telegram-bot · claude-office-skills — Claude Code skill: bots, alerts, group automation
A skill for building Telegram bots: chatbots, notifications, AI assistants and group automation. Useful when you need message handlers and workflows fast.
npx skills add claude-office-skills/skills --skill telegram-bot -a claude-codeWhat telegram-bot does
The skill walks through the full bot lifecycle. Start with @BotFather: /newbot, a name, a username ending in bot, then grab the API token. After that come description, avatar and command list, plus pulling chat.id from getUpdates.
Inside are handler templates for text, commands, photos, documents, voice and location. There are n8n workflows too: a trigger on messages and callback queries, a switch by content type, a model call, and a reply. Voice notes get transcribed with Whisper; images go through a vision model.
Separate sections cover notifications and group work. Alert, build-status and order templates with routing by severity. Welcome messages for new members, spam detection, link filtering and flood control. The skill relies on the telegram-mcp server with telegram_send_message, telegram_get_updates and telegram_send_photo tools — without it some workflows won't run.
What it helps with
- create a bot and get an API token via @BotFather
- handle text, photo and voice messages in one flow
- send build or order alerts routed by severity
- welcome new group members and filter spam
- wire up an AI assistant with recent-message memory
Example prompts for the agent
- Create a Telegram bot: set it up via @BotFather and add /start and /help commands
- Build a voice message handler that transcribes with Whisper
- Send build status alerts to the admin group
How to install telegram-bot
1Skills CLI (Vercel)
npx skills add claude-office-skills/skills --skill telegram-bot -a claude-code2Manually: put the files in place
~/.claude/skills/telegram-bot/SKILL.mdCopy the whole skill folder; for one project use .claude/skills/telegram-bot/
1Skills CLI (Vercel)
npx skills add claude-office-skills/skills --skill telegram-bot -a codex2Manually: put the files in place
.agents/skills/telegram-bot/SKILL.mdFor all projects use ~/.agents/skills/
1Skills CLI (Vercel)
npx skills add claude-office-skills/skills --skill telegram-bot -a cursor2Manually: put the files in place
.cursor/skills/telegram-bot/SKILL.mdCursor also reads .agents/skills/ and .claude/skills/
1GitHub CLI
gh skill install claude-office-skills/skills telegram-bot2Manually: put the files in place
.github/skills/telegram-bot/SKILL.md1Skills CLI (Vercel)
npx skills add claude-office-skills/skills --skill telegram-bot -a gemini-cli2Platform CLI
gemini skills install https://github.com/claude-office-skills/skills --path telegram-botManually: put the files in place
.devin/skills/telegram-bot/SKILL.mdThe legacy .windsurf/skills/ path still works
Skills CLI (Vercel)
npx skills add claude-office-skills/skills --skill telegram-bot -a opencodeClaude.ai
Zip the folder and upload it in Settings → Capabilities → Skills
Step by step, with commands for each platform: How to install skills in Claude Code: five ways
Questions about telegram-bot
How do I install telegram-bot in Claude?
Run npx skills add claude-office-skills/skills --skill telegram-bot -a claude-code. The skill lands in ~/.claude/skills/telegram-bot/SKILL.md.
What does telegram-bot do?
It helps you build Telegram bots: message handlers, AI assistants, notification systems and group automation. Templates for n8n and code examples are included.
What does the skill need to work?
The telegram-mcp server with telegram_send_message, telegram_get_updates and telegram_send_photo tools. Some workflows won't run without it.
Source text
SKILL.md · MIT
---
name: telegram-bot
description: "Telegram bot development - chatbots, notifications, AI assistants, and group automation"
version: "1.0.0"
author: claude-office-skills
license: MIT
category: messaging
tags:
- telegram
- bot
- chatbot
- automation
- notifications
department: Engineering
models:
recommended:
- claude-sonnet-4
mcp:
server: telegram-mcp
tools:
- telegram_send_message
- telegram_get_updates
- telegram_send_photo
capabilities:
- bot_development
- message_handling
- ai_integration
- notification_system
- group_management
languages:
- en
- zh
related_skills:
- discord-bot
- whatsapp-automation
- ai-agent-builder
- slack-workflows
---
# Telegram Bot
Build Telegram bots for chatbots, notifications, AI assistants, and group automation. Based on n8n's Telegram workflow templates.
