vincelele/ai-fomo-skills

This project is a set of skills for AI coding agents (Codex, Claude Code, OpenClaw, Cursor, and similar tools that load skill or rule files) aimed at people who feel buried by the flood of AI news. The author's framing is that the real problem is not access to information but the judgment to decide what is worth reading. The README is in Chinese, and the project is positioned against generic AI daily digests or prompt collections. There are three skills in the repo. The first, ai-fomo-init, asks the user about their background, current goals, and personal criteria, so later judgments can be made in context. The second, ai-fomo-sources, imports content from blogs, RSS feeds, GitHub repos, GitHub Trending, X, podcasts, transcripts, or pasted text, and saves each item as a raw snapshot for later review. The third, ai-fomo, processes those snapshots and splits them into three buckets: write now (save into the personal knowledge base), ask first (needs a question back), and skip (not worth long-term storage). The intended users are AI founders, product managers, and independent developers who already use agent tools and want to spend less attention on low-value content. Over time, the workspace acts as a personal AI knowledge base that the agent re-reads and updates, rather than a static bookmark folder. The README is clear that this is not a SaaS product and not a way to scrape and republish other people's content. A notable piece is integration with Xiaoyuzhou, a Chinese podcast app. The skills can install a local bridge, log in via SMS code, batch-import episodes from a user's subscription inbox, scrape episode comments, and run audio transcription through DashScope's Fun-ASR API. All such steps support a dry-run mode, and the README has detailed guidance about not committing real workspace data, tokens, or transcripts back to the repo. Installation is meant for non-coders: the suggested path is to paste a short Chinese instruction into the agent and ask it to install the three skills. Command-line users get a Codex skill installer script or a git clone plus copy. The status is alpha, scheduled collection and full subscription export are listed as missing, and the project is MIT licensed.