com um clique
daily-ai-news-digest
// Fetches articles from 92 Karpathy-curated RSS feeds, scores them with an LLM, selects the top 3, and delivers a formatted digest to Telegram every morning.
// Fetches articles from 92 Karpathy-curated RSS feeds, scores them with an LLM, selects the top 3, and delivers a formatted digest to Telegram every morning.
| name | daily-ai-news-digest |
| description | Fetches articles from 92 Karpathy-curated RSS feeds, scores them with an LLM, selects the top 3, and delivers a formatted digest to Telegram every morning. |
| trigger | 0 8 * * * |
| metadata | {"openclaw":{"requires":{"env":["TELEGRAM_BOT_TOKEN","TELEGRAM_CHAT_ID"],"bins":["py"]},"primaryEnv":"TELEGRAM_BOT_TOKEN","emoji":["📰"]}} |
A morning briefing skill that monitors 92 RSS feeds curated by Andrej Karpathy, surfaces the most important AI and tech stories from the last 24 hours, and delivers a clean digest to Telegram.
Each morning the skill wakes up, pulls fresh articles from a hand-picked list of RSS feeds, uses an LLM to score and rank them by relevance and significance, then writes a concise summary of the top 3 stories and pushes it to a Telegram chat.
92 RSS feeds
│
▼
fetch_rss.py — parallel feed fetching
│
▼
filter — keep only articles published in the last 24 hours
│
▼
LLM scoring — rate each article for relevance, novelty, and impact
│
▼
select top 3 — ranked by score, deduplicated by URL
│
▼
summarise — one-paragraph summary per article via LLM
│
▼
categorise — group into themes (Breaking, Important, Notable)
│
▼
Telegram — formatted message delivered to configured chat
| Variable | Description |
|---|---|
TELEGRAM_BOT_TOKEN | Token for the Telegram bot that sends the digest |
TELEGRAM_CHAT_ID | ID of the Telegram chat or channel to deliver to |
Copy .env.example to .env and fill in these values before running.
The skill runs on the cron expression 0 8 * * * — every day at 08:00 UTC.
To change the time, update the trigger field in the frontmatter above. For example:
"0 7 * * 1-5" — weekdays at 07:00 UTC"0 6 * * *" — daily at 06:00 UTC# Install dependencies
pip install -r requirements.txt
# Run the full pipeline
python skill.py
# Run only the RSS fetch step
python scripts/fetch_rss.py