Enables an agent to send, edit, and reply to Telegram messages using rich HTML or Markdown formatting.
- GitHub stars
- 1,703
- Files
- 4
- Last commit
- Nov 11, 2026
- License
- Apache-2.0
- Allowed tools
- —
What does the telegram skill do?
telegram enables an agent to send, edit, and reply to Telegram messages using rich HTML or Markdown formatting.
Summary generated from the skill's SKILL.md.
About the telegram skill
Communicating with users, groups, or channels via Telegram is what this skill enables for AI agents. It allows an agent to send new messages, reply to specific existing messages, edit previously sent content, or push proactive notifications. Developers can use it to manage outbound communication through the Telegram Bot API, ensuring that agents can stay active in a conversation without waiting for a prompt.
When an agent needs to deliver structured information like tables, math blocks, or collapsible details, it should load the telegram_rich.py script to use rich HTML or Markdown formatting. For simple, single-paragraph replies, the telegram_send.py script serves as a fallback. The skill relies on a Telegram Bot token and requires specific identifiers like chat_id and message_id to function. It is designed to handle complex scenarios, such as communicating with other bots in a group or providing real-time progress updates by editing a single message as a task evolves.
What an agent can do with it
- 1Send structured reportsAn agent can deliver complex data using HTML tags for headings, lists, and tables.
- 2Provide real-time task updatesAn agent can notify a user of its progress by updating a single message instead of sending multiple new ones.
- 3Interact with other botsIn a group setting, an agent can intentionally trigger another bot using specific mention patterns.
- 4Handle voice message inputsThe agent can process incoming audio by transcribing it before generating a response.
Written from the skill's SKILL.md. Every use case is checked against a quote from it.
How to install the telegram skill
Each command copies the skill folder into the client's skills directory. Restart the client or start a new session to load it.Install the telegram skill in Claude Code
~/.claude/skills/telegram/SKILL.md
$ npx degit bubbuild/bub/src/skills/telegram ~/.claude/skills/telegramInstall the telegram skill in Claude Code (project)
.claude/skills/telegram/SKILL.md
$ npx degit bubbuild/bub/src/skills/telegram .claude/skills/telegramInstall the telegram skill in Codex
~/.agents/skills/telegram/SKILL.md
$ npx degit bubbuild/bub/src/skills/telegram ~/.agents/skills/telegramInstall the telegram skill in Gemini CLI
~/.gemini/skills/telegram/SKILL.md
$ npx degit bubbuild/bub/src/skills/telegram ~/.gemini/skills/telegramInstall the telegram skill in Claude.ai
npx degit bubbuild/bub/src/skills/telegram telegram && zip -r telegram.zip telegram # claude.ai → Customize → Skills → + → Create skill → Upload a skill → telegram.zip
Automated checks
Ran Nov 15, 2026Full report →Passed our automated checks. That is not a security audit: read the scripts before you let an agent run them. How checks work
- Source repository linked and reachablebubbuild/bub/src/skills/telegram is public and reachablegithub.com/bubbuild/bub/tree/HEAD/src/skills/telegram ↗Pass
- Open source license (OSI)Apache-2.0spdx.org/licenses/Apache-2.0.html ↗Pass
- Commit or release in the last yearLast commit 2026-11-11github.com/bubbuild/bub/commits ↗Pass
- No download-and-run or credential theft in scripts3 scripts read; no download-and-run, remote eval, base64 execution or credential uploadPass
- No hidden instructions in SKILL.mdNo override phrases, safeguard switches, credential requests or hidden charactersgithub.com/bubbuild/bub/blob/HEAD/src/skills/telegram/SKILL. ↗Pass
- Pre-approved tools are scopedNo pre-approved tools; the agent asks before using onegithub.com/bubbuild/bub/blob/HEAD/src/skills/telegram/SKILL. ↗Pass
- SKILL.md follows the Agent Skills formatValid name and description; says when to use it; 185 linesgithub.com/bubbuild/bub/blob/HEAD/src/skills/telegram/SKILL. ↗Pass
SKILL.md
View on GitHubTelegram Skill
Agent-facing execution guide for Telegram outbound communication.