## Overview
This skill covers:
- Bot setup and configuration
- Message handling patterns
- AI-powered assistants
- Notification workflows
- Group automation
---
## Bot Setup
### Creating a Bot
```yaml
setup_steps:
1. create_bot:
- open: @BotFather on Telegram
- command: /newbot
- provide: bot_name
- provide: bot_username (must end in 'bot')
- receive: API_token
2. configure_bot:
- command: /setdescription
- command: /setabouttext
- command: /setuserpic
- command: /setcommands
3. get_chat_id:
- start: conversation with bot
- call: https://api.telegram.org/bot{TOKEN}/getUpdates
- extract: chat.id from response
```
### Bot Commands
```yaml
commands:
- command: /start
description: "Start the bot"
- command: /help
description: "Show available commands"
- command: /status
description: "Check system status"
- command: /subscribe
description: "Subscribe to notifications"
- command: /unsubscribe
description: "Unsubscribe from notifications"
```
---
## Message Handlers
### Basic Message Handler
```yaml
workflow: "Telegram Message Handler"
trigger: telegram_message
handlers:
text_message:
action: |
1. Parse message text
2. Determine intent
3. Process request
4. Send response
command:
pattern: "^/"
action: route_to_command_handler
photo:
action: |
1. Download photo
2. Process with vision AI
3. Respond with analysis
document:
action: |
1. Download document
2. Extract content
3. Process and respond
voice:
action: |
1. Download audio
2. Transcribe with Whisper
3. Process text
4. Respond (text or voice)
location:
action: |
1. Extract coordinates
2. Lookup local info
3. Respond with relevant data
```
### n8n Workflow
```yaml
workflow: "Telegram Bot n8n"
nodes:
- name: "Telegram Trigger"
type: "n8n-nodes-base.telegramTrigger"
parameters:
updates: ["message", "callback_query"]
- name: "Route Message Type"
type: "n8n-nodes-base.switch"
parameters:
rules:
- output: 0
condition: "{{ $json.message.text.startsWith('/') }}"
- output: 1
condition: "{{ $json.message.photo }}"
- output: 2
condition: "{{ $json.message.voice }}"
- output: 3
fallback: true
- name: "Process with AI"
type: "n8n-nodes-base.openAi"
parameters:
model: "gpt-4"
messages:
- role: "system"
content: "You are a helpful Telegram assistant."
- role: "user"
content: "{{ $json.message.text }}"
- name: "Send Response"
type: "n8n-nodes-base.telegram"
parameters:
chatId: "{{ $json.message.chat.id }}"
text: "{{ $json.response }}"
```
---
## AI-Powered Bot
### GPT-4 Integration
```yaml
ai_bot:
name: "AI Assistant Bot"
system_prompt: |
You are a helpful AI assistant on Telegram.
Guidelines:
- Be concise (Telegram has message limits)
- Use emojis appropriately
- Format with markdown when helpful
- Ask clarifying questions if needed
features:
- conversational_memory: true
- context_window: last_10_messages
- tools: [web_search, calculator, weather]
message_formatting:
max_length: 4096
split_long_messages: true
use_markdown: true
```
### Multi-Modal Bot
```yaml
multimodal_bot:
handlers:
text:
model: gpt-4
action: chat_completion
image:
model: gpt-4-vision
action: analyze_and_respond
voice:
transcribe: whisper
process: gpt-4
respond: text_or_voice
document:
extract: based_on_type
summarize: gpt-4
respond: text
```
---
## Notification System
### Alert Bot
```yaml
workflow: "System Alert Bot"
triggers:
- source: monitoring_system
event: alert
- source: ci_cd
event: build_status
- source: ecommerce
event: new_order
notification_templates:
alert:
format: |
🚨 *Alert: {severity}*
*Service:* {service}
*Message:* {message}
*Time:* {timestamp}
[View Dashboard]({dashboard_link})
build:
format: |
{status_emoji} *Build {status}*
*Project:* {project}
*Branch:* {branch}
*Commit:* `{commit_short}`
{details}
order:
format: |
🛒 *New Order!*
*Order:* #{order_id}
*Customer:* {customer}
*Total:* ${total}
*Items:* {item_count}
routing:
by_severity:
critical: [admin_group, on_call_user]
warning: [team_group]
info: [logging_channel]
```
### Scheduled Notifications
```yaml
scheduled_notifications:
daily_digest:
schedule: "9am daily"
template: |
📊 *Daily Summary - {date}*
📈 Sales: ${sales} ({change})
👥 New users: {new_users}
🎫 Open tickets: {tickets}
Have a great day! ☀️
weekly_report:
schedule: "Monday 9am"
template: weekly_metrics_report
reminder:
trigger: custom_event
template: |
⏰ *Reminder*
{reminder_text}
Scheduled by: {creator}
```
---
## Group Automation
### Welcome Bot
```yaml
group_bot:
on_member_join:
action: |
1. Check if new member
2. Send welcome message
3. Share rules
4. Suggest introduction
template: |
👋 Welcome to {group_name}, {user_name}!
