Agent Skills: Telegram Scraper Run

Telegram Scraper Run

Runs the Telegram Scraper Agent manually for testing or unscheduled scanning.

When to use

• "run telegram scraper"
• "scan telegram channels"
• "find new channels with AI"
• "check competitors on Telegram"

Input

Optional:

• --dry-run - test run without notifications
• --category <name> - scan only one category (competitors/industry/advertising)
• --no-messages - skip reading messages (faster)
• --notify-test - notification test only

How to execute

Full run (production)

cd $AGENTS_PATH/telegram-scraper
python3 telegram_scraper_agent.py

Dry-run tests

# Test without notifications
python3 telegram_scraper_agent.py --dry-run

# Single category
python3 telegram_scraper_agent.py --category competitors --dry-run

# Without reading messages (faster)
python3 telegram_scraper_agent.py --no-messages --dry-run

Notification test

python3 telegram_scraper_agent.py --notify-test

Unit tests

python3 test_telegram_scraper.py

Output

Agent outputs:

1. Progress to stderr (channel scanning)
2. Summary to stdout (results from Claude)
3. Telegram notification (if high-value channels found)

Data is saved to:

$PROJECT_ROOT/data/telegram_scraper/
├── YYYY-MM-DD/           # Dated results
│   ├── competitors_channels.json
│   ├── competitors_ad_contacts.csv
│   ├── industry_channels.json
│   ├── advertising_channels.json
│   └── messages/
└── latest/               # Symlinks to most recent

Checking results

# Latest results
ls -l $PROJECT_ROOT/data/telegram_scraper/latest/

# Top 5 channels (competitors)
cat $PROJECT_ROOT/data/telegram_scraper/latest/competitors_channels.json | jq '.[0:5]'

# Ad contacts
cat $PROJECT_ROOT/data/telegram_scraper/latest/competitors_ad_contacts.csv

# Agent log
cat $PROJECT_ROOT/data/telegram_scraper/agent_log.json | jq '.[-5:]'

Configuration

Edit config:

code $PROJECT_ROOT/data/telegram_scraper_config.json

Config structure:

{
  "categories": {
    "competitors": {
      "keywords": ["annotation", "data labeling", "cvat"],
      "exclude": ["spam", "crypto"],
      "scan_posts": 10
    }
  },
  "min_subscribers": 100,
  "min_score": 10,
  "notification_threshold": 30
}

Launchd Schedule

Agent runs automatically twice daily (9:00, 18:00).

# Check status
launchctl list | grep telegram-scraper

# Load schedule
launchctl load ~/Library/LaunchAgents/com.yourcompany.telegram-scraper.plist

# Unload schedule
launchctl unload ~/Library/LaunchAgents/com.yourcompany.telegram-scraper.plist

# View logs
tail -f $GOOGLE_TOOLS_PATH/logs/telegram_scraper.log
tail -f $GOOGLE_TOOLS_PATH/logs/telegram_scraper.err

Troubleshooting

Session Expired Error

If Telegram session is invalid:

# Refresh session
cd $TG_TOOLS_PATH
python3 -m tg_utils.auth

No Results

• Check keywords in config (too specific?)
• Verify session: cd $TG_TOOLS_PATH && python3 -m tg_utils.auth
• Run with --dry-run for debug

Rate Limited

• Normal: agent waits and retries
• FloodWaitError > 5 min: channel skipped
• Solution: decrease scan_posts in config

Manual Scraping (without the agent)

If the agent is not working:

cd $TG_TOOLS_PATH/tools

# Find channels with ad contacts
python3 tg_scrape.py ads --keywords "annotation,labeling" --posts 10

# List channels
python3 tg_scrape.py channels --keywords "ai,ml" --output channels.csv

# Read messages
python3 tg_scrape.py messages "Channel Name" --days 7 --limit 50

Next steps

After scraping:

1. Add contacts to CRM: use add-lead skill
2. Write outreach: use telegram-send skill
3. Adjust config: edit config file and re-run

Related skills

• telegram-session - update Telegram session
• add-lead - add found contacts to CRM
• telegram-send - message ad contacts
• daily-briefing - include findings in morning briefing