Env vars:
BUB_TELEGRAM_TOKEN=${config.telegram.token}
Required Inputs
Collect these before execution:
chat_id(required)message_id(required for edit or reply)- message content (required for send/edit)
reply_to_message_id(required when you need a threaded reply)
Rich Messages (Bot API 10.1+, June 2026) — PREFERRED
For any message that benefits from structured formatting — headings, bullet/numbered lists, tables, blockquotes, collapsible <details>, captioned media, math blocks, code blocks with language hints, or mixed media — use the dedicated telegram_rich.py script with rich-html-style markup. This is the first-choice path; fall back to plain markdown (telegram_send.py) only for short single-paragraph replies.
The Rich Message API uses sendRichMessage and editMessageText with a rich_message payload. Exactly one of html or markdown must be supplied. Supported HTML tags include <b>, <i>, <u>, <s>, <tg-spoiler>, <code>, <pre>, <mark>, <sub>, <sup>, <h1>-<h6>, <p>, <ul>/<ol>/<li>, <blockquote>, <aside>, <details>/<summary>, <table>, <figure>/<figcaption>, <img>/<video>/<audio>, <tg-collage>, <tg-slideshow>, <tg-map>, <tg-math>, <tg-math-block>, <tg-reference>, <tg-emoji>, <tg-time>. Full reference: https://core.telegram.org/bots/api#rich-html-style
# Send rich message (HTML)
uv run ./scripts/telegram_rich.py \
--chat-id <CHAT_ID> \
--html "<h2>Title</h2><p>Body with <b>bold</b> and <code>code</code>.</p>"
# Send rich message (HTML from heredoc via file)
uv run ./scripts/telegram_rich.py \
--chat-id <CHAT_ID> \
--html-file /tmp/payload.html \
--reply-to <MESSAGE_ID>
# Send rich message (Markdown instead of HTML)
uv run ./scripts/telegram_rich.py \
--chat-id <CHAT_ID> \
--markdown "## Title\n\n- item one\n- item two"
# Edit an existing message to become a rich message
uv run ./scripts/telegram_rich.py \
--chat-id <CHAT_ID> \
--edit <MESSAGE_ID> \
--html "<p>Updated content.</p>"
For multi-line rich HTML, prefer --html-file over passing the payload on the command line to avoid shell escaping pitfalls. The script reads the file as UTF-8 and strips a single trailing newline.
Execution Policy
- If handling a direct user message in Telegram and
message_idis known, send a reply message (--reply-to). - If source metadata says sender is a bot (
sender_is_bot=true), do not use reply mode, but send a normal message and prefix content with@<sender_username>(or the provided source username). If the user doesn't have a username, use--source-user-idto mention viatg://user?id=link. - For long-running tasks, optionally send one progress message, then edit that same message for final status.
- ALWAYS pass message content via stdin using heredoc pipe and
--message -(or--text -). NEVER embed message text directly in shell arguments — special characters like',",$,!will be mangled or cause syntax errors. - Prefer
telegram_rich.py(rich-html-style) overtelegram_send.pyfor anything beyond a single short paragraph. See the "Rich Messages" section above.
Bot to co-Bot Communication
In Telegram groups, communicate with another bot using only these patterns:
- Reply directly to the other bot's message when
message_idis available. - Use an explicit command mention such as
/command@OtherBotwhen you need to invoke that bot intentionally. - Do not assume free-form group text will reach another bot.
Active Response Policy
When this skill is in scope, prefer proactive and timely Telegram updates:
- Send an immediate acknowledgment for newly assigned tasks
- Send progress updates for long-running operations using message edits
- Send completion notifications when work finishes
- Send important status or failure notifications without waiting for follow-up prompts
- If execution is blocked or fails, send a problem report immediately with cause, impact, and next action
Recommended pattern:
- Send a short acknowledgment reply
- Continue processing
- If blocked, edit or send an issue update immediately
- Edit the acknowledgment message with final result when possible
Voice Message Policy
When the inbound Telegram message is voice:
- Transcribe the voice input first (use STT skill if available)
- Prepare response content based on transcription
- Prefer voice response output (use TTS skill if available)
- If voice output is unavailable, send a concise text fallback and state limitation
Reaction Policy
When an inbound Telegram message warrants acknowledgment but does not merit a full reply, use a Telegram reaction as the response. But when any explanation or details are needed, use a normal reply instead.
