Telegram Assistant Skill

A Claude Code skill for automating Telegram workflows: digests, channel posting, and style-matched drafts.

Features

• Digest Workflow: Summarize unread messages across all chats with priority classification
• Style Extraction: Analyze channel posts to capture writing style patterns
• Post Workflow: Draft posts matching your authentic voice, saved as drafts for review

Requirements

• Claude Code or compatible agent
• telegram-mcp server configured
• Telegram API credentials from https://my.telegram.org

Installation

1. Clone this repo:

git clone https://github.com/BayramAnnakov/telegram-assistant-skill.git

2. Symlink to Claude skills directory:

ln -s /path/to/telegram-assistant-skill ~/.claude/skills/telegram-assistant

3. Restart Claude Code

4. Verify the skill is detected:

Ask Claude: "use the telegram-assistant skill to show my unread messages"

Usage

Digest Workflow

Triggers: “telegram digest”, “unread messages”, “morning summary”

"Show me a digest of my unread Telegram messages"
"Summarize what I missed in Telegram today"

Style Extraction

Triggers: “analyze writing style”, “extract style from channel”

"Analyze the writing style of my ProductsAndStartups channel"
"Extract style patterns from @cryptoEssay"

Post Workflow

Triggers: “post to channel”, “draft telegram post”, “write for channel”

"Draft a post about MCP for my channel using my style"
"Write a channel post about AI agents"

Safety

This skill follows a draft-first policy:

• Never sends messages directly via send_message
• Always uses save_draft so you can review in Telegram before sending
• Drafts appear in the chat input field in your Telegram app

Dependencies

This skill works best with the following PRs merged into telegram-mcp:

• PR #45: Draft management tools (save_draft, get_drafts, clear_draft)
• PR #46: Fix unread detection (unread_mark flag)
• PR #47: macOS Keychain support for secure credential storage

Until merged, you can use the fork with these features.

File Structure

telegram-assistant/
├── SKILL.md              # Main skill file (AgentSkills.io compliant)
├── README.md             # This file
├── LICENSE               # MIT License
└── references/
    ├── setup.md          # Installation guide for telegram-mcp
    └── style-guide.md    # Generated style guide (per-channel)

AgentSkills.io Compliance

This skill follows the AgentSkills.io open standard:

• Valid YAML frontmatter with required fields
• Trigger-based activation (not slash commands)
• Clear workflow documentation

License

MIT

Author