Skip to content

SKILL.md

Latest commit

 

History

History
36 lines (27 loc) · 1.62 KB

SKILL.md

File metadata and controls

36 lines (27 loc) · 1.62 KB

ETH24 - Daily Topic Digest

You are running ETH24, a daily digest tool that surfaces the top tweets for a configured topic.

Pipeline

  1. Crawl - Run python3 crawl.py to fetch tweets via Grok x_search (contextual discovery) and X API v2 (keyword search with engagement metrics). Requires X_BEARER_TOKEN, XAI_API_KEY, and optionally ANTHROPIC_API_KEY environment variables. Output: output/YYYY-MM-DD/crawled.json

  2. Rank - Read the crawled data from output/YYYY-MM-DD/crawled.json. Select up to 10 tweets by ecosystem importance. Filter out spam (airdrop scams, engagement farming, hashtag spam). Write one-line commentary for each. On quiet days, include fewer stories. If nothing clears the bar, return 0 stories.

  3. Output - Save the ranked data to output/YYYY-MM-DD/ranked.json. Default mode (cli) prints plain text to stdout and saves cli.txt. Tweet mode formats a post preview and saves thread.txt.

Ranking Guidelines

  • Read config.json for topic, brand, voice, and search terms
  • Commentary: 1-2 short sentences. Tell the reader why this matters. Don't restate the tweet.
  • Be accurate. Don't claim "first" or "biggest" unless certain.
  • No emojis. No emdashes. Use hyphens.
  • Include only stories that are genuinely important. Fewer is better than filler.
  • Write "highlights": a comma-separated preview of the day's biggest stories (under 200 chars).

Output Format

{
  "stories": [
    {
      "commentary": "One sentence.",
      "tweet_url": "https://x.com/handle/status/ID",
      "handle": "handle"
    }
  ],
  "highlights": "Story A, Story B, Story C",
  "date_label": "M/D/YY"
}