Telegram

Telegram

Send, read, and respond to Telegram messages via the Claudicle daemon adapter. Thin skill layer on top of the existing python-telegram-bot v22 polling adapter at ~/.claudicle/daemon/adapters/telegram_adapter.py.

When to Use This Skill

• Sending a Telegram message from a Claude Code session
• Checking for unhandled Telegram messages in the inbox
• Processing Telegram messages through the /telegram-respond cognitive pipeline
• Managing Telegram memory context (working memory, user models, soul state)

Prerequisites

• python-telegram-bot v21+ installed (pip install python-telegram-bot)
• TELEGRAM_BOT_TOKEN in ~/.config/env/secrets.env
• Bot: @claudicle_bot (ID: 8646522411)
• Tom's Telegram ID: 633125581

Environment Variables

Quick Start

# Start the listener (background daemon)
python3 ~/.claudicle/adapters/telegram/telegram_listen.py --bg

# Check for unhandled messages
cd ~/.claude/skills/telegram/scripts && python3 telegram_check.py

# Send a message
cd ~/.claude/skills/telegram/scripts && python3 telegram_send.py 633125581 "Hello from Claudicle!"

# Process inbox through cognitive pipeline
/telegram-respond

Available Scripts

telegram_send.py -- Send a Telegram message

python3 ~/.claudicle/adapters/telegram/telegram_post.py CHAT_ID "Hello!"
python3 telegram_send.py CHAT_ID "Reply to this" --reply-to 42
python3 telegram_send.py CHAT_ID --stdin    # Read text from stdin
python3 telegram_send.py --help

telegram_check.py -- Check inbox for unhandled messages

python3 telegram_check.py                    # Unhandled telegram:* entries
python3 telegram_check.py --all              # All entries (including handled)
python3 telegram_check.py --limit 5          # Last 5
python3 telegram_check.py --mark-read MSG_ID # Mark entry as handled
python3 telegram_check.py --mark-all-read    # Mark all as handled

Reads from ~/.claudicle/daemon/inbox.jsonl (shared with Slack/Discord).

telegram_memory.py -- Memory context CLI

python3 telegram_memory.py load-context 633125581
python3 telegram_memory.py user-model tg:633125581
python3 telegram_memory.py soul-state

Thin CLI over ~/.claudicle/adapters/shared/claudicle_memory.py.

Listener Management

The listener is the existing Claudicle adapter script (not duplicated):

python3 ~/.claudicle/adapters/telegram/telegram_listen.py --bg      # Start
python3 ~/.claudicle/adapters/telegram/telegram_listen.py --status   # Check
python3 ~/.claudicle/adapters/telegram/telegram_listen.py --stop     # Stop

Channel Format

telegram:{chat_id} -- e.g., telegram:633125581 (private), telegram:-100123456789 (group)

Architecture

Claude Code session
  |
  |-- /telegram-respond  (cognitive pipeline: 6 XML tags)
  |-- telegram_send.py   (wraps Bot API send_message)
  |-- telegram_check.py  (reads shared inbox.jsonl)
  |-- telegram_memory.py (thin CLI over claudicle_memory)
  |
  v
~/.claudicle/adapters/telegram/     (existing, not duplicated)
  telegram_listen.py                 (polling daemon)
  telegram_post.py                   (CLI post)
  _telegram_utils.py                 (split_message, config)
  |
  v
~/.claudicle/adapters/shared/claudicle_memory.py  (3-tier memory routing)

Security

• DM access is DENY-by-default. Set CLAUDICLE_TELEGRAM_ALLOWED_USERS=633125581 to allow Tom.
• User input is sanitized for XML tags before cognitive pipeline processing.
• WebFetch removed from allowed tools in daemon mode subprocess.