telegram-skills

agent
Guvenlik Denetimi
Gecti
Health Gecti
  • License — License: MIT
  • Description — Repository has a description
  • Active repo — Last push 0 days ago
  • Community trust — 37 GitHub stars
Code Gecti
  • Code scan — Scanned 12 files during light audit, no dangerous patterns found
Permissions Gecti
  • Permissions — No dangerous permissions requested

Bu listing icin henuz AI raporu yok.

SUMMARY

Telegram skills for AI agents (Claude Code, Codex, Cursor): Rich Messages, streaming AI replies, HTML widgets, polls, stickers and Mini Apps via the Telegram Bot API.

README.md

Telegram Skills for AI Agents

Telegram skills for Claude Code, Codex CLI, Cursor, and any agent supporting the Agent Skills standard (agentskills.io). Teach your AI agent to send Rich Messages (structured documents, tables, galleries, maps, uploaded media), stream AI replies, build interactive HTML widgets, run polls, and make sticker packs from any Telegram bot. Guide: Telegram skills for Claude Code and Codex.

Bot API 10.1 (June 11, 2026) introduced Rich Messages: bots can send document-grade structured messages and stream AI-generated replies with a native "thinking" block. Bot API 10.2 (July 14, 2026) added outgoing block JSON plus explicit media bindings, including Telegram file_id, HTTP URLs, and multipart uploads.

These skills work through raw HTTP calls to the Bot API — no bot framework required, any stack.

Bot API 10.3 (August 24, 2026) adds native rich-message button rows and compact tables. Styled inline keyboard buttons use primary, success, and danger. The dependency-free Python UI helper builds validated literal block payloads and native inline buttons; it performs no network calls. See consumer integration.

Polls (Bot API 9.1–10.1): up to 12 options, a 1024-char description with media, so a post and its poll fit in one message. See tg-polls.

Install

Universal (recommended)

npx skills add serejaris/telegram-skills

# Install a single skill
npx skills add serejaris/telegram-skills --skill tg-rich-messages

Claude Code (native marketplace)

/plugin marketplace add serejaris/telegram-skills
/plugin install telegram-skills@telegram-skills

Codex CLI

npx skills add serejaris/telegram-skills
# or manually: cp -r skills/* ~/.agents/skills/
# Inside a clone of this repo, skills are auto-discovered via .agents/skills/

Manual copy

# Claude Code
cp -r skills/* ~/.claude/skills/

# Codex CLI / other agents
cp -r skills/* ~/.agents/skills/

# or per-project
cp -r skills/* your-project/.claude/skills/
cp -r skills/* your-project/.agents/skills/

Then ask your agent — e.g. "send this report to my Telegram channel as a rich message".

Skills

Skill What it does
tg-rich-messages Core reference: markup, outgoing block JSON, media bindings and uploads, preflight, limits, raw HTTP sending
tg-markdown-to-rich Convert Markdown into a rich message and bind file_id, URL, or uploaded media
tg-rich-streaming Stream LLM output into a chat: draft animation, thinking block, mandatory finalization
tg-polls Polls and quizzes from a bot or a user account: text + poll in one message via description, media, 12 options, all poll settings and limits
tg-rich-digest Digest and channel-article patterns: flat layout, preview + collapsed full version, media as evidence, preflight gate
tg-reply-rewrite Rewrite a raw agent reply for a non-engineer on a phone: what happened and how urgent, what it means, a ready sentence to send back, every fact unchanged
tg-operator-cards Cards a bot sends to its own operator about one event (payment, order, enrollment): outcome and person on the first line, one fact per line, warning in place of the normal line, decision footer, escaped dynamic text
tg-widgets Interactive HTML message widgets (HTML bubbles, sendWidget): how to write them by Telegram's docs, limits, input, state, which chats accept a widget in production
tg-sticker-pack Static stickers from a meme frame or video: cutout, caption on top (WordArt, Impact, tags), offline format check, publishing through @Stickers from a user account or createNewStickerSet from a bot

Reference

License

MIT

Yorumlar (0)

Sonuc bulunamadi