← ClaudeAtlas

telegram-digestlisted

Builds a digest of new posts from open Telegram channels and publishes it as an artifact — no API key, bot, or account login required. In Claude Cowork it saves a config file and remembers channels between runs, so it can be scheduled to run automatically. In a regular chat it works as a one-off: no memory between sessions. Use for requests like 'make a digest', 'set up a Telegram channel digest', 'what's new in the channels', 'summarize Telegram channels'.
tatianahub/telegram-digest-skill · ★ 1 · AI & Automation · score 69
Install: claude install-skill tatianahub/telegram-digest-skill
# Telegram Channel Digest Reads new posts from public channels, builds a digest out of them, and remembers where it stopped so it doesn't repeat itself next time. ## 0. First run If the config (see section 1) isn't found yet, this is the first run. **Step 1 — check whether you can save a real file.** If you have working file tools (Cowork: `bash_tool`/`create_file`), you can remember channels between runs — go to step 2. If not (a regular chat, no file access), say so plainly: this run will be a one-off, nothing will be remembered after this conversation ends, and there's no point asking where to store a config. Then just ask for channels and language and build a single digest — skip the rest of this section. **Step 2 — set up persistent tracking (Cowork only):** 1. Ask for the list of channels: usernames without `@` and without `https://t.me/`, one per line or comma-separated. 2. Ask which language to write the digest in — offer the language the user is writing in as the default, but let them pick something else. 3. For each channel: `WebFetch https://t.me/s/<channel>/99999`, grab the maximum id from the page (no need to analyze the posts yet). 4. Write the config file (see section 1 for format) to the working directory. 5. State explicitly: "Setup done — tracking N channels from this point on, digest language: <language>. The first digest will be built on the next run" — don't try to build a digest from history that doesn't exist. 6. If the user wants this to run auto