Please:
1. Read the /rules
2. Introduce yourself
3. Ask questions anytime!
Enjoy your stay! 🎉
on_member_leave:
action: optional_goodbye
moderation:
- spam_detection: auto_delete + warn
- link_filtering: whitelist_only
- flood_control: rate_limit
```
### Poll & Survey Bot
```yaml
poll_bot:
commands:
/poll:
format: "/poll Question | Option1 | Option2 | ..."
action: create_poll
/survey:
action: start_multi_question_survey
/results:
action: show_aggregated_results
features:
- anonymous_voting: optional
- multiple_choice: optional
- scheduled_close: optional
- export_results: to_sheets
```
---
## Inline Features
### Inline Keyboard
```yaml
inline_keyboard:
example:
text: "Choose an option:"
keyboard:
- row:
- text: "Option 1"
callback_data: "opt_1"
- text: "Option 2"
callback_data: "opt_2"
- row:
- text: "Cancel"
callback_data: "cancel"
callback_handler:
opt_1:
action: process_option_1
response: "You selected Option 1"
opt_2:
action: process_option_2
response: "You selected Option 2"
cancel:
action: cancel_operation
response: "Operation cancelled"
```
### Inline Query
```yaml
inline_mode:
trigger: "@botname query"
results:
- type: article
title: "{query} result"
description: "Click to send"
message_text: "Result for: {query}"
- type: photo
photo_url: "{search_result_image}"
- type: gif
gif_url: "{gif_url}"
```
---
## Output Example
**Request**: "Create a Telegram notification bot for server alerts"
**Output**:
```markdown
# Server Alert Telegram Bot
## Setup
1. Create bot with @BotFather
2. Get token: `123456:ABC-DEF...`
3. Get chat ID for alerts channel
## n8n Workflow
```yaml
nodes:
1. Webhook (receives alerts)
2. Format Message
3. Telegram Send
```
## Alert Templates
**Critical Alert:**
```
🚨 CRITICAL ALERT
Server: production-web-01
Status: DOWN
Time: 2026-01-30 14:32:05 UTC
Details:
- CPU: 98%
- Memory: 95%
- Disk: 89%
[View Grafana] [Acknowledge]
```
**Recovery:**
```
✅ RECOVERED
Server: production-web-01
Downtime: 5 minutes
Status: All systems normal
Incident resolved automatically.
```
## Implementation
```javascript
// Send alert function
async function sendAlert(severity, message, details) {
const emoji = {
critical: '🚨',
warning: '⚠️',
info: 'ℹ️',
success: '✅'
};
const text = `${emoji[severity]} *${severity.toUpperCase()}*\n\n${message}\n\n${details}`;
await telegram.sendMessage({
chat_id: ALERT_CHANNEL_ID,
text: text,
parse_mode: 'Markdown'
});
}
```
## Features
- Severity-based routing
- Inline action buttons
- Acknowledgment tracking
- Escalation rules
```
---
*Telegram Bot Skill - Part of Claude Office Skills*
© claude-office-skills · MIT · claude-office-skills/skills
Similar skills
SKILL.md skills catalog →telegram-bot by other authors comes first.