Command Templates
Paths are relative to this skill directory.
# Send message (ALWAYS use heredoc stdin, never inline text in arguments)
cat << 'EOF' | uv run ${SKILL_DIR}/scripts/telegram_send.py --chat-id <CHAT_ID> --token "$BUB_TELEGRAM_TOKEN" --message -
Your message content here.
Special characters are safe: $100, "quotes", 'apostrophes', !exclamation
EOF
# Reply to a specific message
cat << 'EOF' | uv run ${SKILL_DIR}/scripts/telegram_send.py --chat-id <CHAT_ID> --token "$BUB_TELEGRAM_TOKEN" --reply-to <MESSAGE_ID> --message -
Reply content here.
EOF
# Source message sender is bot: no direct reply, use @username style
cat << 'EOF' | uv run ${SKILL_DIR}/scripts/telegram_send.py --chat-id <CHAT_ID> --token "$BUB_TELEGRAM_TOKEN" --source-is-bot --source-username <USERNAME> --message -
Message to a bot using @username mention.
EOF
# Source message sender is bot without username: use tg://user?id= link
cat << 'EOF' | uv run ${SKILL_DIR}/scripts/telegram_send.py --chat-id <CHAT_ID> --token "$BUB_TELEGRAM_TOKEN" --source-is-bot --source-user-id <USER_ID> --source-display-name "Display Name" --message -
Message to a user/bot using tg://user?id= mention.
EOF
# Edit an existing message
cat << 'EOF' | uv run ${SKILL_DIR}/scripts/telegram_edit.py --chat-id <CHAT_ID> --token "$BUB_TELEGRAM_TOKEN" --message-id <MESSAGE_ID> --text -
Updated content here.
EOF
When sending message to a bot, either use --reply-to argument or pass --source-is-bot with --source-username otherwise the bot will not receive the message.
For other actions that not covered by these scripts, use curl to call Telegram Bot API directly with the provided token.
Script Interface Reference
telegram_rich.py (preferred for structured messages)
--chat-id,-c: required--html: rich message content as HTML--html-file: read rich message HTML from a file (preferred for multi-line payloads)--markdown: rich message content as Markdown--markdown-file: read rich message Markdown from a file--rtl: render right-to-left--skip-entity-detection: skip URL/email/mention auto-detection--reply-to,-r: optional message ID to reply to--disable-notification: send silently--edit MESSAGE_ID: edit an existing message instead of sending--token,-t: optional (normally not needed)
telegram_send.py
--chat-id,-c: required, supports comma-separated ids--message,-m: required (use-to read from stdin)--reply-to,-r: optional--token,-t: optional (normally not needed)--source-is-bot: optional flag, disables reply mode and adds mention prefix--source-username: optional, uses@usernamestyle mention when set--source-user-id: optional, usestg://user?id=link mention when username is not available--source-display-name: optional, display name for user ID mention (defaults to "User")
telegram_edit.py
--chat-id,-c: required--message-id,-m: required--text,-t: required (use-to read from stdin)--token: optional (normally not needed)
Bot to co-Bot Communication
In Telegram groups, communicate with another bot using only these patterns:
- Reply directly to the other bot's message when
message_idis available. - Use an explicit command mention such as
/command@OtherBotwhen you need to invoke that bot intentionally. - Do not assume free-form group text will reach another bot.
Files in this skill
4telegram skill questions
How do I install the telegram skill in Claude Code?
Run npx degit bubbuild/bub/src/skills/telegram ~/.claude/skills/telegram. Claude Code loads skills from ~/.claude/skills (all projects) or .claude/skills (one project) and uses a skill when its description matches the task.
Which clients can use the telegram skill?
Any client that reads the Agent Skills format: Claude Code, Codex, Gemini CLI and claude.ai (upload as a ZIP). Scripts in the skill run only where the client allows code execution.
How should I send long or structured messages?
Use the telegram_rich.py script with rich-html-style markup for headings, lists, tables, and other structured elements.
What is the best way to handle special characters in message content?
Always pass message content via stdin using a heredoc pipe to avoid shell escaping errors with characters like quotes or dollar signs.
How can an agent communicate with another bot in a group?
Reply directly to the other bot's message or use an explicit command mention like /command@OtherBot.