when new photo found:

NEW_PHOTO_PATH:/path/to/downloaded/photo.jpg

File Tracking

Track processed photos in data/processed.json:

{
  "processed": ["photo1.jpg", "photo2.jpg"],
  "pending": []
}

Tips

• First-person POV content performs well ("Look what I just saw...")
• Keep captions authentic, not overly polished
• Works great for conferences, interesting sightings, daily moments
• Consider time-of-day context when drafting

Requirements

• gdown Python package for Google Drive access
• Vision-capable model for image analysis
• Twitter/X credentials for posting (optional)