tech-news-digest

• Daily or weekly tech briefings: You want a structured, ranked summary of what happened across AI, crypto, and open-source—without manually checking dozens of sources.
• Discord/email newsletter automation: You need a repeatable cron-driven workflow that posts digests to a Discord channel or delivers formatted HTML emails on a schedule.
• Tracking KOLs and GitHub activity: You follow specific Twitter/X accounts (AI researchers, founders) and want their posts alongside GitHub release notes and trending repos in one feed.
• Custom source monitoring: You want to monitor your own RSS feeds, subreddits, or GitHub repos and have results quality-scored and deduplicated against previous digests.

• Real-time alerting: This skill runs in batch/scheduled mode; it is not designed for instant push notifications when breaking news occurs.
• Non-tech domains: Sources and topic definitions are pre-configured around AI, crypto, and frontier tech. Adapting it to unrelated domains (e.g., sports, finance) requires substantial config rewriting.
• Environments without Python 3.8+: The pipeline requires Python 3 to run; it cannot operate in purely browser-based or no-code environments.

The run-pipeline.py script fetches from RSS feeds, Twitter/X KOLs, GitHub releases, GitHub Trending repos, Reddit, and web search simultaneously using parallel workers. All 6 fetch steps run concurrently and complete in approximately 30 seconds.

merge-sources.py scores every article using engagement metrics and source priority, then deduplicates by title similarity (85% threshold) and penalizes items already seen in previous digests. Output is topic-grouped and sorted by score.

Ships with 62 RSS feeds (AI labs, tech blogs, crypto media), 48 Twitter/X KOLs, 28 GitHub repos, 13 subreddits, and 4 web search topic queries—covering LLM, AI agents, crypto, and frontier tech with no setup required.

Includes ready-made templates for Discord (mobile-optimized, 2000-char aware), HTML email (with executive summary and archive links), and styled A4 PDF (Noto Sans CJK SC, emoji headers, page numbers via weasyprint).

Twitter collection supports twitterapi.io (preferred, ~$3–5/mo) or the official X API v2 bearer token, with auto-detection. Web search supports Brave or Tavily with automatic rate-limit-aware switching between parallel and sequential query modes.

Custom sources.json and topics.json files in workspace/config/ override or extend defaults without modifying the skill bundle. Add new sources, disable noisy ones, or redefine topic search queries and display limits.

Schedule a daily cron job referencing digest-prompt.md to run the full pipeline over the past 48 hours, then post a bullet-formatted summary to a Discord channel—covering top LLM news, trending GitHub repos, and KOL tweets.

Generate a 7-day rolling digest with 10–15 items per section, enrich top articles with full text, and deliver a styled HTML email with an optional PDF attachment using the built-in email template and send-email.py.

Use fetch-github.py to track releases from projects like LangChain, vLLM, DeepSeek, and Llama over the past week, scored and deduplicated against prior digests, then surface only genuinely new releases.

Run source-health.py on pipeline outputs to track per-source success and failure rates over 7 days, identify feeds or API endpoints with >50% failure rates, and clean up or replace unreliable sources.